打卡图像的分层人脸比对识别方法研究

Research on the Method of Hierachical Human Face Identification for the Image with Identity

具有身份的人脸图像比对要求具有高的识别率和实时性。本文针对打卡人脸图像,提出了一种基于加权模板匹配和SVM的分层人脸识别方法。该方法利用Gabor小波变换进行人脸图像特征提取,采用贡献分析法分析特征的贡献权重,在待测人脸图像比对识别时,采用加权模板匹配进行比对,通过两个阈值的判断,在既不能认为比对正确和不正确的情形下,再采用SVM和库中人脸图像进行识别比对。实验结果表明,基于该方法的人脸比对识别率高、实时性好,可用于实时打卡人脸比对。

Human face image contrast requires a high recognition rate and the real-timing.The article in allusion to the clocking image of human face,proposing a recognition based on the elastic template matching and stratified human face.This method,using the Gabor wavelet transformation,carries on the feature extraction of human face image.And it analyses these features to weight the contribution of the personalized human face by the contribution analytic method.When the human face images which are waiting for test are comparing and recognizing,using the weighted template matching to compare,through the judgment of the two thresholds costs,and when it＇s hard to identify whether the contrast is correct or incorrect,we can reuse SVM and the human face images of the database to identify and contrast.The experimental result shows that the human face contrast based on the method has a high recognition rate and good real-timing,so as to used for the real-time clocking human face contrast.