Aiming at the problem of face recognition under partial occlusion,an HOG feature-based face recognition algorithm based on dual attribute model was proposed.The face image was reconstructed by linear regression to get the error image.Then,the dual attribute model was used to fuse the feature vector information of the error image and the global feature information to obtain the dual attribute feature vector information.At the same time,the HOG operator was used to segment the target image evenly to extract the local HOG feature information.The dual attribute eigenvectors and partial block vectors were weighted by the whole classifier.Experiments in AR and Yale-B face database proved that this algorithm had the characteristics of strong robustness and high recognition rate for different regions of human face and different levels of occlusion recognition.%针对局部遮挡条件下的人脸识别问题,提出一种基于双属性模型的HOG特征人脸识别算法.采用线性回归分类对人脸图像进行重构得到误差图像;通过双属性模型将误差图像的特征向量信息与全局特征信息进行融合,得到双属性特征向量信息.同时,采用HOG算子对目标图像进行均匀分块提取局部的HOG特征信息;通过整体分类器对双属性特征向量和局部分块向量进行加权分类.在AR和Yale-B人脸数据库实验证明该算法对人脸的不同区域、不同程度的遮挡识别具有鲁棒性强、识别率高的特点.
展开▼
