CAJ | 학술논문

为了克服因人脸图像检测引起的配准不稳定性和小样本引起的维数灾难,由一副二维人脸图像通过上下左右平移生成4个图像,把生成的图像与原来的图像一起加入训练样本集,构成新的训练图像集.基于二维图像,结合图像局部结构信息,设计了准则函数,获得双投影矩阵,抽取人脸特征.对待识别人脸图像,由它的扰动图像设计识别方法.与传统的人脸识别方法相比,该方法的识别效果更好;Yale和 ORL 人脸数据库上的实验结果验证了该方法的有效性.
위료극복인인검도상검측인기적배준불은정성화소양본인기적유수재난,유일부이유인검도상통과상하좌우평이생성4개도상,파생성적도상여원래적도상일기가입훈련양본집,구성신적훈련도상집.기우이유도상,결합도상국부결구신식,설계료준칙함수,획득쌍투영구진,추취인검특정.대대식별인검도상,유타적우동도상설계식별방법.여전통적인검식별방법상비,해방법적식별효과경호;Yale화 ORL 인검수거고상적실험결과험증료해방법적유효성.
@@@@In order to overcome the instability caused by face detection and curses of dimensionality resulted from small sample size, four new images from one two-dimensional image is generated by horizontal and vertical translation, and a new image data-base is formed by adding these new images to original image database. Combined with local structure information, this paper designs objective function and obtains dual projection matrix based on two-dimensional images. The proposed method is carried out on disturbing image database. Compared with the traditional face recognition method, it has a better recognition perfor-mance, and the experimental results on Yale and ORL face image database show that it is effective and robust.