Tracking of wheelchair users in dense crowds

This paper proposes a tracking algorithm for wheelchair users fusing Inertial Measurement Unit (IMU) data and trajectory information of surrounding people. This approach is based on a hybrid tracking system for laboratory experiments with pedestrians in dense crowds consisting of cameras and IMUs. A camera system on the ceiling forms the basic detection and tracking system. Current experiments with heterogeneous crowds have arisen new requirements to this system. Participating wheelchair users tend to get occluded by surrounding people which leads to incomplete trajectories. To fill those temporary gaps the data of IMUs that are attached to the wheelchairs are used. The IMU trajectories are calculated with an Inertial Navigation System (INS) algorithm. The orientation is calculated with a sensor fusion filter and the distance is estimated by numerical integration of the acceleration measurements. To limit the drift the calculated velocity of the wheelchair is restricted to the velocity of the pushing person. Preliminary studies were performed to investigate this approach and resulting trajectories are presented