CAJ | 학술논문

通过提取人脸图像的Gabor特征,结合Adaboost进行人脸识别.针对Gabor特征维数高、冗余大的特点,引入Adaboost算法进行特征选择降低特征向量的维数,对于大量的Gabor特征进行选取.同时采用单一正样本集合和多个负样本集合分别进行训练的方法构建多个强分类器级联的层级分类器.在YaLe库上进行测试的结果验证了该法的有效性.
통과제취인검도상적Gabor특정,결합Adaboost진행인검식별.침대Gabor특정유수고、용여대적특점,인입Adaboost산법진행특정선택강저특정향량적유수,대우대량적Gabor특정진행선취.동시채용단일정양본집합화다개부양본집합분별진행훈련적방법구건다개강분류기급련적층급분류기.재YaLe고상진행측시적결과험증료해법적유효성.
Through the Gabor feature extraction of face images, face recognition is conducted with Adaboost. According to the characteristics of the Gabor features of high dimension, redundancy, Adaboost algorithm is used to reduce the dimension of feature vectors in feature selection and select Gabor features. Hierarchical classifier while using the single positive sample set and a plurality of negative samples were training set to construct multiple classifier. Test results on YaLe face database verify the effectiveness of the method.