摘要
人体检测与定位是计算机视觉领域的研究热点与难点之一,其被广泛应用于人机交互和人机协作等诸多领域。提出一种基于区域和关键点特征相结合的双目视觉人体检测与定位方法,首先利用HOG与SVM相结合的检测框架实现人体的检测,然后利用MSER算法提取人体区域的MSER特征,最后利用局部不变算子ORB描述MSER区域的特征,进而实现立体匹配与人体的定位。实验表明本文提出的方法不仅能缩小特征检测的范围,而且能提高特征之间匹配的准确度。
Human detection and location is one of the research hotspots and difficulties in the field of computer vision. It is widely used in many fields such as human-computer interaction and human-computer collaboration. A method of human detection and location based on binocular vision combined with region and key points is proposed in this study. First of all, HOG combined with SVM detection framework is adopted to realize the human detection. Then, the MSER algorithm is applied to extract feature of the human. Finally the ORB local invariant operator is used to describe MSER region feature, and ultimately realize stereo matching and positioning of the human. The experiments show that the proposed approach not only narrows the scope of feature detection, but enhances the accuracy of the matched features.
出处
《北京联合大学学报》
CAS
2015年第3期38-43,48,共7页
Journal of Beijing Union University
基金
国家科技支撑课题(2014BAK08B02
2015BAH55F03)
国家自然科学基金(61271369
61372148)
北京市自然基金(4152016
4152018)
北京联合大学"人才强校项目"(BPHR2014E02
BPHR2014A04)