Pedestrian Detection and Tracking for Counting Applications in Crowded Situations

This paper describes a vision based pedestrian detection and tracking
 system which is able to count people in very crowded situations like
 escalator entrances in underground stations. The proposed system
 uses motion to compute regions of interest and prediction of movements,
 extracts shape information from the video frames to detect individuals,
 and applies texture features to recognize people. A search strategy
 creates trajectories and new pedestrian hypotheses and then filters
 and combines those into accurate counting events. We show that counting
 accuracies up to 98 % can be achieved.