This article presents an approach for the automatic semantic segmentation of floorplan images, predicting room boundaries (walls, doors, windows) and semantic labels of room types. A multi-task network was designed to represent and learn inherent dependencies by combining a Convolutional Neural Network to generate suitable features with a Graph Convolutional Network (GCN) to capture long-range dependencies. In particular, a Self-Constructing Graph module is applied to automatically induce an input graph for the GCN. Experiments on different datasets demonstrate the superiority and effectiveness of the multi-task network compared to state-of-the-art methods. The accurate results not only allow for subsequent vectorization of the existing floorplans but also for automatic inference of layout graphs including connectivity and adjacency relations. The latter could serve as basis to automatically sample layout graphs for architectural planning and design, predict missing links for unobserved parts for as-built building models and learn important latent topological and architectonic patterns.

doi:10.1016/j.autcon.2024.105649

@article{Knechtel2024FloorplanSCG,
    abstract = {This article presents an approach for the automatic semantic segmentation of floorplan images, predicting room boundaries (walls, doors, windows) and semantic labels of room types. A multi-task network was designed to represent and learn inherent dependencies by combining a Convolutional Neural Network to generate suitable features with a Graph Convolutional Network (GCN) to capture long-range dependencies. In particular, a Self-Constructing Graph module is applied to automatically induce an input graph for the GCN. Experiments on different datasets demonstrate the superiority and effectiveness of the multi-task network compared to state-of-the-art methods. The accurate results not only allow for subsequent vectorization of the existing floorplans but also for automatic inference of layout graphs including connectivity and adjacency relations. The latter could serve as basis to automatically sample layout graphs for architectural planning and design, predict missing links for unobserved parts for as-built building models and learn important latent topological and architectonic patterns.},
    author = {Knechtel, Julius and Rottmann, Peter and Haunert, Jan-Henrik and Dehbi, Youness},
    doi = {10.1016/j.autcon.2024.105649},
    journal = {Automation in Construction},
    pages = {105649},
    title = {Semantic Floorplan Segmentation Using Self-Constructing Graph Networks},
    url = {https://www.sciencedirect.com/science/article/pii/S0926580524003856},
    volume = {166},
    year = {2024}
}

Immersive virtual reality (IVR) allows viewing abstract concepts and entities in a three dimensional (3D) visuospatial environment. In this paper, we innovatively introduced IVR technology into the verification of the as-built state of electric line networks in buildings. On the one hand, using a reasoning-based estimation of electric networks as a starting point, we demonstrated the usability of IVR technology for verifying installed utilities in buildings. On the other hand, we established the communication between the Reasoner and the practitioner and also simulated the verification action of electric line networks in buildings in the real world. The principal findings of this work pave the way for a subsequent and systematic evaluation of the different reasoning strategies for estimating and generating the as-built state of building utilities.

doi:10.1007/978-3-031-43699-4_8

@inproceedings{knechtel2024immersiveVRElectricNetworks,
    abstract = {Immersive virtual reality (IVR) allows viewing abstract concepts and entities in a three dimensional (3D) visuospatial environment. In this paper, we innovatively introduced IVR technology into the verification of the as-built state of electric line networks in buildings. On the one hand, using a reasoning-based estimation of electric networks as a starting point, we demonstrated the usability of IVR technology for verifying installed utilities in buildings. On the other hand, we established the communication between the Reasoner and the practitioner and also simulated the verification action of electric line networks in buildings in the real world. The principal findings of this work pave the way for a subsequent and systematic evaluation of the different reasoning strategies for estimating and generating the as-built state of building utilities.},
    address = {Cham},
    author = {Knechtel, Julius and Li, Weilian and Orgeig, Yannick and Haunert, Jan-Henrik and Dehbi, Youness},
    booktitle = {Recent Advances in 3D Geoinformation Science (Proceedings of the 18th 3D GeoInfo Conference 2023, Munich)},
    doi = {10.1007/978-3-031-43699-4\_8},
    editor = {Kolbe, Thomas H. and Donaubauer, Andreas and Beil, Christof},
    isbn = {978-3-031-43699-4},
    pages = {129--143},
    publisher = {Springer Nature Switzerland},
    title = {{I}mmersive {V}irtual {R}eality to {V}erify the {A}s-built {S}tate of {E}lectric {L}ine {N}etworks in {B}uildings},
    year = {2024}
}

Knowledge on the utilities hidden in the wall, e.g., electric lines or water pipes, is indispensable for work safety and valuable for planning. Since most of the existing building stock originates from the pre-digital era, no models as understood for Building Information Modeling (BIM) exist. To generate these models often labor-intensive procedures are necessary; however, recent research has dealt with the efficient generation and verification of a building’s electric network. In this context, a reliable measurement method is a necessity. In this paper we test different measurement techniques, such as point-wise measurements with hand-held devices or area-based techniques utilizing thermal imaging. For this purpose, we designed and built a simulation environment that allows various parameters to be manipulated under controlled conditions. In this scenario the low-cost handheld devices show promising results, with a precision between 92% and 100% and a recall between 89% and 100%. The expensive thermal imaging camera is also able to detect electric lines and pipes if there is enough power on the line or if the temperature of the water in the pipe and the environment’s temperature are sufficiently different. Nevertheless, while point-wise measurements can directly yield results, the thermal camera requires post-processing in specific analysis software. The results reinforce the idea of using reasoning methods in both the do-it-yourself and commercial sector, to rapidly gather information about hidden installations in a building without prior technical knowledge. This paves the way for, e.g., exploring the possibilities of an implementation and presentation in augmented reality (AR).

doi:10.5194/isprs-annals-X-1-W1-2023-263-2023

@article{knechtel2023sensorSuitability,
    abstract = {Knowledge on the utilities hidden in the wall, e.g., electric lines or water pipes, is indispensable for work safety and valuable for planning. Since most of the existing building stock originates from the pre-digital era, no models as understood for Building Information Modeling (BIM) exist. To generate these models often labor-intensive procedures are necessary; however, recent research has dealt with the efficient generation and verification of a building’s electric network. In this context, a reliable measurement method is a necessity. In this paper we test different measurement techniques, such as point-wise measurements with hand-held devices or area-based techniques utilizing thermal imaging. For this purpose, we designed and built a simulation environment that allows various parameters to be manipulated under controlled conditions. In this scenario the low-cost handheld devices show promising results, with a precision between 92\% and 100\% and a recall between 89\% and 100\%. The expensive thermal imaging camera is also able to detect electric lines and pipes if there is enough power on the line or if the temperature of the water in the pipe and the environment’s temperature are sufficiently different. Nevertheless, while point-wise measurements can directly yield results, the thermal camera requires post-processing in specific analysis software. The results reinforce the idea of using reasoning methods in both the do-it-yourself and commercial sector, to rapidly gather information about hidden installations in a building without prior technical knowledge. This paves the way for, e.g., exploring the possibilities of an implementation and presentation in augmented reality (AR).},
    author = {Knechtel, Julius and Behmann, Jan and Haunert, Jan-Henrik and Dehbi, Youness},
    doi = {10.5194/isprs-annals-X-1-W1-2023-263-2023},
    journal = {ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
    pages = {263--270},
    title = {{S}uitability {A}ssessment of {D}ifferent {S}ensors to {D}etect {H}idden {I}nstallations for {A}s-built {BIM}},
    url = {https://isprs-annals.copernicus.org/articles/X-1-W1-2023/263/2023/},
    volume = {X-1/W1-2023},
    year = {2023}
}

doi:10.1145/3615891.3628012

@inproceedings{inproceedings,
    address = {Hamburg, Germany},
    author = {Bonerath, Annika and Orgeig, Yannick and Haunert, Jan-Henrik and Dehbi, Youness},
    booktitle = {Proceedings of the 6th ACM SIGSPATIAL International Workshop on GeoSpatial Simulation},
    doi = {10.1145/3615891.3628012},
    month = {11},
    pages = {19-22},
    title = {Integrating Optimization-Based Spatial Unit Allocation into a Multi-Agent Model for the Simulation of Urban Growth},
    year = {2023}
}

doi:10.1145/3615899.3627931

@inproceedings{arzoumanidis2023catchment,
    address = {Hamburg, Germany},
    author = {Lukas Arzoumanidis and Axel Forsch and Jan-Henrik and Youness Dehbi},
    booktitle = {Proc. 1st ACM SIGSPATIAL Workshop on Sustainable Mobility},
    doi = {10.1145/3615899.3627931},
    title = {Catchment cell visualization for multi-modal public transportation networks},
    year = {2023}
}

Abstract Detecting and collecting public opinion via social media can provide near real-time information to decision-makers, which plays a vital role in urban disaster management and sustainable development. However, there has been little work focusing on identifying the perception and the sentiment polarity expressed by users during and after disasters, particularly regional flood events. In this article, we comprehensively analyze tweets data related to the “European floods in 2021” over time, topic, and sentiment, forming a complete workflow from data processing, topic modeling, sentiment analysis, and topic and sentiment prediction. The aim is to address the following research questions: (1) What are the public perception and main concerns during and after floods? (2) How does the public sentiment change during and after floods? Results indicate that there is a significant correlation between a flood's trend and the heat of corresponding tweets. The three topics that receive the most public concern are: (1) climate change and global warming; (2) praying for the victims: and (3) disaster situations and information. Negative sentiments are predominant during the floods and will continue for some time. We tested five different classifiers, of which TextCNN-attention turned out to deliver the best predictions in topic and sentiment prediction, and performed well for sparse flood tweets, it can be used to predict the topic and sentiment polarity of a single tweet in real-time during the flood events. Our findings can help disaster agencies to better understand the dynamics of social networks and develop stronger situational awareness towards a disaster, which can contribute to scientifically justified decision-making in urban risk management and also meet the challenges associated with the global sustainable development goal 11 (SDGs) on Sustainable Cities and Communities.

