Abstract:

To solve the problems in non-rigid object tracking, such as the significant and rapid variation in shape as well as appearance,an efficient tracking algorithm based on joint matching of SIFT features is proposed in this paper. Firstly, local SIFT key points of the object template and the current searching area are detected and described. Then, the key points are matched via the improved 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. When constructing the similarity matrix, the improved joint matching strategy exploits the rotation and scale invariance of SIFT features, and considers the spatial information of the key points, which effectively enhances the feature matching accuracy. This paper successfully introduces the improved matching strategy to the SIFT matching tracking, which overcomes the defect of the traditional SIFT matching algorithm in non-rigid object tracking. The experimental results indicate that the proposed algorithm is robust to object’s non-rigid deformation, scale change and background distraction.