论文部分内容阅读
针对压缩跟踪(CT)算法中存在特征单一,发生遮挡情况时易丢失目标的问题,提出基于Y通道Haar-like特征的压缩跟踪算法.为了更好地表示目标,该算法基于YUV格式图像的Y通道随机生成位置、大小的Haar-like特征;然后在预测点附近搜索目标位置,最后提出一种遮挡控制策略来缓解短暂遮挡,用卡方统计法去判断是否存在遮挡以及是否需要更新模板参数.对不同视频的测试结表明,该方法在目标存在光照变化、位置移动、遮挡的情况下,均能取得良好的跟踪效果.与原始压缩感知算法相比,本算法降低了目标中心位置的平均误差,减少了因遮挡而导致目标丢失的情况.“,”To solve the problems that CT (compressive tracking) algorithm extract few features and fail to track targets stably in occlusion,compressive tracking based on Y-channel Haar-like feature is proposed.Our algorithm describes that the target is represented with Haar-like features generated from Y channels of YUV with completely random location and size to represent the target better;and the next step is searching positive samples in the vicinity of predicted target to reduce the search range thus saving search time;at last,a block control strategy which is using Chi-square statistics to judge whether there is occlusion and whether to need to update the template parameter is proposed to alleviate short occlusion.The proposed algorithm is tested with variant video sequences and the results show that it performs favorably at the circumstances of the illumination changing,target position changing and occlusion.As compared with the origin compressive tracking (CT),this algorithm reduces the average error of the target center position as well as the circumstances of the loss of target in occlusion.