Mean Shift Tracking Algorithm Based On Graph Cuts Image Segmentation

Object tracking is to find the most similar parts to the targets in video sequences, which is an important topic in computer vision. Object tracking is used in many fields such as military, space technology, supervisory control, biomedicine and robot. In recent years, it has achieved considerable success. Many tracking algorithms are proposed to solve the problems of object tracking. Among the various tracking algorithms, Mean Shift tracking algorithm has become popular due to its simplicity, good real-time and efficiency.Mean Shift tracking algorithm is a modeling mechanism based on statistical probability density function (pdf). In the process of tracking, the target area is selected in the first frame by user, and then the corresponding target model can be established. In the subsequent frames, the best candidate areas are searched iteratively by Mean Shift algorithm based on the Bhattacharyya similarity function. The Mean Shift tracking algorithm shows good performance, but when the target and background are quite similar or the target moves fast, it will be difficult for the Mean Shift tracking algorithm to find the best candidate areas, even the target will be lost.In this paper, the Mean Shift object tracking algorithm based on Graph Cuts image segmentation is proposed to solve the defects mentioned above. The new algorithm uses image segmentation in each frame of the video sequences. The image segmentation algorithm based on Graph Cuts is quite fast in computing and gives the precise segmentation. For the videos in which the target and background are quite similar, the new algorithm separates the target from background effectively, makes the target model and candidate model more precise, then improves the performance of tracking. The experimental analysis and comparison between the Mean Shift algorithm and the new algorithm prove that the new algorithm shows good robustness for the video sequences in which the target and background are quite similar or the target moves fast.