Hybrid procedures for 3D perception of humans for the social assistance robotics in public and domestic application scenarios
The automated perception of people in dynamic public and domestic environments is an ability which, in addition to assistance robotics, is also fundamental in many other fields of application such as the monitoring of safety-critical areas, occupational health and safety and production assistance. Despite several decades of intensive research work, this is still a challenge, especially in real dynamic environments. There are many situations in which objects in the surrounding area change the otherwise typical human appearance, e.g. by concealment or lighting fluctuations. Atypical poses also lead to inadequate results with the current detection methods. Examples of this are people lying on sofas or on the floor.
This project deals with new hybrid approaches for the robust perception of humans which are to make use of the specific advantages of 3D sensor technology and to compensate for the various disadvantages by merging other feature types.