Information

  • Publication Type: Master Thesis
  • Workgroup(s)/Project(s):
  • Date: June 2026
  • Date (Start): 18. January 2024
  • Date (End): 24. June 2026
  • TU Wien Library:
  • Diploma Examination: 2026
  • First Supervisor: Stefan OhrhallingerORCID iD
  • Keywords: point clouds, change detection, change tracking

Abstract

To know the surrounding environment is required in different applications, such as augmented reality, robotics and autonomous vehicles. While deep learning technologies allow advanced methods for image analysis and object detection, another important aspect of monitoring the environment is detecting changes, distinguishing removed or added objects and rotated or shifted objects, which is challenging especially in very large point clouds for example from outdoor environments that are observed with large epochs in between, even when the changed points are marked. To develop such applications, there are further barriers to overcome: changing viewpoint, occluded objects, observing objects from different angles, computational efficiency for a real-time application.\\ \\Information about changes can potentially be used to train deep learning algorithms for the semantic understanding of environments. Additionally, we would like to answer large-scale queries such as which objects of a certain class such as trees, poles or traffic signed have changed, or an inventory of how many cars are in a city during the time of a given scanning, or how does the ratio of cars and bicycles has changed after a given time-frame (for example presumably as a result of new bike-lanes). This could be useful for city agencies.\\ \\To observe the environment and to generate input, depth sensors such as Light Detection and Ranging (LiDAR) sensors are commonly used besides RGB cameras, which provides depth values allowing the generation of 3D point clouds. 3D point clouds provide more information about spatial locations of different objects compared to RGB images.

Additional Files and Images

No additional files or images.

Weblinks

No further information available.

BibTeX

@mastersthesis{csecsy-fcd,
  title =      "Fast Change Detection and Tracking in Large Point Clouds",
  author =     "Csaba Csecsy",
  year =       "2026",
  abstract =   "To know the surrounding environment is required in different
               applications, such as augmented reality, robotics and
               autonomous vehicles.  While deep learning technologies allow
               advanced methods for image analysis and object detection,
               another important aspect of monitoring the environment is
               detecting changes, distinguishing removed or added objects
               and rotated or shifted objects, which is challenging
               especially in very large point clouds for example from
               outdoor environments that are observed with large epochs in
               between, even when the changed points are marked. To develop
               such applications, there are further barriers to overcome:
               changing viewpoint, occluded objects, observing objects from
               different angles, computational efficiency for a real-time
               application.\\ \\Information about changes can potentially
               be used to train deep learning algorithms for the semantic
               understanding of environments. Additionally, we would like
               to answer large-scale queries such as which objects of a
               certain class such as trees, poles or traffic signed have
               changed, or an inventory of how many cars are in a city
               during the time of a given scanning, or how does the ratio
               of cars and bicycles has changed after a given time-frame
               (for example presumably as a result of new bike-lanes). This
               could be useful for city agencies.\\ \\To observe the
               environment and to generate input, depth sensors such as
               Light Detection and Ranging (LiDAR) sensors are commonly
               used besides RGB cameras, which provides depth values
               allowing the generation of 3D point clouds. 3D point clouds
               provide more information about spatial locations of
               different objects compared to RGB images.",
  month =      jun,
  address =    "Favoritenstrasse 9-11/E193-02, A-1040 Vienna, Austria",
  school =     "Research Unit of Computer Graphics, Institute of Visual
               Computing and Human-Centered Technology, Faculty of
               Informatics, TU Wien",
  keywords =   "point clouds, change detection, change tracking",
  URL =        "https://www.cg.tuwien.ac.at/research/publications/2026/csecsy-fcd/",
}