Research And Implementation Of Deformation Target Tracking Algorithm

Research of target tracking technology has attracted widespread attention because of its practical value in behavior analysis, guidance for military and other areas.Target tracking, detecting, extracting, identifying and tracking target in the image sequence, is the basis for further analysis and understanding targets. Because of the larger attitude change for deformation target, for example, athletes, dancers,the approach that the image in the bounding box of the fixed scale as a positive sample to learn object model is no longer applicable.In recent years, using color feature segmentation and learning object model has been successfully applied to track the deformation target and shows robustness to deformation,but that color feature is susceptible to similar background interference causes the low precision.The paper studies tracking of the deformation target in the motion background, the main work includes the following aspects: Firstly, study on the existed tracking algorithm and in order to overcome the problem of tracking drift caused by the similar background interference during tracking the deformed target, this paper proposes a tracking method based on saliency segmentation and target detection. By way of saliency segmentation via graph-based manifold ranking in the super-pixel level, the method achieves significant, high-quality target pixels and uses the quantized values of these pixels color and gradient, and the relative positional relationship with the center of the target to denote the target model, which is used for tracking target by detecting the target center position. With the global color features as the drift constraints corrects for the central position drift because of target continued deformation. During tracking, the tracking results for constraint of saliency segmentation and global color features and update the target model with saliency segmentation results.Secondly, the proposed algorithm is implemented in IDE Visual Studio 2010, with C++ and Open CV,and the main modules include the saliency segmentation, the target model management, the color feature management and the performance analysis.Finally, the paper selects the public test data sets and compares with other algorithms on precision and accuracy through detailed qualitative and quantitative experiments. The experimental result indicates that the proposed method improves the precision and stability of tracking the deformed target and has the stronger robustness.