An Improved Multi-Target Tracking Algorithm for Pedestrian Counting
Tian Xia, Hong Fan, Suping Yu, Liping Zhang, Jiajing Wen · Journal of Physics Conference Series · 2018
In view of the traditional Camshift algorithm is easy to lose the target in the case of serious colour interference and occlusion, and the traditional method of people counting is inefficient and has low accuracy. This paper uses three-frame difference method and Gaussian Mixture Model to detect the pedestrians, and then an improved multi-target tracking algorithm is used to implement automatic tracking of multiple pedestrians and draw a motion trajectory. Based on this, the motion direction is judged and pedestrians are counted. In the tracking process, the problem of colour interference and severe occlusion is solved according to the trajectory prediction. The experimental results show that the proposed algorithm not only improves the tracking accuracy, but also realizes the automatic pedestrian counting. It also greatly improves the efficiency of the number of people's statistics and can accurately count the number of people in the presence of overlap and occlusion.