单样本三层人脸图像智能识别仿真研究  

Simulation Research on Intelligent Recognition of Three Face Images With Single Sample

作　　者：黄丽达[1,2] 陈燕[1] 魏源良 

机构地区：[1]广西大学计算机与电子信息学,广西南宁530005 [2]广西艺术学院,广西南宁530022

出　　处：《计算机仿真》2018年第1期361-364,共4页Computer Simulation

摘　　要：对于单样本三层人脸图像的高效识别,能够有效实现人脸智能识别技术。对单样本三层人脸图像的智能识别,需要先给出人脸图像序列线特征的提取过程,进行图像的直线匹配,完成对人脸单样本三层图像的识别。传统方法从人脸图像中提取五官轮廓,对其进行规则化处理,但忽略了对人脸图像进行直线匹配,导致识别精度偏低。提出一种基于直线特征匹配的人脸单样本三层图像智能识别方法。对人脸图像进行小波尺度分解,进行Harris自适应阈值角点检测,给出人脸图像序列线特征的提取与拟合过程,结合人脸图像直线特征的角度及灰度近似性测度等约束完成人脸图像的直线匹配,利用面-面相交原理实现序列图像中人脸特征直线的单样本三层图像重建。实验仿真证明,所提方法具有较好的稳定性和可扩展性,并且识别精度较高。ABSTRACT：An intelligent recognition method for three-layer face images with single sample is proposed. Firstly, the wavelet-scale decomposition for face image is carried out and Harris comer of adaptive threshold is detected. Then extraction and fitting process of linear feature of face image sequence is provided. Moreover, the linear matching of the face image integrated with constraint such as angle of the linear feature and measurement of gray level similarity is completed. Finally, reconstruction of the three-layer face image of characteristic line in sequence image is a- chieved. Simulation results prove that the method has good stability and expandability, and high recognition preci-

关 键 词：三层图像 人脸图像 智能识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]