基于门禁系统的人脸识别算法——Gabor小波 变换

摘要

Gabor小波变换对人脸图像上的部分结点(基准点)进行多尺度、多方向的Gabor小波变换，提取人脸的纹理特征，然后再对提取的Gabor特征进行后续的特征选择、降维、特征再提取以及分类识别等处理。结合弹性束图匹配算法，利用人脸的基准特征点构造拓扑图，保留了局部特征的拓扑结构作为整体特征。将模拟进化算法与基于局部特征的人脸识别方法相结合，整体的拓扑结构约束作为进化算法中的整体优化，局部特征的相似度对比作为进化算法中的局部优化，即可完成人脸面部特征点的正确匹配，完成高精度的人脸识别。在检测过程中只要抓住基准点，就可保留总体人脸信息，同时局部特征也得到了增强，这样可减少运算量，提高识别率和识别速度。 Gabor wavelet transform of face images on the part of the node (reference point) did the Gabor wavelet transform with directions, and extracted the texture feature of human face firstly, then extracted the Gabor feature selection and feature for subsequent dimension reduction and feature extraction and classification recognition processing. Combined with elastic graph matching algorithm, beam of face of the base feature points topology structure, keeping the local features of topology as a whole, which was simulated evolutionary algorithm combined with a face recognition method based on local characteristics, the overall topology constraints as whole optimization in the evolutionary algorithm, the local characteristics of similarity comparison as the local optimization evolutionary algorithm, could complete the correct face facial feature points matching, complete the high accuracy of face recognition. Just caught in the process of testing datum which could retain the overall face information, local characteristics had been enhanced at the same time, this could reduce the computational complexity and improve the recognition rate and recognition speed.