Using Image Registration through Spatial Transformation to Reduce Vehicle Window Glare for Occupant Detection

摘要

This paper examines minimizing the effects of window glare caused by the sun during daytime dynamic image collection so as to permit enhanced counting or identification of persons in a moving vehicle. The use of spatial transformation through image registration permits obtaining improved information in an adverse collection environment caused by factors such as glare. Once the glare is removed a filter can be applied to remove such effects as a dirty windshield, targeting of moving objects can be performed through subtraction, segmentation can be applied, and a classifier can be used to count or attempt to identify persons in the vehicle.