基于HSV模型和特征点匹配的行人重识别算法  

Person re-identification algorithm based on HSV model and keypoints matching

作　　者：彭志勇[1] 常发亮[1] 刘洪彬[1] 别秀德 

机构地区：[1]山东大学控制科学与工程学院,山东济南250061

出　　处：《光电子．激光》2015年第8期1575-1582,共8页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金(61273277);高等学校博士学科点专项科研基金(20130131110038);教育部留学回国人员科研启动基金(20101174);山东省自然科学基金(ZR2011FM032)资助项目

摘　　要：提出了一种基于HSV模型和特征点匹配相结合的行人重识别算法。首先根据改进的HSV空间颜色量化策略,比对两幅行人图像的躯干和腿部主颜色是否一致,以快速确定备选目标;然后对备选目标,利用环形Gabor滤波器组生成多尺度图像,再利用改进的FAST算法和BRIEF算法对多尺度图像进行特征点提取与描述,最后利用暴力算法和随机抽样一致性算法进行特征点匹配和提纯,以达到较好的匹配效果。实验结果表明,本文提出的行人重识别算法具有较高的识别准确率,识别速度达到12frames。Abstract：In the application of video content analyses and multimedia retrieval, person re-identification is a critical technique, which has great 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 rule out images whose main colors are different from the target, and then tests remaining images by matching keypoints. Then, the method of pre-testing pedestri- an images compares main colors of the torso and legs of two images according to an improved color quantization strategy of HSV space, which achieves better pre-testing effects and raises recognition speed of the algorithm. The method of keypoints matching first takes advantage of circularly symmetrical Ga- bor filters to generate multi-scale images, then extracts and describes keypoints by improved FAST and BRIEF algorithms, matches and purifies keypoints by Brute force algorithm and Random sample consen- sus algorithm, which achieves better matching effects than SIFT, SURF and ORB algorithms. For testing the validity of the proposed algorithm,we establish a pedestrian image library including 600 images. Ex- perimental results show that the proposed algorithm can identify pedestrian targets accurately by the speed of 12 frames per second.

关 键 词：行人重识别 HSV模型 特征点匹配 环形Gabor滤波器 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]