doi:https://doi.org/10.1111/tgis.13097

@article{li2023socialMediaDisaster,
    abstract = {Abstract Detecting and collecting public opinion via social media can provide near real-time information to decision-makers, which plays a vital role in urban disaster management and sustainable development. However, there has been little work focusing on identifying the perception and the sentiment polarity expressed by users during and after disasters, particularly regional flood events. In this article, we comprehensively analyze tweets data related to the “European floods in 2021” over time, topic, and sentiment, forming a complete workflow from data processing, topic modeling, sentiment analysis, and topic and sentiment prediction. The aim is to address the following research questions: (1) What are the public perception and main concerns during and after floods? (2) How does the public sentiment change during and after floods? Results indicate that there is a significant correlation between a flood's trend and the heat of corresponding tweets. The three topics that receive the most public concern are: (1) climate change and global warming; (2) praying for the victims: and (3) disaster situations and information. Negative sentiments are predominant during the floods and will continue for some time. We tested five different classifiers, of which TextCNN-attention turned out to deliver the best predictions in topic and sentiment prediction, and performed well for sparse flood tweets, it can be used to predict the topic and sentiment polarity of a single tweet in real-time during the flood events. Our findings can help disaster agencies to better understand the dynamics of social networks and develop stronger situational awareness towards a disaster, which can contribute to scientifically justified decision-making in urban risk management and also meet the challenges associated with the global sustainable development goal 11 (SDGs) on Sustainable Cities and Communities.},
    author = {Li, Weilian and Haunert, Jan-Henrik and Knechtel, Julius and Zhu, Jun and Zhu, Qing and Dehbi, Youness},
    doi = {https://doi.org/10.1111/tgis.13097},
    journal = {Transactions in GIS},
    number = {6},
    pages = {1766-1793},
    title = {Social {M}edia {I}nsights on {P}ublic {P}erception and {S}entiment {D}uring and {A}fter {D}isasters: {T}he {E}uropean {F}loods in 2021 as a {C}ase {S}tudy},
    volume = {27},
    year = {2023}
}

doi:10.5194/ica-abs-6-11-2023

@article{arzoumanidis2023SCGHistMaps,
    author = {Arzoumanidis, Lukas and Knechtel, Julius and Haunert, Jan-Henrik and Dehbi, Youness},
    doi = {10.5194/ica-abs-6-11-2023},
    journal = {Abstracts of the ICA},
    pages = {11},
    title = {{S}elf-{C}onstructing {G}raph {C}onvolutional {N}etworks for {S}emantic {S}egmentation of {H}istorical {M}aps},
    volume = {6},
    year = {2023}
}

This article addresses the augmentation of existing building models by a-priori not observable structures such as electric installations. The aim is to unambiguously determine an electric network in an incremental manner with a minimum number of local measurements, e.g. using wire detectors, by suggesting the next measurement. Different reasoning strategies, e.g. utilizing graph-theoretical algorithms, have been presented and tested based on a hypothesis which is generated using Mixed Integer Linear Programming (MILP) while incorporating standards regarding the installation of electric wiring and findings from previous measurements. The presented method has been successfully applied on simulated and real-world buildings, it saves up to 80% of the necessary measurements compared to an exhaustive verification of the whole existing electric network, and paves the way for efficiently extending existing models, e.g. GIS models, with information on hidden utilities. This opens up new opportunities to model further infrastructures, e.g. water pipes, in future research.

doi:10.1016/j.autcon.2022.104571

@article{dehbi2022incrementalConstraint-based,
    abstract = {This article addresses the augmentation of existing building models by a-priori not observable structures such as electric installations. The aim is to unambiguously
determine an electric network in an incremental manner with a minimum number of local measurements, e.g. using wire detectors, by suggesting the next measurement.
Different reasoning strategies, e.g. utilizing graph-theoretical algorithms, have been presented and tested based on a hypothesis which is generated using Mixed Integer
Linear Programming (MILP) while incorporating standards regarding the installation of electric wiring and findings from previous measurements. The presented
method has been successfully applied on simulated and real-world buildings, it saves up to 80% of the necessary measurements compared to an exhaustive verification
of the whole existing electric network, and paves the way for efficiently extending existing models, e.g. GIS models, with information on hidden utilities. This opens
up new opportunities to model further infrastructures, e.g. water pipes, in future research.},
    author = {Dehbi, Youness and Knechtel, Julius and Niedermann, Benjamin and Haunert, Jan-Henrik},
    doi = {10.1016/j.autcon.2022.104571},
    journal = {Automation in Construction},
    pages = {104571},
    title = {Incremental Constraint-based Reasoning for Estimating As-built Electric Line Routing in Buildings},
    volume = {143},
    year = {2022}
}

The objective of disaster scenes is to share location-based risk information to a large audience in an effective and intuitive way. However, current studies on three-dimensional (3D) representation for dam-break floods have the following limitations: (1) they are lacking a reasonable logic to organize the whole process of dam-break floods, (2) they present information in a way that cannot be easily understood by laypersons. Geospatial storytelling helps to create exciting experiences and to explain complex relationships of geospatial phenomena. This article proposes a three-dimensional virtual representation method for the whole process of dam-break floods from a geospatial storytelling perspective. The creation of a storyline and a storytelling-oriented representation of dam-break floods are discussed in detail. Finally, a prototype system based on WebGL is developed to conduct an experiment analysis. The results of the experiment show that the proposed method can effectively support 3D representation of the spatiotemporal process of dam-break floods. Furthermore, the statistical results indicate that the storytelling is useful for assisting participants in understanding the occurrence and development of dam-break floods, and is applicable to the popularization of disaster science for the general public.

doi:10.1080/17538947.2022.2118877

@article{li2022storytelling,
    abstract = {The objective of disaster scenes is to share location-based risk information to a large audience in an effective and intuitive way.
However, current studies on three-dimensional (3D) representation for dam-break floods have the following limitations: (1) they are lacking a reasonable logic to organize the whole process of dam-break floods,
(2) they present information in a way that cannot be easily understood by laypersons. Geospatial storytelling helps to create exciting experiences and to explain complex relationships of geospatial phenomena.
This article proposes a three-dimensional virtual representation method for the whole process of dam-break floods from a geospatial storytelling perspective.
The creation of a storyline and a storytelling-oriented representation of dam-break floods are discussed in detail. Finally, a prototype system based on WebGL is developed to conduct an experiment analysis.
The results of the experiment show that the proposed method can effectively support 3D representation of the spatiotemporal process of dam-break floods.
Furthermore, the statistical results indicate that the storytelling is useful for assisting participants in understanding the occurrence and development of
dam-break floods, and is applicable to the popularization of disaster science for the general public.},
    author = {Weilian, L. and Jun, Z. and Haunert, J.-H. and Lin, F. and Qing, Z. and Dehbi, Y.},
    doi = {10.1080/17538947.2022.2118877},
    journal = {International Journal of Digital Earth},
    pages = {1637--1656},
    title = {Three-dimensional virtual representation for the whole process of dam-break floods from a geospatial storytelling perspective},
    volume = {15},
    year = {2022}
}

Building footprints are a prerequisite for many tasks such as urban mapping and planning. Such structures are mostly derived using airborne laser scanning which reveals rather roof structures than the underlying hidden footprint boundary. This paper introduces an approach to extract a 2D building boundary from a 3D point cloud stemming from either terrestrial scanning or via close-range sensing using a mobile platform, e.g. drone. To this end, a pipeline of methods including non-parametric kernel density estimation (KDE) of an underlying probability density function, a solution of the Travelling Salesperson Problem (TSP), outlier elimination and line segmentation are presented to extract the underlying building footprint. KDE turns out to be suitable to automatically determine a horizontal cut in the point cloud. An ordering of the resulting points in this cut using a shortest possible tour based on TSP allows for the application of existing line segmentation algorithms, otherwise dedicated to indoor segmentation. Outliers in the resulting segments are removed using DBSCAN. The segments are then generalized leading to the final footprint geometry. We applied our approach on real-world examples and achieved an IoU between 0.930 and 0.998 assessed by ground truth footprints from both authoritative and VGI data.

doi:10.5194/isprs-annals-X-4-W2-2022-233-2022

@article{rottmann2022AutomaticBuilding,
    abstract = {Building footprints are a prerequisite for many tasks such as urban mapping and planning. Such structures are mostly derived using
airborne laser scanning which reveals rather roof structures than the underlying hidden footprint boundary. This paper introduces an
approach to extract a 2D building boundary from a 3D point cloud stemming from either terrestrial scanning or via close-range sensing
using a mobile platform, e.g. drone. To this end, a pipeline of methods including non-parametric kernel density estimation (KDE)
of an underlying probability density function, a solution of the Travelling Salesperson Problem (TSP), outlier elimination and line
segmentation are presented to extract the underlying building footprint. KDE turns out to be suitable to automatically determine a
horizontal cut in the point cloud. An ordering of the resulting points in this cut using a shortest possible tour based on TSP allows for
the application of existing line segmentation algorithms, otherwise dedicated to indoor segmentation. Outliers in the resulting segments
are removed using DBSCAN. The segments are then generalized leading to the final footprint geometry. We applied our approach on
real-world examples and achieved an IoU between 0.930 and 0.998 assessed by ground truth footprints from both authoritative and
VGI data.},
    author = {Peter Rottmann and Jan-Henrik Haunert and Youness Dehbi},
    doi = {10.5194/isprs-annals-X-4-W2-2022-233-2022},
    journal = {ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
    pages = {233--240},
    title = {Automatic Building Footprint extraction from 3D laserscans},
    volume = {X-4/W2-2022},
    year = {2022}
}

doi:10.5194/ica-abs-5-89-2022

@article{korir2022modalsplit,
    address = {Vienna},
    author = {Korir, Victor and Forsch, Axel and Dehbi, Youness and Haunert, Jan-Henrik},
    doi = {10.5194/ica-abs-5-89-2022},
    journal = {Abstracts of the ICA},
    month = {sep},
    pages = {89},
    title = {Visualizing the Modal Split in Public Transportation Networks},
    volume = {5},
    year = {2022}
}

Terrestrial laser scanning has become more and more popular in recent years. The according planning of the standpoint network is a crucial issue influencing the overhead and the resulting point cloud. Fully static approaches are both cost and time extensive, whereas fully kinematic approaches cannot produce the same data quality. Stop-and-go scanning, which combines the strengths of both strategies, represents a good alternative solution. In the scanning process, the standpoint planning is by now mostly a manual process based on expert knowledge and relying on the surveyor's experience. This paper provides a method based on Mixed Integer Linear Programming (MILP) ensuring an optimal placement of scanner standpoints considering all scanner-related constraints (e.g. incidence angle), a full coverage of the scenery, a sufficient overlap for the subsequent registration and an optimal route planning solving a Traveling Salesperson Problem (TSP). This enables the fully automatic application of autonomous systems for providing a complete model while performing a stop-and-go laser scanning, e.g. with the Spot robot from Boston Dynamics. Our pre-computed solution, i.e. standpoints and trajectory, has been evaluated surveying a real-world environment and successfully compared with a precise LoD2 building model of the underlying scene. The performed ICP-based registration issued from our fully automatic pipeline turns out to be a very good and safe alternative of the otherwise laborious target-based registration.

