WG II/4: 3D Scene Reconstruction and Analysis

Our Mission

ISPRS Working Group II/4 aims to make progress in the automatic recognition and 3D reconstruction of objects in complex scenes from images, point clouds, and other sensor data. The emphasis is on scenes characterized by the occurrence of different object classes, e.g. urban scenes including vegetation, buildings, roads, street furniture, cars and pedestrians, or indoor scenes. One of the main applications of interest of our WG is the generation of high-resolution 3D City Models, but the scope of the WG also includes any form of object detection in complex environments, e.g. in the context of robotics or autonomous driving. It is also our goal to evaluate methods for object extraction and the suitability of different sensors for that task.

To this purpose, the WG organizes workshops to exchange the latest developments on object recognition and reconstruction, collaborating closely with other WGs both from TC II and from other TCs. Furthermore, the WG will continue organizing the ISPRS labelling benchmark and intends to contribute to the review process for a Theme Issue of the ISPRS journal on subjects related to its terms of reference.

 

Working Group Officers:



Chair  

Franz RottensteinerFranz Rottensteiner
Institute of Photogrammetry and GeoInformation
Leibniz Universität Hannover
30167 Hannover
Germany
+49 511 762 3893

home page

Co-Chair  

Bruno ValletBruno Vallet
MATIS Laboratory
Institut Géographique National (IGN)
Saint-Mandé CEDEX
France
+33 1 4398 8081

home page

Co-Chair  

Markus GerkeMarkus Gerke
Institute of Geodesy and Photogrammetry
Technische Universität Braunschweig
Pockelsstraße 3
DE-38106 Braunschweig
Germany
+49 531/ 391–945 70

home page

Secretary  

Martin WeinmannMartin Weinmann
Institute of Photogrammetry and Remote Sensing
Karlsruhe Institute of Technology (KIT)
Englerstraße 7
76131 Karlsruhe
Germany
+49 721 608 47302

home page

Terms of Reference:

  • Models and techniques for extracting features, geometrical primitives and objects from data acquired by airborne and/or terrestrial sensors, including object recognition and 3D object reconstruction, and possibly integrating information about multiple object classes and their relations within complex scenes.
  • Classification and semantic segmentation of point clouds and surface meshes with or without radiometric information.
  • Generation and update of high-resolution 3D city models and road databases, including mesh based, polyhedral, parametric and multiscale representations possibly with level-of-detail (LOD) and (semantic) attributes, and texturing of the resultant 3D models. 
  • Object detection, recognition and 3D reconstruction in the context of robotics or autonomous driving.
  • Multimodal data fusion: performing any of the tasks mentioned above by exploiting the complementarity of using different viewpoints (space-borne, nadir/oblique aerial, UAV, fixed/mobile terrestrial), different sensor types (monoscopic/stereoscopic images, LiDAR, (In)SAR), and existing data (traditional cartographic products, CAD models, urban GIS).
  • Assessment of efficiency and quality and of their dependence on the quality of the input data, including uncertainty analysis and uncertainty propagation, for any of the tasks mentioned above.