Multiple faces tracking using local statistics

Abstract

Our project is to design algorithms to count people in vehicles such as buses from surveillance cameras' video streams. This article presents a method of detection and tracking of multiple faces in a video by using a model of first and second order local moments. The three essential steps of our system are the skin color modeling, the probabilistic shape model and Bayesian decision, and the tracking. An iterative process estimates the position and shape of multiple faces in images, and tracks them. Tracking updates an object history including all spatial and temporal information about this object. Location and size of this tracking object are predicted by constant speed motion analysis and learned trajectories. Results on office and buses video are promising.