doi:10.5194/isprs-annals-V-4-2022-129-2022

@article{knechtel2022OptimalPositionPath,
    abstract = {Terrestrial laser scanning has become more and more popular in recent years.
The according planning of the standpoint network is a crucial issue influencing the overhead and the resulting point cloud. Fully static approaches are both cost and time extensive, whereas fully kinematic approaches cannot produce the same data quality.
Stop-and-go scanning, which combines the strengths of both strategies, represents a good alternative solution. In the scanning process, the standpoint planning is by now mostly a manual process based on expert knowledge and relying on the surveyor's experience. This paper provides a method based on Mixed Integer Linear Programming (MILP) ensuring an optimal placement of scanner standpoints considering all scanner-related constraints (e.g. incidence angle),
a full coverage of the scenery, a sufficient overlap for the subsequent registration and an optimal route planning solving a Traveling Salesperson Problem (TSP). This enables the fully automatic application of autonomous systems for providing a complete model while performing a stop-and-go laser scanning, e.g. with the Spot robot from Boston Dynamics.
Our pre-computed solution, i.e. standpoints and trajectory, has been evaluated surveying a real-world environment and successfully compared with a precise LoD2 building model of the underlying scene. The performed ICP-based registration issued from our fully automatic pipeline turns out to be a very good and safe alternative of the otherwise laborious target-based registration.},
    author = {Knechtel, Julius and Klingbeil, Lasse and Haunert, Jan-Henrik and Dehbi, Youness},
    doi = {10.5194/isprs-annals-V-4-2022-129-2022},
    journal = {ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
    pages = {129--136},
    title = {Optimal position and path planning for stop-and-go laserscanning for the acquisition of 3D building models},
    url = {https://www.isprs-ann-photogramm-remote-sens-spatial-inf-sci.net/V-4-2022/129/2022/},
    volume = {V-4-2022},
    year = {2022}
}

The complexity of disasters creates a significant challenge in the knowledge acquisition of the public. With the development of geospatial technologies, maps, geographic information science (GIS), and virtual geographic environment (VGE) are widely used to represent disaster information and help the public better understand disaster risk. However, the application, design, and specific challenges have not been investigated comprehensively in disaster information representation thus far. This article presents the weaknesses and strengths of the existing methods for representing disaster information in the last decades, and then the authors contribute some basic ideas for efficient disaster knowledge communication. The objective of this paper is to provide a clear image that improves users’ understanding of disaster information and bridge the communication gaps in disaster management. Finally, we suggest readers applying further creative thinking strategies to address the challenges associated with communicating disaster knowledge.

doi:10.1111/tgis.12922

@article{dehbi2020improving,
    abstract = {The complexity of disasters creates a significant challenge in the knowledge acquisition of the public.
With the development of geospatial technologies, maps, geographic information science (GIS), and virtual geographic environment (VGE) are widely used to represent disaster information and help the public better
understand disaster risk. However, the application, design, and specific challenges have not been investigated comprehensively in disaster information representation thus far.
This article presents the weaknesses and strengths of the existing methods for representing disaster information in the last decades, and then the authors contribute some basic ideas for efficient disaster knowledge communication.
The objective of this paper is to provide a clear image that improves users’ understanding of disaster information and bridge the communication gaps in disaster management.
Finally, we suggest readers applying further creative thinking strategies to address the challenges associated with communicating disaster knowledge.},
    author = {Li, W. and Zhu, J. and Pirasteh, S. and Zhu, Q. and Fu, L. and Wu, J. and Hu, Y. and Dehbi, Y.},
    doi = {10.1111/tgis.12922},
    eprint = {https://onlinelibrary.wiley.com/doi/pdf/10.1111/tgis.12922},
    journal = {Transactions in GIS},
    number = {3},
    pages = {1376-1398},
    title = {Investigations of disaster information representation from a geospatial perspective: progress, challenges and recommendations},
    url = {https://onlinelibrary.wiley.com/doi/abs/10.1111/tgis.12922},
    volume = {26},
    year = {2022}
}

The positioning of laser scanners for indoor surveying is still a time and cost expensive process. This article proposes an optimization approach for computing an admissible sensor placement with the minimal number of sensor view point positions. The approach facilitates both wall and floor surveying based on a floorplan of the study object. Optimal solutions are calculated by solving an Integer Linear Program that respects manufacturer specifications incorporating constraints such as full coverage. To enable a subsequent co-registration of the scans, a flow-based constraint formulation ensuring the connectivity of the selected positions in an appropriately defined geometric intersection graph is introduced. The method has been evaluated on real-world objects and compared to heuristic methods that have frequently been used for related problems. Our solutions outperform heuristic approaches regarding both running time and the number of TLS stations. In a case study with a larger floorplan of an institute building and with different parameter settings, our method resulted in a solution with at least two stations less compared to a solution generated by an expert.

doi:10.1016/j.isprsjprs.2021.07.013

@article{dehbi2021optimalscan,
    abstract = {The positioning of laser scanners for indoor surveying is still a time and cost expensive process. This article proposes an optimization approach for computing an admissible sensor placement with the minimal number of sensor view point positions. The approach facilitates both wall and floor surveying based on a floorplan of the study object. Optimal solutions are calculated by solving an Integer Linear Program
that respects manufacturer specifications incorporating constraints such as full coverage. To enable a subsequent co-registration of the scans, a flow-based constraint formulation ensuring the connectivity of the selected positions in an appropriately defined geometric intersection graph is introduced. The method has been evaluated on real-world objects and compared to heuristic methods that have frequently been used for related problems. Our solutions outperform heuristic approaches regarding both running time and the number of TLS stations. In a case study with a larger floorplan of an institute building and with different parameter settings,
our method resulted in a solution with at least two stations less compared to a solution generated by an expert.},
    author = {Dehbi, Youness and Leonhardt, Johannes and Oehrlein, Johannes and Haunert, Jan-Henrik},
    doi = {10.1016/j.isprsjprs.2021.07.013},
    issn = {0924-2716},
    journal = {ISPRS Journal of Photogrammetry and Remote Sensing},
    pages = {103-116},
    title = {Optimal scan planning with enforced network connectivity for the acquisition of three-dimensional indoor models},
    url = {https://www.sciencedirect.com/science/article/pii/S0924271621001982},
    volume = {180},
    year = {2021}
}

Accurate and robust positioning of vehicles in urban environments is of high importance for autonomous driving or mobile mapping. In mobile mapping systems, a simultaneous mapping of the environment using laser scanning and an accurate positioning using GNSS is targeted. This requirement is often not guaranteed in shadowed cities where GNSS signals are usually disturbed, weak or even unavailable. We propose a novel approach which incorporates prior knowledge, i.e. 3D city model of the environment, and improves the trajectory. The recorded point cloud is matched with the semantic city model using a point-to-plane ICP. A pre-classification step enables an informed sampling of appropriate matching points. Random Forest is used as classifier to discriminate between facade and remaining points. Local inconsistencies are tackled by a segment-wise partitioning of the point cloud where an interpolation guarantees a seamless transition between the segments. The general

doi:10.1111/tgis.12719

@article{dehbi2020improving,
    abstract = {Accurate and robust positioning of vehicles in urban environments is of high importance for autonomous driving or mobile mapping. In mobile mapping systems, a simultaneous mapping of the environment using laser scanning and an accurate positioning using GNSS is targeted. This requirement is often not guaranteed in shadowed cities where GNSS signals are usually disturbed, weak or even unavailable. We propose a novel approach which incorporates prior knowledge, i.e. 3D city
model of the environment, and improves the trajectory. The recorded point cloud is matched with the semantic city model using a point-to-plane ICP. A pre-classification step enables an informed sampling of appropriate matching points. Random Forest is used as classifier to discriminate between facade and remaining points. Local inconsistencies are tackled by a segment-wise partitioning of the point cloud where an interpolation guarantees a seamless transition between the segments. The general},
    author = {Lucks, L. and Klingbeil, L. and Pl\"{u}mer, L. and Dehbi, Y.},
    doi = {10.1111/tgis.12719},
    journal = {Transactions in GIS},
    number = {1},
    pages = {238-260},
    title = {Improving Trajectory Estimation Using 3D City Models and Kinematic Point Clouds},
    volume = {25},
    year = {2021}
}

This article proposes a novel method for the 3D-reconstruction of LoD2 buildings from LIDAR data. We propose an active sampling strategy which applies a cascade of filters focusing on promising samples in an early stage, thus avoiding the pitfalls of RANSAC based approaches. Filters are based on prior knowledge represented by (non-parametric) density distributions. In our approach samples are pairs of surflets, i.e. 3D points together with normal vectors derived from a plane approximation of their neighborhood. Surflet pairs provide parameters for model candidates such as azimuth, inclination and ridge height, as well as parameters estimating internal precision and consistency. This provides a ranking of roof model candidates and leads to a small number of promising hypotheses. Building footprints are derived in a preprocessing step using machine learning methods, in particular Support Vector Machines (SVM).

doi:10.1111/tgis.12659

@article{dehbi2020robust,
    abstract = {This article proposes a novel method for the 3D-reconstruction of LoD2 buildings from LIDAR data. We propose an active sampling strategy which applies a cascade of filters focusing on promising samples in an early stage, thus avoiding the pitfalls of RANSAC based approaches. Filters are based on prior knowledge represented by (non-parametric) density distributions. In our approach samples are pairs of surflets, i.e. 3D points together with normal vectors derived from a plane approximation of
their neighborhood. Surflet pairs provide parameters for model candidates such as azimuth, inclination and ridge height, as well as parameters estimating internal precision and consistency. This provides a ranking of roof model candidates and leads to a small number of promising hypotheses. Building footprints are derived in a preprocessing
step using machine learning methods, in particular Support Vector Machines (SVM).},
    author = {Dehbi, Y. and Henn, A. and Gr\"{o}ger, G. and Stroh, V. and Pl\"{u}mer, L.},
    doi = {10.1111/tgis.12659},
    journal = {Transactions in GIS},
    number = {1},
    pages = {112-133},
    title = {Robust and Fast Reconstruction of Complex Roofs with Active Sampling from 3D Point Clouds},
    volume = {25},
    year = {2021}
}

Accurate reconstruction of roofs with dormers is challenging. Without careful separation of the dormer points from the points on the roof surface, the estimation of the roof areas is distorted. The characteristic distortion of the density distribution in comparison to the expected normal distribution is the starting point of our method. We propose a hierarchical method which improves roof reconstruction from LiDAR point clouds in a model-based manner separating dormer points from roof points using classification methods. The key idea is to exploit probability density functions (PDFs) to reveal roof properties and design skilful features for a supervised learning method using support vector machines (SVMs). The approach is tested based on real data as well as simulated point clouds.

