目标遮挡情况下的压缩跟踪算法

The compressive tracking under occlusion

针对经典压缩跟踪算法在目标被遮挡时容易导致目标丢失的问题,提出了一种基于目标遮挡情况下的压缩跟踪算法。该方法首先依据分类器的最大响应值判断目标是否被遮挡。若发生遮挡则利用基于颜色直方图特征的粒子滤波算法进行跟踪预测,即将遮挡前提取的目标颜色直方图与粒子的颜色直方图进行相似性比较。为确保目标再现时能及时准确地捕捉其位置,再利用Harris角点特征进一步验证,并将预测的位置作为目标位置继续压缩跟踪。仿真结果表明,该算法能够准确地判断遮挡的发生,平均跟踪成功率较经典的压缩跟踪算法提高了24%,有效提高了跟踪的鲁棒性。

In view of the problem that the classic compressive tracking algorithm leads to the loss of target easily when the target is under the occlusion condition,a compressive tracking algorithm under occlusion is proposed. Firstly,the method judges the occlusion according to classifier’s maximum response; if the occlusion occurs,use particle filter algorithm based on the color histogram feature for tracking prediction,and compare the color histogram of target that extracted before occlusion with particle’s color histogram.Then use the Harris corner features to further validate to capture the location of the object timely and accurately until the target reappears,and regard the predicted location as the target position to compressive tracking. The simulation results show that the proposed algorithm can judge the occurrence of occlusion accurately. The average tracking success rate is 24% higher than that of the classical compressive tracking algorithm,which improves the robustness of tracking effectively.