一种基于序贯检测机制的运动目标跟踪算法

摘要

为了克服单个目标跟踪算法在复杂环境下跟踪精度不高的问题，提出了一种基于序贯检测机制的运动目标跟踪算法。该算法首先用基于颜色特征的粒子滤波估计最优跟踪窗口；以此跟踪窗口和目标的相似度决定是否采用稀疏场主动轮廓方法，又以目标轮廓和目标的相似度决定是否需要Camshift对轮廓进行修正。实验表明，对比基于颜色的粒子滤波和Camshift方法，本文方法能够在不同的复杂环境和目标有尺度、旋转、视角等姿态变化的情况下，具有更好的跟踪精度和鲁棒性。【其它文摘A single method for object tracking has low accuracy in a complex environment. To overcome this limitation, a novel object tracking method is studied in this paper, based on sequential detection scheme. First, particle filter object tracking approach with color feature was applied to estimate optimal tracking window, and sparse field active contours is performed or not based on the similarity between object and resulting window, and then the similarity between object and the obtained contour is calculated in sequence to determine whether Camshift should be employed to modify the object contour. And thus the accuracy of moving object tracking will be greatly improved. Experiments demonstrated that compared with the classic particles filter object tracking approach with color feature and Camshift algorithm, the proposed algorithm can track the moving target successfully in different scenarios and it can handle target with scale, orientation, and view changes; moreover, it has better robustness and accuracy of object tracking.