doi:10.1111/tgis.12658

@article{dehbi2020probability,
    abstract = {Accurate reconstruction of roofs with dormers is challenging. Without careful separation of the dormer points from the points on the roof surface, the
estimation of the roof areas is distorted. The characteristic distortion of the density distribution in comparison to the expected normal distribution is the
starting point of our method. We propose a hierarchical method which improves roof reconstruction from LiDAR point clouds in a model-based manner
separating dormer points from roof points using classification methods. The key idea is to exploit probability density functions (PDFs) to reveal roof
properties and design skilful features for a supervised learning method using support vector machines (SVMs). The approach is tested based on real data
as well as simulated point clouds.},
    author = {Dehbi, Y. and Koppers, S. and Pl\"umer, L.},
    doi = {10.1111/tgis.12658},
    journal = {Transactions in GIS},
    number = {1},
    pages = {44-70},
    title = {Looking for a needle in a haystack: Probability density based classification and reconstruction of dormers from 3D point clouds},
    volume = {25},
    year = {2021}
}

The automatic generation of travel-time maps is a prerequisite for many fields of application such as tourist assistance and spatial decision support systems. The task is to determine outlines of zones that are reachable from a user's location in a given amount of time. In this work, we focus on travel-time maps with a formally guaranteed separation property in the sense that a zone exactly contains the part of the road network that is reachable within a pre-defined time from a given starting point and start time. In contrast to other automated methods, our approach generates schematized travel-time maps that reduce the visual complexity by representing each zone by an octilinear polygon. We aim at octilinear polygons with a small number of bends to further optimize the legibility of the map. The reachable parts of the road network are determined by the integration of timetable information for different modes of public transportation and pedestrian walkways based on a multi-modal time-expanded network. Moreover, the generated travel-time maps visualize multiple travel times using a map overlay of different time zones. In experiments on real-world data we compare our schematic visualizations to travel-time maps created with other visualization techniques.

doi:10.1111/tgis.12821

@article{forsch2021isochrones,
    abstract = {The automatic generation of travel-time maps is a prerequisite for many fields of application such as tourist assistance and spatial decision support systems. The task is to determine outlines of zones that are reachable from a user's location in a given amount of time. In this work, we focus on travel-time maps with a formally guaranteed separation property in the sense that a zone exactly contains the part of the road network that is reachable within a pre-defined time from a given starting point and start time. In contrast to other automated methods, our approach generates schematized travel-time maps that reduce the visual complexity by representing each zone by an octilinear polygon. We aim at octilinear polygons with a small number of bends to further optimize the legibility of the map. The reachable parts of the road network are determined by the integration of timetable information for different modes of public transportation and pedestrian walkways based on a multi-modal time-expanded network. Moreover, the generated travel-time maps visualize multiple travel times using a map overlay of different time zones. In experiments on real-world data we compare our schematic visualizations to travel-time maps created with other visualization techniques.},
    author = {Forsch, Axel and Dehbi, Youness and Niedermann, Benjamin and Oehrlein, Johannes and Rottmann, Peter and Haunert, Jan-Henrik},
    doi = {10.1111/tgis.12821},
    journal = {Transactions in GIS},
    number = {6},
    pages = {3233--3256},
    title = {Multimodal Travel-Time Maps with Formally Correct and Schematic Isochrones},
    volume = {25},
    year = {2021}
}

Unmanned Aerial Vehicles (UAVs) are used for the inspection of areas which are otherwise difficult to access. Autonomous monitoring and navigation requires a background knowledge on the surroundings of the vehicle. Most mission planing systems assume collision- free pre-defined paths and do not tolerate a GPS signal outage. Our approach makes weaker assumptions. This paper introduces a mission planing platform allowing for the integration of environmental prior knowledge such as 3D building and terrain models. This prior knowledge is integrated to pre-compute an octomap for collision detection. The semantically rich building models are used to specify semantic user queries such as roof or facade inspection. A reasoning process paves the way for semantic mission planing of hidden and a-priori unknown objects. Subsequent scene interpretation is performed by an incremental parsing process.

doi:10.5194/isprs-archives-XLIII-B1-2020-521-2020

@article{dehbi2020mission,
    abstract = {Unmanned Aerial Vehicles (UAVs) are used for the inspection of areas which are otherwise difficult to access. Autonomous monitoring and navigation requires a background knowledge on the surroundings of the vehicle. Most mission planing systems assume collision-
free pre-defined paths and do not tolerate a GPS signal outage. Our approach makes weaker assumptions. This paper introduces a mission planing platform allowing for the integration of environmental prior knowledge such as 3D building and terrain models. This
prior knowledge is integrated to pre-compute an octomap for collision detection. The semantically rich building models are used to specify semantic user queries such as roof or facade inspection. A reasoning process paves the way for semantic mission planing of
hidden and a-priori unknown objects. Subsequent scene interpretation is performed by an incremental parsing process.},
    author = {Dehbi, Y. and Klingbeil, L. and Pl\"umer, L.},
    doi = {10.5194/isprs-archives-XLIII-B1-2020-521-2020},
    journal = {ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
    pages = {521--526},
    series = {ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences (ISPRS Congress, Nice)},
    title = {UAV mission planning for automatic exploration and semantic mapping},
    volume = {XLIII-B1-2020},
    year = {2020}
}

3D city models in Level-of-Detail 2 (LoD2) are nowadays inevitable for many applications such as solar radiation calculation and energy demand estimation. City-wide models are required which can solely be acquired by fully automatic approaches. In this paper we propose a novel method for the 3D-reconstruction of LoD2 buildings with structured roofs and dormers from LIDAR data. We apply a hybrid strategy which combines the strengths of top-down and bottom-up methods. The main contribution is the introduction of an \textitactive sampling strategy which applies a cascade of filters focusing on promising samples in an early stage and avoiding the pitfalls of RANSAC based approaches. Such filters are based on prior knowledge represented by (non-parametric) density distributions. Samples are pairs of surflets, i.e. 3D points together with normal vectors derived from a plane approximation of their neighborhood. Surflet pairs imply immediately important roof parameters such as azimuth, inclination and ridge height, as well as parameters for internal precision and consistency, giving a good base for assessment and ranking. Ranking of samples leads to a small number of promising hypotheses. Model selection is based on predictions for example of ridge positions which can easily be falsified based on the given observations. Our approach does not require building footprints as prerequisite. They are derived in a preprocessing step using machine learning methods, in particular Support Vector Machines (SVM).

doi:10.5194/isprs-annals-IV-4-W8-43-2019

@inproceedings{dehbi2019active,
    abstract = {3D city models in Level-of-Detail 2 (LoD2) are nowadays inevitable for many applications such as solar radiation calculation and energy demand estimation.
City-wide models are required which can solely be acquired by fully automatic approaches.
In this paper we propose a novel method for the 3D-reconstruction of LoD2 buildings with structured roofs and dormers from LIDAR data.
We apply a hybrid strategy which combines the strengths of top-down and bottom-up methods.
The main contribution is the introduction of an \textit{active sampling} strategy which applies a cascade of filters focusing on promising samples in an early stage and avoiding the pitfalls of RANSAC based approaches.
Such filters are based on prior knowledge represented by (non-parametric) density distributions.
Samples are pairs of surflets, i.e. 3D points together with normal vectors derived from a plane approximation of their neighborhood.
Surflet pairs imply immediately important roof parameters such as azimuth, inclination and ridge height, as well as parameters for internal precision and consistency, giving a good base for assessment and ranking.
Ranking of samples leads to a small number of promising hypotheses.
Model selection is based on predictions for example of ridge positions which can easily be falsified based on the given observations.
Our approach does not require building footprints as prerequisite.
They are derived in a preprocessing step using machine learning methods, in particular Support Vector Machines (SVM).},
    author = {Dehbi, Y. and Henn, A. and Gr\"oger, G. and Stroh, V. and Pl\"umer, L.},
    booktitle = {Proc. 14th 3D Geoinfo Conference},
    doi = {10.5194/isprs-annals-IV-4-W8-43-2019},
    pages = {43--50},
    series = {ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
    title = {Active Sampling and Model Based Prediction for Fast and Robust Detection and Reconstruction of Complex Roofs in 3D Point Clouds},
    volume = {IV-4/W8},
    year = {2019}
}

LoD2 models include roof shapes and thus provide added value over their LoD1 counterparts for some applications such as estimating the solar potential of rooftops. However, because of laborious acquisition workflows they are more difficult to obtain than LoD1 models and are thus less prevalent in practice. This paper explores whether the type of the roof of a building can be inferred from semantic LoD1 data, potentially leading to their free upgrade to LoD2, in a broader context of a workflow for their generation without aerial campaigns. Inferring rooftop information has also other uses: data quality and verification of existing data, supporting roof reconstruction, and enriching LoD0/LoD1 data with the attribute of the roof type. We tested a RandomForest classifier that analyses attributes of buildings predicting the type of the roof. Experiments carried out on the 3D city model of Hamburg using 12 attributes achieve an accuracy of 85% in identifying the roof type from sparse data using a multiclass classification. The performance of binary classification hits the roof: 92% accuracy in predicting whether a roof is flat or not. It turns out that the two most useful variables are footprint area and building height (i.e. LoD1 models without any semantics, or LoD0 with such information), and using only them also yields relatively accurate results.

doi:10.5194/isprs-annals-IV-4-W8-27-2019

@inproceedings{dehbi2019towards,
    abstract = {LoD2 models include roof shapes and thus provide added value over their LoD1 counterparts for some applications such as estimating the solar potential of rooftops.
However, because of laborious acquisition workflows they are more difficult to obtain than LoD1 models and are thus less prevalent in practice.
This paper explores whether the type of the roof of a building can be inferred from semantic LoD1 data, potentially leading to their free upgrade to LoD2, in a broader context of a workflow for their generation without aerial campaigns.
Inferring rooftop information has also other uses: data quality and verification of existing data, supporting roof reconstruction, and enriching LoD0/LoD1 data with the attribute of the roof type.
We tested a RandomForest classifier that analyses attributes of buildings predicting the type of the roof.
Experiments carried out on the 3D city model of Hamburg using 12 attributes achieve an accuracy of 85\% in identifying the roof type from sparse data using a multiclass classification.
The performance of binary classification hits the roof: 92\% accuracy in predicting whether a roof is flat or not.
It turns out that the two most useful variables are footprint area and building height (i.e. LoD1 models without any semantics, or LoD0 with such information), and using only them also yields relatively accurate results.},
    author = {Biljecki, F. and Dehbi, Y.},
    booktitle = {Proc. 14th 3D Geoinfo Conference},
    doi = {10.5194/isprs-annals-IV-4-W8-27-2019},
    pages = {27--34},
    series = {ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
    title = {Raise the roof: Towards generating LoD2 models without aerial surveys using machine learning},
    volume = {IV-4/W8},
    year = {2019}
}

