基于ORB特征点匹配的多目标跟踪算法

Multi-target Tracking Algorithm Based on ORB Feature Points Matching

视频中的多目标跟踪是计算机视觉领域的一个重要问题.针对多目标跟踪过程中由于目标缩放、旋转、扭曲以及遮挡等问题的存在导致目标易丢失的问题,提出了一种基于ORB特征点的多目标跟踪算法.首先利用"运动边缘生长"算法得到目标块,再利用目标块的ORB特征与特征模板的匹配来实现目标的关联.通过匹配过程,目标模板能够得到实时更新,即去除由噪声带来的过时的特征信息并添加进新的特征信息,保证特征模板的实时有效性,进而提高跟踪过程中匹配的可靠性.实验结果表明,本算法能够实时有效地处理目标由于形变以及局部遮挡而导致跟踪性能下降甚至跟踪目标丢失的问题,具有较好的鲁棒性.

Video multi-target tracking is an important problem in the field of computer vision. In order to solve the problem where the target is easy to lose due to target scaling, rotation, distortion and occlusion, this paper proposed a multi-target tracking algorithm based on ORB feature point. First, the ＂moving edge growth＂ algorithm was used to obtain target blocks. Next, targets were associated by matching the ORB features of target blocks with feature templates. The matching process enables target templates to be updated in the real time by eliminating the old feature information from the noise and adding new feature information. In this way, effectiveness of feature template is guaranteed in the real time, resulting in greater reliability of matching during the tracking process. Experimental results demonstrate the ability of the proposed algorithm to robustly track targets even if the targets are deformed or partially occluded.