基于决策级的多源人脸融合识别

仅依赖于可见光谱进行人脸识别,在实际应用中存在很多问题。目前的研究表明,红外人脸识别也可以作为一种很好的生物鉴定技术。但是,红外图像也有自己的缺陷,其中眼镜的存在对识别的影响很大。文章提出了一种对可见光和红外人脸进行融合识别的方法。首先用PCA提取人脸主分量,计算测试样本与各类的欧氏距离,并通过构造的转换函数获得子决策;然后在决策级用D-S证据理论对子决策进行融合,得到最优决策。实验结果表明,该方法有效地提高了识别率。

Multisource Image Fusion for Face Recognition Based on Decision Level

Face recognition based on visible spectrum has many difficulties in practical application.Recent research has showed that infrared face recognition can also be a good technique for biologic identification,However,infrared image has its flaws,especially suffered from the infection of eyeglasses.A method fusing visual and infrared images for face recognition is proposed.Firstly,extracting face principal components by PCA,and computing the Euclid distances between testing sample and classes,then getting sub-decisions by a transform function.Finally,the optimal decision can be gained by using the D-S evidential theory to fuse sub-decisions.Experimental results show that recognition rate can be improved effectively.