基于特征点匹配的步态识别

摘要

基于图像特征点匹配的算法思想,结合步态能量图(GEI),提出了一种适用于2幅GEI匹配的步态识别方法.在GEI中采用改进的FAST算法提取特征点,并采用具有良好特征描述性能的BRIEF算法描述特征点.考虑到GEI匹配不要求特征点具有旋转不变性,提出了一种质心角约束条件加速特征点的匹配.在CASIA数据库B库上的实验结果表明:方法在识别率和特征计算时间上均具有良好的表现.%A new gait energy image(GEI)-matching method for gait recognition based on algorithm of feature points in image matching is presented.Improved FAST corner detection is employed to extract the ferture points and BRIEF descriptor is used to describe them.Considering that rotation invariance is not required,a constraint condition called centroid angle constraint is proposed to speed up the matching.Experiments on CASIA Dataset B prove that the proposed approach has encouraging performance in both recognition rate and computation time.