尺度自适应CBWH跟踪算法研究

传统Mean Shift跟踪算法存在固定核窗宽导致目标尺度定位和空间定位不准确的问题。本文在背景加权的均值漂移算法(corrected background-weighted histogram ,CBWH Mean Shift)精确的目标定位基础之上，在RGB颜色空间下使用目标背景加权模型生成目标显著特征的颜色概率图，对其进行阈值分割和图像处理后获取二值图像，以此计算不变矩来调整下一帧的跟踪窗口,并在满足一定条件时及时更新背景加权模型以适应复杂背景下的跟踪任务。实验结果表明，上述方法能够自适应地更新核函数的带宽，提高了算法跟踪尺度变化目标的准确性和鲁棒性。 

Scale Adaptive Corrected Background-weighted Histogram Mean Shift Tracking Algorithm

Classic Mean Shift based tracking algorithm uses fixed kernel-bandwidth leads to scale and spatial localization inaccuracy. Based on CBWH algorithm which gives the precise location of a moving object, the proposed algorithm generates a color probability distribution with object background weighted model in RGB color space, and calculates the invariant moment to resize the tracking window of the next frame.To meet certain conditions, the background weighted model will be updated timely to adapt to the complex background of tracking. Experimental results show that the proposed algorithm improves the robustness of object tracking by self-adaptive kernel-bandwidth updating.