基于HSV模型和特征点匹配的行人重识别算法

In the application of video content analyses and multimedia retrieval, person re-identification is a critical technique, which has great realistic significance in the field of criminal investigation. This paper proposes a person re-identification algorithm based on HSV model and keypoints matching. It first utilizes HSV model to pre-test pedestrian images and quickly rules out images of which main colors are different from the target, and then tests remaining images by matching keypoints. The method of keypoints matching takes advantage of circularly symmetrical Gabor filters to generate multi-scale features, extracts and matches keypoints by improved FAST and BRIEF algorithms, which achieves better matching effects than SIFT algorithm. Experimental results show that the proposed algorithm can identify pedestrian targets accurately by the speed of 12 frames per second.