3D building models including roofs are a key prerequisite in many fields of applications such as the estimation of solar suitability of rooftops. The accurate reconstruction of roofs with dormers is sometimes challenging. Without careful separation of the dormer points from the points on the roof surface, the estimation of the roof areas is distorted in a most characteristic way, which then let the dormer points appear as white noise. The characteristic distortion of the density distribution of the defects by dormers in comparison to the expected normal distribution is the starting point of our method. We propose a hierarchical method which improves roof reconstruction from LiDAR point clouds in a model-based manner separating dormer points from roof points using classification methods. The key idea is to exploit probability density functions (PDFs) to reveal roof properties and design skilful features for a supervised learning method using support vector machines (SVMs). Properties of the PDFs of measures such as residuals of model-based estimated roof models are used among others. A clustering step leads to a semantic segmentation of the point cloud enabling subsequent reconstruction. The approach is tested based on real data as well as simulated point clouds. The latter allow for experiments for various roof and dormer types with different parameters using an implemented simulation toolbox which generates virtual buildings and synthetic point clouds.

doi:10.5194/isprs-archives-XLII-4-W16-177-2019

@inproceedings{dehbi2019probability,
    abstract = {3D building models including roofs are a key prerequisite in many fields of applications such as the estimation of solar suitability of rooftops. The accurate reconstruction of roofs with dormers is sometimes
challenging. Without careful separation of the dormer points from the points on the roof surface, the estimation of the roof areas is distorted in a most characteristic way, which then let the dormer points appear as white
noise. The characteristic distortion of the density distribution of the defects by dormers in comparison to the expected normal distribution is the starting point of our method. We propose a hierarchical method which improves
roof reconstruction from LiDAR point clouds in a model-based manner separating dormer points from roof points using classification methods. The key idea is to exploit probability density functions (PDFs) to reveal roof
properties and design skilful features for a supervised learning method using support vector machines (SVMs). Properties of the PDFs of measures such as residuals of model-based estimated roof models are used among others.
A clustering step leads to a semantic segmentation of the point cloud enabling subsequent reconstruction. The approach is tested based on real data as well as simulated point clouds. The latter allow for experiments for various
roof and dormer types with different parameters using an implemented simulation toolbox which generates virtual buildings and synthetic point clouds.},
    author = {Dehbi, Y. and Koppers, S. and Pl\"umer, L.},
    booktitle = {Proc. 6th International Conference on Geomatics and Geospatial Technology},
    doi = {10.5194/isprs-archives-XLII-4-W16-177-2019},
    pages = {177--184},
    series = {ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
    title = {Probability density based classification and reconstruction of roof structures from 3D point clouds},
    volume = {XLII-4/W16},
    year = {2019}
}

Accurate and robust positioning of vehicles in urban environments is of high importance for many applications (e.g. autonomous driving or mobile mapping). In the case of mobile mapping systems, a simultaneous mapping of the environment using laser scanning and an accurate positioning using GNSS is targeted. This requirement is often not guaranteed in shadowed cities where GNSS signals are usually disturbed, weak or even unavailable. Both, the generated point clouds and the derived trajectory are consequently imprecise. We propose a novel approach which incorporates prior knowledge, i.e. 3D building model of the environment, and improves the point cloud and the trajectory. The key idea is to benefit from the complementarity of both GNSS and 3D building models. The point cloud is matched to the city model using a point-to-plane ICP. An informed sampling of appropriate matching points is enabled by a pre-classification step. Support vector machines (SVMs) are used to discriminate between facade and remaining points. Local inconsistencies are tackled by a segment-wise partitioning of the point cloud where an interpolation guarantees a seamless transition between the segments. The full processing chain is implemented from the detection of facades in the point clouds, the matching between them and the building models and the update of the trajectory estimate. The general applicability of the implemented method is demonstrated on an inner city data set recorded with a mobile mapping system.

doi:10.5194/isprs-annals-IV-4-W9-35-2019

@inproceedings{dehbi2019Improving,
    abstract = {Accurate and robust positioning of vehicles in urban environments is of high importance for many applications (e.g. autonomous driving or mobile mapping). In the case of mobile mapping systems, a simultaneous mapping of the environment using laser scanning
and an accurate positioning using GNSS is targeted. This requirement is often not guaranteed in shadowed cities where GNSS signals are usually disturbed, weak or even unavailable. Both, the generated point clouds and the derived trajectory are consequently imprecise.
We propose a novel approach which incorporates prior knowledge, i.e. 3D building model of the environment, and improves the point cloud and the trajectory. The key idea is to benefit from the complementarity of both GNSS and 3D building models. The point
cloud is matched to the city model using a point-to-plane ICP. An informed sampling of appropriate matching points is enabled by a pre-classification step. Support vector machines (SVMs) are used to discriminate between facade and remaining points. Local
inconsistencies are tackled by a segment-wise partitioning of the point cloud where an interpolation guarantees a seamless transition between the segments. The full processing chain is implemented from the detection of facades in the point clouds, the matching
between them and the building models and the update of the trajectory estimate. The general applicability of the implemented method is demonstrated on an inner city data set recorded with a mobile mapping system.},
    author = {Dehbi, Y. and Lucks, L. and Behmann, J. and klingbeil, L. and Pl\"umer, L.},
    booktitle = {Proc. 4th International Conference on Smart Data and Smart Cities},
    doi = {10.5194/isprs-annals-IV-4-W9-35-2019},
    pages = {35--42},
    series = {ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
    title = {Improving GPS Trajectories using 3D City Models and Kinematic Point Clouds},
    volume = {IV-4/W9},
    year = {2019}
}

Understanding the criteria that bicyclists apply when they choose their routes is crucial for planning new bicycle paths or recommending routes to bicyclists. This is becoming more and more important as city councils are becoming increasingly aware of limitations of the transport infrastructure and problems related to automobile traffic. Since different groups of cyclists have different preferences, however, searching for a single set of criteria is prone to failure. Therefore, in this paper, we present a new approach to classify trajectories recorded and shared by bicyclists into different groups and, for each group, to identify favored and unfavored road types. Based on these results we show how to assign weights to the edges of a graph representing the road network such that minimumweight paths in the graph, which can be computed with standard shortest-path algorithms, correspond to adequate routes. Our method combines known algorithms for machine learning and the analysis of trajectories in an innovative way and, thereby, constitutes a new comprehensive solution for the problem of deriving routing preferences from initially unclassified trajectories. An important property of our method is that it yields reasonable results even if the given set of trajectories is sparse in the sense that it does not cover all segments of the cycle network.

doi:10.5194/isprs-annals-IV-4-W7-107-2018

@inproceedings{OehrleinEtAl2018,
    abstract = {Understanding the criteria that bicyclists apply when they choose their routes is crucial for planning new bicycle paths or recommending routes to bicyclists. This is becoming more and more important as city councils are becoming increasingly aware of limitations of the transport infrastructure and problems related to automobile traffic. Since different groups of cyclists have different preferences, however, searching for a single set of criteria is prone to failure. Therefore, in this paper, we present a new approach to classify trajectories recorded and shared by bicyclists into different groups and, for each group, to identify favored and unfavored road types. Based on these results we show how to assign weights to the edges of a graph representing the road network such that minimumweight paths in the graph, which can be computed with standard shortest-path algorithms, correspond to adequate routes. Our method combines known algorithms for machine learning and the analysis of trajectories in an innovative way and, thereby, constitutes a new comprehensive solution for the problem of deriving routing preferences from initially unclassified trajectories. An important property of our method is that it yields reasonable results even if the given set of trajectories is sparse in the sense that it does not cover all segments of the cycle network.},
    author = {Oehrlein, J. and Förster, A. and Schunck, D. and Dehbi, Y. and Roscher, R. and Haunert, J.-H.},
    booktitle = {Proc. 3rd International Conference on Smart Data and Smart Cities},
    doi = {10.5194/isprs-annals-IV-4-W7-107-2018},
    pages = {107--114},
    series = {ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
    title = {Inferring Routing Preferences of Bicyclists from Sparse Sets of Trajectories},
    volume = {IV-4/W7},
    year = {2018}
}

Many researchers used expensive 3D laser scanning techniques to derive indoor models. Few papers describe the derivation of indoor models based on sparse data such as footprints. They assume that floorplans and rooms are rather rectangular and that information on functional use is given. This paper addresses the automatic learning of a classifier which predicts the functional use of housing rooms. The classification is based on features which are widely available such as room areas and orientation. These features are extracted from an extensive database of annotated rooms. A Bayesian classifier is applied which delivers probabilities of competing class hypotheses. In a second step, functional uses are used to predict the shape of the rooms in a further classification.

doi:10.5194/isprs-annals-IV-4-33-2018

@inproceedings{DehbiEtAl2018,
    abstract = {Many researchers used expensive 3D laser scanning techniques to derive indoor models. Few papers describe the derivation of indoor models based on sparse data such as footprints. They assume that floorplans and rooms are rather rectangular and that information on functional use is given. This paper addresses the automatic learning of a classifier which predicts the functional use of housing rooms. The classification is based on features which are widely available such as room areas and orientation. These features are extracted from an extensive database of annotated rooms. A Bayesian classifier is applied which delivers probabilities of competing class hypotheses. In a second step, functional uses are used to predict the shape of the rooms in a further classification.},
    author = {Dehbi, Y. and Gojayeva, N. and Pickert, A. R. and Haunert, J.-H. and Pl\"{u}mer, L.},
    booktitle = {Proc. ISPRS Technical Commission IV Symposium},
    doi = {10.5194/isprs-annals-IV-4-33-2018},
    series = {ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
    title = {Room shapes and functional uses predicted from sparse data},
    volume = {IV-4:33--40},
    year = {2018}
}

3D city and building models according to CityGML encode the geometry, represent the structure and model semantically relevant building parts such as doors, windows and balconies. Building information models support the building design, construction and the facility management. In contrast to CityGML, they include also objects which cannot be observed from the outside. The three dimensional indoor models characterize a missing link between both worlds. Their derivation, however, is expensive. The semantic automatic interpretation of 3D point clouds of indoor environments is a methodically demanding task. The data acquisition is costly and difficult. The laser scanners and image-based methods require the access to every room. Based on an approach which does not require an additional geometry acquisition of building indoors, we propose an attempt for filling the gaps between 3D building models and building information models. Based on sparse observations such as the building footprint and room areas, 3D indoor models are generated using combinatorial and stochastic reasoning. The derived models are expanded by a-priori not observable structures such as electric installation. Gaussian mixtures, linear and bi-linear constraints are used to represent the background knowledge and structural regularities. The derivation of hypothesised models is performed by stochastic reasoning using graphical models, Gauss-Markov models and MAP-estimators.

