基于离散模板精匹配的人脸识别算法

通过离散多模板算法可实现人脸图像识别的精匹配.先检测人脸图像边缘,并对边缘所有点进行最小平方椭圆拟合.根据嘴部的宽度和高度比例,给定域值,确定嘴部候选区域.然后运用训练过的支持向量机(SVM)分类器验证嘴部区域并进行嘴角定位.再采用离散多模板方法对嘴部进行匹配识别.在自建库和ORL库上试验表明,在验证率为91.2%、95.3%情况下,该方法可获得92%、97%的识别准确率.

Face Recognition Algorithm Based on Discrete Multi-Template Matching

The algorithm of discrete multi-templates realizes accurate matching of face recognition. Firstly edge extracting of face images is applied, then least-square ellipse fitting is performed on all detected edge points in face images According to the proportions of mouth width and mouth height, a threshold is given to confirm the regions waiting for selections, then a trained support vector machine classifier is applied to identify mouth regions, and localizing algorithm on corners of the mouth is used in corresponding regions. At last, discrete multi-templates are put on to accomplish the matching and recognition of face images. The experiments are carried on self-building library and ORL library. The results indicated that, while the certification of support vector machine classifier reaches 91.2% and 95.3%, the corresponding recognition can achieve 92% and 97%.