加权的超像素级时空上下文目标跟踪

目标跟踪是计算机视觉中的热点问题，而目标运动的复杂背景、光照变换和尺度变化等因素大大的影响着目标跟踪的准确性。总结当前比较热门的几种跟踪算法的优缺点，针对时空上下文算法的不足提出了改进方法：加权的超像素级时空上下文目标跟踪算法（weighted super pixel level spatio-temporal context，WSSTC）。该算法利用像素的特征信息对目标上下文区域进行聚类，形成超像素级区域，并通过时间上下文中超像素块特征的相似性，对空间上下文进行加权处理，建立了超像素级的目标外观模型。实验结果表明，加权的超像素级时空上下文目标跟踪算法在目标跟踪中具有更好的准确性和鲁棒性。

Object tracking on weighted super pixel level spatio-temporal context

Object tracking is a hot topic in computer vision, and the complex background of the target motion, illumination change and scale change greatly affect the accuracy of target tracking. The paper analyzes the advantages and disadvantages of some current popular tracking algorithms, and for the lack of spatio-temporal context algorithm improvements: the tracking algorithm of weighted super pixel level spatio-temporal context. The algorithm uses the feature information of pixels to cluster the target context region, and forms a super pixel level region, and the spatial context is weighted by the similarity of the super pixel region. The experimental results show that the weighted super pixel level spatio-temporal context target tracking algorithm has better accuracy and robustness in target tracking.