A multi-target tracking approach combined with occlusion segmentation

Multiple-target tracking in complex scenes is one of the most complicated problems in computer vision. Handling the occlusion between objects is the key issue in multiple target tracking. This paper presents an occlusion segmentation-based method to track multiple people in complex situations which are captured by static monocular cameras. In the proposed method, we calculate the probabilistic histogram of each object´s optical flow vector, then use this motion statistic information along with the color and appearance information to construct a new expression of pixel distance. Finally, a stepwise classification and K-means clustering method are taken advantages of to accomplish occlusion segmentation. Object tracking is handled by a particle filter-based tracking framework, and a probabilistic appearance model is used to find the best particle. Experiments are conducted using public challenging data set PETS 2009. Results show that our approach can improve the performance of the existing tracking approach and handle dynamic occlusions better.