一种基于ORB特征匹配和反投影直方图的目标跟踪方法

A Tracking Method Based on ORB Feature Matching and Histogram Back Projection

针对目前许多跟踪算法跟踪效果随跟踪时间增加而变差的问题,提出一种基于ORB特征匹配结合反投影直方图的跟踪方法;首先提取ORB特征,使用匹配策略对目标和选定模板的ORB特征进行匹配,对比了暴力匹配和局部敏感哈希匹配两种匹配方式;然后提取模板的颜色信息,利用直方图反投影的方式去除在视频图像中与所选目标区域颜色不相关的部分,对比了使用ORB通道的颜色和H通道的颜色信息时跟踪的效果。通过仿真实验和对比,说明这种方案有直观的跟踪效果,而且有一定的实时性。

In allusion to the shortcoming of many tracking algorithm that tracking effect become poor with the time passing, a tracking method based on the ORB feature matching and histogram back projection was proposed. Firstly, the ORB feature was extracted and marching strategy was used to match the ORB feature of the selected target template and target. The brute force （BF） and LSH marching strategy were contras- ted. Secondly, the color information of the target template was extracted and the part of video frame which is uncorrelated with the color of the selected target area was wiped out by the histogram back projection. The tracking effect of using the information of ORB channels and the Hue （H） channel was contrast. The experiment indicates the combined method can obtain a visual tracking result in real time.