doi:10.5194/isprs-annals-IV-4-W5-33-2017

@inproceedings{isprs-annals-IV-4-W5-33-2017,
    abstract = {3D city and building models according to CityGML encode the geometry, represent the structure and model semantically relevant building parts such as doors, windows and balconies. Building information models support the building design, construction and the facility management. In contrast to CityGML, they include also objects which cannot be observed from the outside. The three dimensional indoor models characterize a missing link between both worlds. Their derivation, however, is expensive. The semantic automatic interpretation of 3D point clouds of indoor environments is a methodically demanding task. The data acquisition is costly and difficult. The laser scanners and image-based methods require the access to every room. Based on an approach which does not require an additional geometry acquisition of building indoors, we propose an attempt for filling the gaps between 3D building models and building information models. Based on sparse observations such as the building footprint and room areas, 3D indoor models are generated using combinatorial and stochastic reasoning. The derived models are expanded by a-priori not observable structures such as electric installation. Gaussian mixtures, linear and bi-linear constraints are used to represent the background knowledge and structural regularities. The derivation of hypothesised models is performed by stochastic reasoning using graphical models, Gauss-Markov models and MAP-estimators.},
    author = {Dehbi, Y. and Haunert, J.-H. and Pl\"umer, L.},
    booktitle = {Proc. 12th 3D Geoinfo Conference},
    doi = {10.5194/isprs-annals-IV-4-W5-33-2017},
    pages = {33--39},
    series = {ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences},
    title = {Stochastic and geometric reasoning for indoor building models with electric installations -- bridging the gap between GIS and BIM},
    url = {https://www.isprs-ann-photogramm-remote-sens-spatial-inf-sci.net/IV-4-W5/33/2017/},
    volume = {IV-4/W5},
    year = {2017}
}

This paper presents a novel method for the parameter estimation and model selection for the reconstruction of indoor environments based on sparse observations. While most approaches for the reconstruction of indoor models rely on dense observations, we predict scenes of the interior with high accuracy in the absence of indoor measurements. We use a model-based top-down approach and incorporate strong but profound prior knowledge. The latter includes probability density functions for model parameters and sparse observations such as room areas and the building footprint. The floorplan model is characterized by linear and bi-linear relations with discrete and continuous parameters. We focus on the stochastic estimation of model parameters based on a topological model derived by combinatorial reasoning in a first step. A Gauss-Markov model is applied for estimation and simulation of the model parameters. Symmetries are represented and exploited during the estimation process. Background knowledge as well as observations are incorporated in a maximum likelihood estimation and model selection is performed with AIC/BIC. The likelihood is also used for the detection and correction of potential errors in the topological model. Estimation results are presented and discussed.

doi:10.5194/isprs-annals-IV-2-W4-303-2017

@inproceedings{isprs-annals-IV-2-W4-303-2017,
    abstract = {This paper presents a novel method for the parameter estimation and model selection for the reconstruction of indoor environments based on sparse observations. While most approaches for the reconstruction of indoor models rely on dense observations, we predict scenes of the interior with high accuracy in the absence of indoor measurements. We use a model-based top-down approach and incorporate strong but profound prior knowledge. The latter includes probability density functions for model parameters and sparse observations such as room areas and the building footprint. The floorplan model is characterized by linear and bi-linear relations with discrete and continuous parameters. We focus on the stochastic estimation of model parameters based on a topological model derived by combinatorial reasoning in a first step. A Gauss-Markov model is applied for estimation and simulation of the model parameters. Symmetries are represented and exploited during the estimation process. Background knowledge as well as observations are incorporated in a maximum likelihood estimation and model selection is performed with AIC/BIC. The likelihood is also used for the detection and correction of potential errors in the topological model. Estimation results are presented and discussed.},
    author = {Dehbi, Y. and Loch-Dehbi, S. and Pl\"umer, L.},
    booktitle = {Proc. ISPRS Geospatial Week 2017},
    doi = {10.5194/isprs-annals-IV-2-W4-303-2017},
    pages = {303--310},
    series = {ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences},
    title = {Parameter estimation and model selection for indoor models based on sparse observations},
    url = {https://www.isprs-ann-photogramm-remote-sens-spatial-inf-sci.net/IV-2-W4/303/2017/},
    volume = {IV-2/W4},
    year = {2017}
}

The automatic interpretation of 3D point clouds for building reconstruction is a challenging task. The interpretation process requires highly structured models representing semantics. Formal grammars can describe structures as well as the parameters of buildings and their parts. We propose a novel approach for the automatic learning of weighted attributed context-free grammar rules for 3D building reconstruction, supporting the laborious manual design of rules. We separate structure from parameter learning. Specific Support Vector Machines (SVMs) are used to generate a weighted context-free grammar and predict structured outputs such as parse trees. The grammar is extended by parameters and constraints, which are learned based on a statistical relational learning method using Markov Logic Networks (MLNs). MLNs enforce the topological and geometric constraints. MLNs address uncertainty explicitly and provide probabilistic inference. They are able to deal with partial observations caused by occlusions. Uncertain projective geometry is used to deal with the uncertainty of the observations. Learning is based on a large building database covering different building styles and façade structures. In particular, a treebank that has been derived from the database is employed for structure learning.

doi:10.1111/tgis.12200

@article{DehbiEtAl2017,
    abstract = {The automatic interpretation of 3D point clouds for building reconstruction is a challenging task. The interpretation process requires highly structured models representing semantics. Formal grammars can describe structures as well as the parameters of buildings and their parts. We propose a novel approach for the automatic learning of weighted attributed context-free grammar rules for 3D building reconstruction, supporting the laborious manual design of rules. We separate structure from parameter learning. Specific Support Vector Machines (SVMs) are used to generate a weighted context-free grammar and predict structured outputs such as parse trees. The grammar is extended by parameters and constraints, which are learned based on a statistical relational learning method using Markov Logic Networks (MLNs). MLNs enforce the topological and geometric constraints. MLNs address uncertainty explicitly and provide probabilistic inference. They are able to deal with partial observations caused by occlusions. Uncertain projective geometry is used to deal with the uncertainty of the observations. Learning is based on a large building database covering different building styles and façade structures. In particular, a treebank that has been derived from the database is employed for structure learning.},
    author = {Dehbi, Y. and Hadiji, F. and Gr{\"{o}}ger, Gerhard and Kersting, Kristian and Pl{\"{u}}mer, L.},
    doi = {10.1111/tgis.12200},
    journal = {Transactions in GIS},
    number = {1},
    pages = {134--150},
    title = {Statistical Relational Learning of Grammar Rules for 3D Building Reconstruction},
    url = {http://dx.doi.org/10.1111/tgis.12200},
    volume = {21},
    year = {2017}
}

This paper presents a novel method for the prediction of building floor plans based on sparse observations in the absence of measurements. We derive the most likely hypothesis using a maximum a posteriori probability approach. Background knowledge consisting of probability density functions of room shape and location parameters is learned from training data. Relations between rooms and room substructures are represented by linear and bilinear constraints. We perform reasoning on different levels providing a problem solution that is optimal with regard to the given information. In a first step, the problem is modeled as a constraint satisfaction problem. Constraint Logic Programming derives a solution which is topologically correct but suboptimal with regard to the geometric parameters. The search space is reduced using architectural constraints and browsed by intelligent search strategies which use domain knowledge. In a second step, graphical models are used for updating the initial hypothesis and refining its continuous parameters. We make use of Gaussian mixtures for model parameters in order to represent background knowledge and to get access to established methods for efficient and exact stochastic reasoning. We demonstrate our approach on different illustrative examples. Initially, we assume that floor plans are rectangular and that rooms are rectangles and discuss more general shapes afterwards. In a similar spirit, we predict door locations providing further important components of 3D indoor models.

doi:10.3390/ijgi6030090

@article{Loch-DehbiEtAl2017,
    abstract = {This paper presents a novel method for the prediction of building floor plans based on sparse observations in the absence of measurements. We derive the most likely hypothesis using a maximum a posteriori probability approach. Background knowledge consisting of probability density functions of room shape and location parameters is learned from training data. Relations between rooms and room substructures are represented by linear and bilinear constraints. We perform reasoning on different levels providing a problem solution that is optimal with regard to the given information. In a first step, the problem is modeled as a constraint satisfaction problem. Constraint Logic Programming derives a solution which is topologically correct but suboptimal with regard to the geometric parameters. The search space is reduced using architectural constraints and browsed by intelligent search strategies which use domain knowledge. In a second step, graphical models are used for updating the initial hypothesis and refining its continuous parameters. We make use of Gaussian mixtures for model parameters in order to represent background knowledge and to get access to established methods for efficient and exact stochastic reasoning. We demonstrate our approach on different illustrative examples. Initially, we assume that floor plans are rectangular and that rooms are rectangles and discuss more general shapes afterwards. In a similar spirit, we predict door locations providing further important components of 3D indoor models.},
    article-number = {90},
    author = {Loch-Dehbi, S. and Dehbi, Y. and Pl{\"{u}}mer, L.},
    doi = {10.3390/ijgi6030090},
    issn = {2220-9964},
    journal = {ISPRS International Journal of Geo-Information},
    number = {3},
    title = {Estimation of 3D Indoor Models with Constraint Propagation and Stochastic Reasoning in the Absence of Indoor Measurements},
    url = {http://www.mdpi.com/2220-9964/6/3/90},
    volume = {6},
    year = {2017}
}

Buildings and other man-made objects, for many reasons such as economical or aesthetic, are often characterized by their symmetry. The latter predominates in the design of building footprints and building parts such as façades. Thus the identification and modeling of this valuable information facilitates the reconstruction of these buildings and their parts. This article presents a novel approach for the automatic identification and modelling of symmetries and their hierarchical structures in building footprints, providing an important prior for façade and roof reconstruction. The uncertainty of symmetries is explicitly addressed using supervised machine learning methods, in particular Support Vector Machines (SVMs). Unlike classical statistical methods, for SVMs assumptions on the a priori distribution of the data are not required. Both axial and translational symmetries are detected. The quality of the identified major and minor symmetry axes is assessed by a least squares based adjustment. Context-free formal grammar rules are used to model the hierarchical and repetitive structure of the underlying footprints. We present an algorithm which derives grammar rules based on the previously acquired symmetry information and using lexical analysis describing regular patterns and palindrome-like structures. This offers insights into the latent structures of building footprints and therefore describes the associated façade in a relational and compact way.

