Non-rigid object tracking using joint matching of local features
YU Wangshen · Journal of Xidian University · 2014
To the problem of non-rigid object tracking in video sequence,an efficient tracking algorithm based on joint matching of local features is proposed.The feature matching problem based on key points is translated into the optimization problem of balance assignment,and completed via joint matching of features under the maximization of holistic similarity.The tracking process is as follows.Firstly,local key points of the object template and current searching area are detected and described using the information on their neighborhoods.Then,the key points are marched via the proposed joint matching strategy,and the right matching pairs are picked out.Finally,the object's location and size are calculated according to the matching results,and the tracking result of the current frame is outputted.Experimental results indicate that the proposed algorithm is robust to objects non-rigid deformation,and can cope with the complex background change to obtain a stable tracking result.