摘要
针对基于测地线环的三维人脸识别算法计算复杂度过大的问题,提出一种快速识别算法。提取等距测地线环的刚性部分作为特征,利用新3DWWs度量测地线环间的空间位置关系;以图模型构造不相似度作为匹配指标,采用一种由粗到精的匹配策略对三维人脸进行匹配,从而得到识别结果。在浙大人脸库FaceWarehouse上的实验结果表明,相较于传统测地线环算法,该算法在识别速度方面提高了15倍,且识别率达97%,相较于现有的其他算法识别率较高。
A fast recognition algorithm is proposed to solve the problem of high complexity of 3 Dface recognition algorithm based on geodesic stripes.We extract rigid parts of the equidistant geodesic stripes as features,and exploit new 3 DWWs to measure the spatial position of geodesic stripes.We use the graph model to form dissimilarity as matching targets,and propose a matching strategy from coarse to fine to match 3 Dfaces,and then obtain the recognition results.Experimental results on FaceWarehouse show that our algorithm improves the recognition speed by15 times compared with the traditional algorithm based on geodesic stripes,and the recognition rate is up to 97%,which is higher than some existing algorithms.
作者
项聪颖
周大可
杨欣
Xiang Congying;Zhou Dake;Yang Xin(Department of Automation, Nanjing University of Aeronautics and Astronautics, Naniing 211106, China;Key Laboratory of Optoelectronic Control Technology, Luoyang 471023, China)
出处
《电子测量技术》
2018年第4期81-86,共6页
Electronic Measurement Technology
关键词
三维人脸识别
测地线环
三维加权走查
人脸匹配
图模型
不相似度
three-dimensional face recognition
geodesic stripes
3DWWs
face matching
graph models
dissimilarity