doi:10.1111/tgis.12177

@article{DehbiEtAl2016,
    abstract = {Buildings and other man-made objects, for many reasons such as economical or aesthetic, are often characterized by their symmetry. The latter predominates in the design of building footprints and building parts such as façades. Thus the identification and modeling of this valuable information facilitates the reconstruction of these buildings and their parts. This article presents a novel approach for the automatic identification and modelling of symmetries and their hierarchical structures in building footprints, providing an important prior for façade and roof reconstruction. The uncertainty of symmetries is explicitly addressed using supervised machine learning methods, in particular Support Vector Machines (SVMs). Unlike classical statistical methods, for SVMs assumptions on the a priori distribution of the data are not required. Both axial and translational symmetries are detected. The quality of the identified major and minor symmetry axes is assessed by a least squares based adjustment. Context-free formal grammar rules are used to model the hierarchical and repetitive structure of the underlying footprints. We present an algorithm which derives grammar rules based on the previously acquired symmetry information and using lexical analysis describing regular patterns and palindrome-like structures. This offers insights into the latent structures of building footprints and therefore describes the associated façade in a relational and compact way.},
    author = {Dehbi, Y. and Gr{\"{o}}ger, Gerhard and Pl{\"{u}}mer, L.},
    doi = {10.1111/tgis.12177},
    journal = {Transactions in GIS},
    number = {5},
    pages = {645--663},
    title = {Identification and modelling of translational and axial symmetries and their hierarchical structures in building footprints by formal grammars},
    url = {http://dx.doi.org/10.1111/tgis.12177},
    volume = {20},
    year = {2016}
}

Formal grammars are well suited for the estimation of models with an a-priori unknown number of parameters such as buildings and have proven their worth for 3D modeling and reconstruction of cities. However, the generation and design of corresponding grammar rules is a laborious task and relies on expert knowledge. This thesis presents novel approaches for the reduction of this effort using advanced machine learning methods resulting in automatically learned sophisticated grammar rules. Indeed, the learning of a wide range of sophisticated rules, that reflect the variety and complexity, is a challenging task. This is especially the case if a simultaneous machine learning of building structures and the underlying aggregation hierarchies as well as the building parameters and the constraints among them for a semantic interpretation is expected. Thus, in this thesis, an incremental approach is followed. It separates the structure learning from the parameter distribution learning of building parts. Moreover, the so far procedural approaches with formal grammars are mostly rather convenient for the generation of virtual city models than for the reconstruction of existing buildings. To this end, Inductive Logic Programming (ILP) techniques are transferred and applied for the first time in the field of 3D building modeling. This enables the automatic learning of declarative logic programs, which are equivalent to attribute grammars and separate the representation of buildings and their parts from the reconstruction task. A stepwise bottom-up learning, starting from the smallest atomic features of a building part together with the semantic, topological and geometric constraints, is a key to a successful learning of a whole building part. Only few examples are sufficient to learn from precise as well as noisy observations. The learning from uncertain data is realized using probability density functions, decision trees and uncertain projective geometry. This enables the handling and modeling of uncertain topology and geometric reasoning taking noise into consideration. The uncertainty of models itself is also considered. Therefore, a novel method is developed for the learning of Weighted Attribute Context-Free Grammar (WACFG). On the one hand, the structure learning of façades - context-free part of the Grammar - is performed based on annotated derivation trees using specific Support Vector Machines (SVMs). The latter are able to derive probabilistic models from structured data and to predict a most likely tree regarding to given observations. On the other hand, to the best of my knowledge, Statistical Relational Learning (SRL), especially Markov Logic Networks (MLNs), are applied for the first time in order to learn building part (shape and location) parameters as well as the constraints among these parts. The use of SRL enables to take profit from the elegant logical relational description and to benefit from the efficiency of statistical inference methods. In order to model latent prior knowledge and exploit the architectural regularities of buildings, a novel method is developed for the automatic identification of translational as well as axial symmetries. For symmetry identification a supervised machine learning approach is followed based on an SVM classifier. Building upon the classification results, algorithms are designed for the representation of symmetries using context-free grammars from authoritative building footprints. In all steps the machine learning is performed based on real- world data such as 3D point clouds and building footprints. The handling with uncertainty and occlusions is assured. The presented methods have been successfully applied on real data. The belonging classification and reconstruction results are shown.

@phdthesis{dehbi2016statistical,
    abstract = {Formal grammars are well suited for the estimation of models with an a-priori unknown number of parameters such as buildings and have proven their worth for 3D modeling and reconstruction of cities. However, the generation and design of corresponding grammar rules is a laborious task and relies on expert knowledge. This thesis presents novel approaches for the reduction of this effort using advanced machine learning methods resulting in automatically learned sophisticated grammar rules. Indeed, the learning of a wide range of sophisticated rules, that reflect the variety and complexity, is a challenging task. This is especially the case if a simultaneous machine learning of building structures and the underlying aggregation hierarchies as well as the building parameters and the constraints among them for a semantic interpretation is expected. Thus, in this thesis, an incremental approach is followed. It separates the structure learning from the parameter distribution learning of building parts. Moreover, the so far procedural approaches with formal grammars are mostly rather convenient for the generation of virtual city models than for the reconstruction of existing buildings. To this end, Inductive Logic Programming (ILP) techniques are transferred and applied for the first time in the field of 3D building modeling. This enables the automatic learning of declarative logic programs, which are equivalent to attribute grammars and separate the representation of buildings and their parts from the reconstruction task. A stepwise bottom-up learning, starting from the smallest atomic features of a building part together with the semantic, topological and geometric constraints, is a key to a successful learning of a whole building part. Only few examples are sufficient to learn from precise as well as noisy observations. The learning from uncertain data is realized using probability density functions, decision trees and uncertain projective geometry. This enables the handling and modeling of uncertain topology and geometric reasoning taking noise into consideration. The uncertainty of models itself is also considered. Therefore, a novel method is developed for the learning of Weighted Attribute Context-Free Grammar (WACFG). On the one hand, the structure learning of façades -- context-free part of the Grammar -- is performed based on annotated derivation trees using specific Support Vector Machines (SVMs). The latter are able to derive probabilistic models from structured data and to predict a most likely tree regarding to given observations. On the other hand, to the best of my knowledge, Statistical Relational Learning (SRL), especially Markov Logic Networks (MLNs), are applied for the first time in order to learn building part (shape and location) parameters as well as the constraints among these parts. The use of SRL enables to take profit from the elegant logical relational description and to benefit from the efficiency of statistical inference methods. In order to model latent prior knowledge and exploit the architectural regularities of buildings, a novel method is developed for the automatic identification of translational as well as axial symmetries. For symmetry identification a supervised machine learning approach is followed based on an SVM classifier. Building upon the classification results, algorithms are designed for the representation of symmetries using context-free grammars from authoritative building footprints. In all steps the machine learning is performed based on real- world data such as 3D point clouds and building footprints. The handling with uncertainty and occlusions is assured. The presented methods have been successfully applied on real data. The belonging classification and reconstruction results are shown.},
    author = {Dehbi, Y.},
    school = {Rheinische Friedrich-Wilhelms-Universit{\"{a}}t Bonn, Germany},
    series = {Universitäts- und Landesbibliothek Bonn},
    title = {Statistical relational learning of semantic models and grammar rules for 3D building reconstruction from 3D point clouds},
    type = {Dissertation},
    url = {http://hss.ulb.uni-bonn.de/2016/4532/4532.htm},
    urn = {urn:nbn:de:hbz:5n-45329},
    year = {2016}
}

Data acquisition using unmanned aerial vehicles (UAVs) has gotten more and more attention over the last years. Especially in the field of building reconstruction the incremental interpretation of such data is a demanding task. In this context formal grammars play an important role for the top-down identification and reconstruction of building objects. Up to now, the available approaches expect offline data in order to parse an a-priori known grammar. For mapping on demand an on the fly reconstruction based on UAV data is required. An incremental interpretation of the data stream is inevitable. This paper presents an incremental parser of grammar rules for an automatic 3D building reconstruction. The parser enables a model refinement based on new observations with respect to a weighted attribute context-free grammar (WACFG). The falsification or rejection of hypotheses is supported as well. The parser can deal with and adapt available parse trees acquired from previous interpretations or predictions. Parse trees derived so far are updated in an iterative way using transformation rules. A diagnostic step searches for mismatches between current and new nodes. Prior knowledge on facžades is incorporated. It is given by probability densities as well as architectural patterns. Since we cannot always assume normal distributions, the derivation of location and shape parameters of building objects is based on a kernel density estimation (KDE). While the level of detail is continuously improved, the geometrical, semantic and topological consistency is ensured.

doi:10.5194/isprs-annals-III-3-311-2016

@inproceedings{isprs-annals-III-3-311-2016,
    abstract = {Data acquisition using unmanned aerial vehicles (UAVs) has gotten more and more attention over the last years. Especially in the field of building reconstruction the incremental interpretation of such data is a demanding task. In this context formal grammars play an important role for the top-down identification and reconstruction of building objects. Up to now, the available approaches expect offline data in order to parse an a-priori known grammar. For mapping on demand an on the fly reconstruction based on UAV data is required. An incremental interpretation of the data stream is inevitable. This paper presents an incremental parser of grammar rules for an automatic 3D building reconstruction. The parser enables a model refinement based on new observations with respect to a weighted attribute context-free grammar (WACFG). The falsification or rejection of hypotheses is supported as well. The parser can deal with and adapt available parse trees acquired from previous interpretations or predictions. Parse trees derived so far are updated in an iterative way using transformation rules. A diagnostic step searches for mismatches between current and new nodes. Prior knowledge on facžades is incorporated. It is given by probability densities as well as architectural patterns. Since we cannot always assume normal distributions, the derivation of location and shape parameters of building objects is based on a kernel density estimation (KDE). While the level of detail is continuously improved, the geometrical, semantic and topological consistency is ensured.},
    author = {Dehbi, Y and Staat, C and Mandtler, L and Pl{\"{u}}mer, L},
    booktitle = {Proc. XXIII ISPRS Congress},
    doi = {10.5194/isprs-annals-III-3-311-2016},
    pages = {311--316},
    series = {ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences},
    title = {Incremental refinement of facade models with attribute grammar from 3D point clouds},
    url = {http://www.isprs-ann-photogramm-remote-sens-spatial-inf-sci.net/III-3/311/2016/},
    volume = {III-3},
    year = {2016}
}

This paper introduces a novel method for the automatic derivation of building floorplans and indoor models. Our approach is based on a logical and stochastic reasoning using sparse observations such as building room areas. No further sensor observations like 3D point clouds are needed. Our method benefits from an extensive prior knowledge of functional dependencies and probability density functions of shape and location parameters of rooms depending on their functional use. The determination of posterior beliefs is performed using Bayesian Networks. Stochastic reasoning is complex since the problem is characterized by a mixture of discrete and continuous parameters that are in turn correlated by non-linear constraints. To cope with this kind of complexity, the proposed reasoner combines statistical methods with constraint propagation. It generates a limited number of hypotheses in a model-based top-down approach. It predicts floorplans based on a-priori localised windows. The use of Gaussian mixture models, constraint solvers and stochastic models helps to cope with the a-priori infinite space of the possible floorplan instantiations.

