摘要： 针对部分遮挡人脸识别，提出了一种基于局部区域遮挡判断的人脸识别新方法。利用支持向量机对不同人的不同遮挡情况分别进行训练，确定局部相似度到整体相似度的最佳映射，使它具有最佳类间可分离性。实验表明，该文算法显著提高了部分遮挡人脸识别效果，对无遮挡识别也有所改善。

关键词: 人脸识别, 部分遮挡, 支持向量机, 局部遮挡判决

Abstract: This paper presented a novel partially occluded face recognition method with sub-region occlusion judgment and sub-region similarity fusion. In this work, people trains support vector machines for all occlusion conditions of all reference persons, through which they get the optimal function that composes sub-region similarities into the overall similarity, which guarantees optimal separability between classes. Experiments show that the method improves the recognition performance for both partially occluded and non-occluded faces, especially for the former.

中图分类号: 

• TP391.41