Real-time occupant classification in high dynamic range environments

This paper presents a visual occupant classification system for vehicles. An illumination technique is described for image processing in environments which are characterized by large intensity fluctuations and hence a high optical dynamic range (HDR), in particular for vehicle interior surveillance. The method works by combining a set of images flashed with different radiant intensities. With a constant exposure time for the imager, a single image can be produced with a compressed dynamic range and a simultaneously reduced offset. This enables it to capture a vehicle interior under adverse light conditions without using a high dynamic range camera and without losing image detail. Furthermore, a real-time segmentation approach for stationary cameras is presented which is neither affected by sudden illumination changes nor shadow effects. This approach analyses the frame ratio instead of performing the usual background subtraction.