复杂环境下高效物体跟踪级联分类器

目的 传统跟踪算法在复杂环境下容易发生漂移（drift）现象，通过改进TLD（tracking learning detection）跟踪技术算法，提出了基于Sliding-window的局部搜索和全局搜索策略、积分直方图过滤器和随机Haar-like块特征过滤器。方法 首先，采用积分直方图过滤器可以有效地过滤大量非目标子窗口块，从而减少后续过滤器特征匹配数；其次，利用随机Haar-like块特征过滤器能够解决跟踪算法在复杂环境（多物体、部分或较大区域遮挡、快速运动等）跟踪过程易发生漂移而导致跟踪精度的不足。结果 结合TLD原始过滤器与新提出的两个过滤器组合而成的级联分类器，通过与主流的跟踪算法实验进行对比表明，级联分类器在稳定的背景或复杂环境的跟踪鲁棒性强、跟踪精度高，并且采用了局部和全局搜索策略提高了计算速度。结论 提出的方法在诸多背景环境变化，跟踪物体形变等情况下，能够精确地多尺度跟踪待测目标；结合全局和局部搜索跟踪策略能够有效地克服级联分类器所带来的时间复杂度过高的问题，从而实现实时目标跟踪。

Efficient cascade classifier for object tracking in complex conditions

Objective We improve the TLD algorithm and propose local and global search based on the sliding-window method, the Integral Histogram Filter, and Random Haar-like Feature Filter to solve the drift problem of traditional tracking algorithms in complex conditions. Method First, we use the Integral Histogram Filter to reject the Sliding-window patches as quickly as possible to release the feature matching in the following filters. Then, we use Random Haar-like Feature Filter to overcome the drift problem, which causes a loss of accuracy during the object tracking under complex conditions (multi-object, occlusion, fast movement). Result We ultimately combine filters of the TLD algorithm and two new filters of our proposed. The experimental results show that the proposed approaches compared with the traditional tracking algorithms not only presents robustness and tracking accuracy in stable background or complex conditions, but also obtains the best computing speed with the use of the local and global search. Conclusion The proposed method is able to detect the multi-scale object accurately both in different environment and tracking object deformation. Combining the global and local search strategy can overcome the time consuming effectively to achieve the real-time object tracking.