改进的线段Hausdorff距离人脸识别方法

Improved Line Segment Hausdorff Distance for Face Recognition

针对传统的线段Hausdorff距离(LHD)方法在人脸识别中的缺陷与不足提出了几点改进:把activeshape模型(ASM)得到的人脸配准的结果结合到LHD线段边缘图(LEM)的提取过程,得到了比较好的结果;对人脸不同部分的线段给予不同的权重,并且把线段对的匹配限制在给定的邻域内。实验结果表明,改进的LHD方法比传统的LHD方法在微笑和小角度姿态变化下的识别率分别提高了10.3%和10.0%。

Several improvements on the conventional line segment Hausdorff distance (LHD) for face recognition were presented. Firstly, the result of active shape model (ASM) for face alignment was combined into the extraction procedure of line edge map (LEM). Secondly, different weights were given to lines in different part of face and the match of line pairs were limited in a given neighbor area. Experiments show that the recognition rate of proposed method is 10.3% and 10.0% higher than that of the conventional LHD method under the conditions of both smile and different poses.