Research On Particle Filter Pedestrian Target Tracking Algorithm Based On Occlusion Detection

The object tracking technology has become one of the hottest topics in the field of computer vision in the last decades.Many scholars have done a lot of research on the object tracking technology to promote its rapid development.When applied to video sequences in real-life situation,trackers should cope with numerous difficulties caused by occlusions,similarities between objects,illumination changes and background clutter.Among many challenges in visual tracking,occlusion is one of the most intractable issues.This paper mainly studies the pedestrian target tracking problem in complex occlusion scenes,and its specific research contents are as follows:In order to avoid the tracking drift produced by occlusion,a pedestrian target tracking algorithm based on the variance of reconstruction error(VREPT)is proposed.Under the framework of particle filtering,the proposed algorithm uses principal component analysis(PCA)to model the appearance of the target;And then,to avoid the tracking drift,occlusion detector is introduced in the tracking algorithm.We detect whether the target is in occlusion with the reconstruction error variance in the tracking results of each frame,and then to handle occlusion issue according to the test results;When the target is judged to be in occlusion state,the target template will not be updated.And when the target is judged to be not occlusion,an incremental approach is used to study and update the target appearance model.The experimental results show that the proposed algorithm can effectively deal with occlusion,and can track the target accurately and stably.Aiming at the problem of rapid detection of target occlusion in pedestrians target tracking,a pedestrian tracking algorithm based on high order cumulant(HOCPT)is proposed.In the proposed algorithm,the three order cumulant of the reconstruction error is constructed to detect the occlusion because high order cumulant can effectively suppress the Gauss noise.And then,the occlusion handling scheme is used to solve the problem if the target is in the occlusion state.The experiments revealed that that the proposed algorithm can accurately detect the time when the target is in and out of occlusion,and is impactful to deal with occlusion.