An Unsupervised Algorithm for Video Pedestrian Detection and Estimation

In this paper,we proposed an algorithm to detect pedestrians from a video scene using finite mixture models(FMM).We first derived the background image from the video by averaging multiple frames and then subtracted the background from certain video scene to derive the moving pedestrians.Finally,we imported a learning method of the FMM to detect the pedestrians and estimate the number of pedestrians from the parameters of the FMM.The final result showed the performance of the proposed algorithm.