A Fast Tracking Algorithm for Multiple- Objects in Occlusions

摘要

Occlusion problem is one of difficult points in moving target tracking, which also is one of key techniques to intelligent vision. Based on moving objects detection, minimum enclosing rectangle of human-body top half area is selected as tracing window, which can eliminate the interference of ground shadow, avoid the disturbing of leg movement, and can reduce the amount of computation. Then color histogram of tracing window is calculated by use of the simplified color space, which combines minimum distance matching is used for target matching under non-occlusion; if occlusion occurs, color histogram matching and trajectory prediction is adopted for target matching, with this method, moving object can be tracked correctly under serious occlusion. The experimental results demonstrate that the algorithm is rapid and robust, it can effectively overcome multi-targets tracking in complex scenarios.