Real time driver body pose estimation for novel assistance systems

摘要

The trend towards increasing automation turns the drivers of modern cars more and more into passengers. This increases the scope of potential secondary tasks and increases distraction. For safety reasons cars should therefore be aware of the situation in their interior to be able to evaluate the drivers readiness to assume control. We present an algorithm that can accurately detect the 3D upper body pose of the driver in real time with a single depth camera. We show that it can determine the upper body pose in various real driving situations and with different secondary tasks. We propose that our algorithm can be a building block for various advanced driver observation systems for automated cars.