For object tracking, the mean-shift algorithm with texture feature fusing has fine robustness and real-time ability, but it does not work well when occlusion takes place. Although with particle filtering algorithm this problem can be solved easily, its real-time performance was not good. So the fusion of the advantage of these two algorithms is logical and has been implemented. An algorithm for anti-occlusion object tracking with multi-cue fusing was presented, where the particle filtering algorithm would be switch to when object occlusion took place. However, there was no normalized simulation platform for comparing and evaluating the various algorithms so far. Therefore, a test platform of anti-occlusion object tracking was presented based on Matlab GUI, on which several identical sequences were tested for their comparison to give out a simulation result. It was shown by the result that this algorithm could deal with the occlusion in object tracking effectively and that this simulation platform would not only provide the comparison of different algorithms but also have good mutuality and extensibility. It also could be used to realize centralized management of various algorithms and output visual test results, fully showing the effectiveness of the platform design.%针对目标跟踪中的遮挡问题,融合纹理特征的mean-shift算法鲁棒性实时性好,但遮挡情况下无法有效跟踪,粒子滤波算法能很好地解决这一问题,但实时性不好,因此,融合两种算法的优点,遮挡情况下切换到粒子滤波算法.同时如何比较、评价各种算法,目前还没有一个统一的仿真测试平台,为此给出一个基于Matlab GUI的抗遮挡目标跟踪系统仿真平台,在该平台下,对同一序列进行比较试验,给出仿真实例.结果表明,该算法能有效处理目标跟踪中的遮挡难题,同时该平台不但提供给了统一的界面进行仿真比较,还有很好的交互性与可扩展性,能够实现不同算法的集中管理,并输出可视的实验结果,充分验证了平台设计的有效性.
展开▼
