In this paper, we propose a novel method for tracking multiple objects in video sequences. this approach can track objects in crowded scenes with occlusions and random change directions of object movements, efficiently. this method is an extension to particle filter tracking approach that we named it 3-D particle filter. in the proposed method, the features of objects that cannot be used in the process of posterior probability estimation in particle filter, are used for as third dimension and improve the estimation of 2-d particle filter in each frame. we test our method and compare it with 3 other particle filter methods on the famous data sets such as PETS09. The results show the improvement of accuracy and a reduction of error in tracking, about 12%

Keywords: video motion analysis, motion object tracking, particle filters, 3-D particle filter, single camera videos

Full-Text [PDF 1163 kb]   (1686 Downloads)