3DPersA - Hybride Verfahren zur 3D Personenwahrnehmung für die soziale Assistenzrobotik in öffentlichen und häuslichen Einsatzszenarien

  • Duration:  01.07.2015 - 30.06.2019
  • Funding:  BMBF - Bundesministerium für Bildung und Forschung im Rahmen des Graduiertenforschungskollegs "3D-Technologien in der Mensch-Maschine-Interaktion" (Forschungsallianz 3Dsensation)

The automated perception of people in dynamic operational environments is a capability that is of fundamental importance not only in assistance robotics but also in many other application areas, such as the monitoring of safety-critical areas, occupational safety and production assistance.

This project is to deal with new hybrid approaches for the robust perception of persons, which are to use the specific advantages of 3D sensor technology and compensate for the various disadvantages by fusion with other feature maps. The main sub-problems to be addressed are depth data-based person detection including person tracking in suitable state spaces, person recognition using the depth data of learned person models, and sensor fusion of depth data with other feature maps.

The goal is to create a person perception system for domestic and public use scenarios that is suitable for the real world and that reliably recognises persons even in situations for which current approaches only provide insufficient results.