doi:10.5194/isprs-annals-IV-2-W1-265-2016

@inproceedings{isprs-annals-IV-2-W1-265-2016,
    abstract = {This paper introduces a novel method for the automatic derivation of building floorplans and indoor models. Our approach is based on a logical and stochastic reasoning using sparse observations such as building room areas. No further sensor observations like 3D point clouds are needed. Our method benefits from an extensive prior knowledge of functional dependencies and probability density functions of shape and location parameters of rooms depending on their functional use. The determination of posterior beliefs is performed using Bayesian Networks. Stochastic reasoning is complex since the problem is characterized by a mixture of discrete and continuous parameters that are in turn correlated by non-linear constraints. To cope with this kind of complexity, the proposed reasoner combines statistical methods with constraint propagation. It generates a limited number of hypotheses in a model-based top-down approach. It predicts floorplans based on a-priori localised windows. The use of Gaussian mixture models, constraint solvers and stochastic models helps to cope with the a-priori infinite space of the possible floorplan instantiations.},
    author = {Loch-Dehbi, S. and Dehbi, Y. and Gr\"oger, G. and Plümer, L.},
    booktitle = {Proc. 11th 3D Geoinfo Conference},
    doi = {10.5194/isprs-annals-IV-2-W1-265-2016},
    pages = {265--270},
    series = {ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences},
    title = {Prediction of building floorplans using logical and stochastic reasoning based on sparse observations},
    url = {https://www.isprs-ann-photogramm-remote-sens-spatial-inf-sci.net/IV-2-W1/265/2016/},
    volume = {IV-2/W1},
    year = {2016}
}

The acquisition of detailed information for buildings and their components becomes more and more important. However, an automatic reconstruction needs high-resolution measurements. Such features can be derived from images or 3D laserscans that are e.g. taken by unmanned aerial vehicles (UAV). Since this data is not always available or not measurable at the first for example due to occlusions we developed a reasoning approach that is based on sparse observations. It benefits from an extensive prior knowledge of probability density distributions and functional dependencies and allows for the incorporation of further structural characteristics such as symmetries. Bayesian networks are used to determine posterior beliefs. Stochastic reasoning is complex since the problem is characterized by a mixture of discrete and continuous parameters that are in turn correlated by nonlinear constraints. To cope with this kind of complexity, the implemented reasoner combines statistical methods with constraint propagation. It generates a limited number of hypotheses in a model-based top-down approach. It predicts substructures in building facades - such as windows - that can be used for specific UAV navigations for further measurements.

doi:10.5194/isprsarchives-XL-1-W2-257-2013

@inproceedings{isprs-archives-XL-1-W2-257-2013,
    abstract = {The acquisition of detailed information for buildings and their components becomes more and more important. However, an automatic reconstruction needs high-resolution measurements. Such features can be derived from images or 3D laserscans that are e.g. taken by unmanned aerial vehicles (UAV). Since this data is not always available or not measurable at the first for example due to occlusions we developed a reasoning approach that is based on sparse observations. It benefits from an extensive prior knowledge of probability density distributions and functional dependencies and allows for the incorporation of further structural characteristics such as symmetries. Bayesian networks are used to determine posterior beliefs. Stochastic reasoning is complex since the problem is characterized by a mixture of discrete and continuous parameters that are in turn correlated by nonlinear constraints. To cope with this kind of complexity, the implemented reasoner combines statistical methods with constraint propagation. It generates a limited number of hypotheses in a model-based top-down approach. It predicts substructures in building facades -- such as windows -- that can be used for specific UAV navigations for further measurements.},
    author = {Loch-Dehbi, S. and Dehbi, Y. and Pl\"umer, L.},
    booktitle = {Proc. UAV-g Conference 2013},
    doi = {10.5194/isprsarchives-XL-1-W2-257-2013},
    pages = {257--261},
    series = {ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences},
    title = {Stochastic reasoning for UAV supported reconstruction of 3D building models},
    url = {https://www.int-arch-photogramm-remote-sens-spatial-inf-sci.net/XL-1-W2/257/2013/},
    volume = {XL-1/W2},
    year = {2013}
}

The automatic interpretation of dense three-dimensional (3D) point clouds is still an open research problem. The quality and usability of the derived models depend to a large degree on the availability of highly structured models which represent semantics explicitly and provide a priori knowledge to the interpretation process. The usage of formal grammars for modelling man-made objects has gained increasing interest in the last few years. In order to cope with the variety and complexity of buildings, a large number of fairly sophisticated grammar rules are needed. As yet, such rules mostly have to be designed by human experts. This article describes a novel approach to machine learning of attribute grammar rules based on the Inductive Logic Programming paradigm. Apart from syntactic differences, logic programs and attribute grammars are basically the same language. Attribute grammars extend context-free grammars by attributes and semantic rules and provide a much larger expressive power. Our approach to derive attribute grammars is able to deal with two kinds of input data. On the one hand, we show how attribute grammars can be derived from precise descriptions in the form of examples provided by a human user as the teacher. On the other hand, we present the acquisition of models from noisy observations such as 3D point clouds. This includes the learning of geometric and topological constraints by taking measurement errors into account. The feasibility of our approach is proven exemplarily by stairs, and a generic framework for learning other building parts is discussed. Stairs aggregate an arbitrary number of steps in a manner which is specified by topological and geometric constraints and can be modelled in a recursive way. Due to this recursion, they pose a special challenge to machine learning. In order to learn the concept of stairs, only a small number of examples were required. Our approach represents and addresses the quality of the given observations and the derived constraints explicitly, using concepts from uncertain projective geometry for learning geometric relations and the Wakeby distribution together with decision trees for topological relations.

doi:https://doi.org/10.1016/j.isprsjprs.2010.10.001

@article{Dehbi2011,
    abstract = {The automatic interpretation of dense three-dimensional (3D) point clouds is still an open research problem. The quality and usability of the derived models depend to a large degree on the availability of highly structured models which represent semantics explicitly and provide a priori knowledge to the interpretation process. The usage of formal grammars for modelling man-made objects has gained increasing interest in the last few years. In order to cope with the variety and complexity of buildings, a large number of fairly sophisticated grammar rules are needed. As yet, such rules mostly have to be designed by human experts. This article describes a novel approach to machine learning of attribute grammar rules based on the Inductive Logic Programming paradigm. Apart from syntactic differences, logic programs and attribute grammars are basically the same language. Attribute grammars extend context-free grammars by attributes and semantic rules and provide a much larger expressive power. Our approach to derive attribute grammars is able to deal with two kinds of input data. On the one hand, we show how attribute grammars can be derived from precise descriptions in the form of examples provided by a human user as the teacher. On the other hand, we present the acquisition of models from noisy observations such as 3D point clouds. This includes the learning of geometric and topological constraints by taking measurement errors into account. The feasibility of our approach is proven exemplarily by stairs, and a generic framework for learning other building parts is discussed. Stairs aggregate an arbitrary number of steps in a manner which is specified by topological and geometric constraints and can be modelled in a recursive way. Due to this recursion, they pose a special challenge to machine learning. In order to learn the concept of stairs, only a small number of examples were required. Our approach represents and addresses the quality of the given observations and the derived constraints explicitly, using concepts from uncertain projective geometry for learning geometric relations and the Wakeby distribution together with decision trees for topological relations.},
    author = {Dehbi, Y and Pl\"umer, L},
    doi = {https://doi.org/10.1016/j.isprsjprs.2010.10.001},
    issn = {0924-2716},
    journal = {ISPRS Journal of Photogrammetry and Remote Sensing},
    note = {Quality, Scale and Analysis Aspects of Urban City Models},
    number = {2},
    pages = {166--6176},
    title = {Learning grammar rules of building parts from precise models and noisy observations},
    url = {http://www.sciencedirect.com/science/article/pii/S0924271610000924},
    volume = {66},
    year = {2011}
}

Building reconstruction and building model generation nowadays receives more and more attention. In this context models such as formal grammars play a major role in 3D geometric modelling. Up to now, models have been designed manually by experts such as architects. Hence, this paper describes an Inductive Logic Programming based approach for learning semantic models and grammar rules of buildings and their parts. Due to their complex structure and their important role as link between the building and its outside, stairs are presented as an example. ILP is introduced and applied as machine learning method. The learning process is explained and the learned models and results are presented.

@inproceedings{dehbi2010,
    abstract = {Building reconstruction and building model generation nowadays receives more and more attention. In this context models such as formal grammars play a major role in 3D geometric modelling. Up to now, models have been designed manually by experts such as architects. Hence, this paper describes an Inductive Logic Programming based approach for learning semantic models and grammar rules of buildings and their parts. Due to their complex structure and their important role as link between the building and its outside, stairs are presented as an example. ILP is introduced and applied as machine learning method. The learning process is explained and the learned models and results are presented.},
    author = {Dehbi, Y. and Schmittwilken, J{\"{o}}rg and Pl{\"{u}}mer, L.},
    booktitle = {Proc. 24th Workshop on Constraint Logic Programming (WLP 2010)},
    pages = {45--56},
    publisher = {German University in Cairo},
    title = {Learning Semantic Models and Grammar Rules of Building Parts},
    year = {2010}
}

Building reconstruction and building model generation nowadays receives more and more attention. In this context models such as formal grammars play a major role in 3D geometric modelling. Up to now, models have been designed manually by experts such as architects. Hence, this paper describes an Inductive Logic Programming (ILP) based approach for learning semantic models and grammar rules of buildings and their parts. Due to their complex structure and their important role as link between the building and its outside, straight stairs are presented as an example. ILP is introduced and applied as machine learning method. The learning process is explained and the learned models and results are presented.

@inproceedings{dehbi2009,
    abstract = {Building reconstruction and building model generation nowadays receives more and more attention. In this context models such as formal grammars play a major role in 3D geometric modelling. Up to now, models have been designed manually by experts such as architects. Hence, this paper describes an Inductive Logic Programming (ILP) based approach for learning semantic models and grammar rules of buildings and their parts. Due to their complex structure and their important role as link between the building and its outside, straight stairs are presented as an example. ILP is introduced and applied as machine learning method. The learning process is explained and the learned models and results are presented.},
    author = {Dehbi, Y. and Schmittwilken, J{\"{o}}rg and Pl{\"{u}}mer, L.},
    booktitle = {Proc. of the ISPRS WG II/2+3+4 and Cost Workshop. Lund, Sweden 2009},
    series = {ISPRS Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences},
    title = {Learning Semantic Models and Grammar Rules of Building Parts},
    volume = {XXXVIII-2/W11},
    year = {2009}
}