基于形变模型的红外人脸鲁棒识别*

针对人脸识别中的光照、表情和遮挡变化三大难题，引进热红外人脸克服光照变化，并且采用融合局部形变模型的人脸分类方法克服表情和遮挡变化。该方法将热红外测试人脸看成人脸库的线性组合，并用形变模型表示，通过 最小优化求解组合系数，根据系数的稀疏性进行人脸识别。为了进一步提高算法的鲁棒性，采用人脸分片加权的策略。在Equinox人脸库上通过大量实验表明：基于红外光的人脸识别性能明显高于可见光对光照变化的影响；融合局部形变模型的人脸识别方法可以有效地提高识别率且克服红外人脸识别中的眼镜干扰与表情变化问题。

Robust infrared face recognition based on deformable model

There are three main barriers toward face recognition: the variation of illumination, expression change and occlusion. A method of thermal infrared face image recognition is put forward to tackle illumination change, and an algorithm of integration local deformable model is proposed to overcome the problem of expression change and occlusion. This method cast the thermal infrared test face image as a linear combination of face database and used deformable model representation; Match optimization deformable model for solving combinatorial coefficient, according to the sparse nature of coefficients for classification. To further improve the robustness of the algorithm, use partition-based scheme. Conducting extensive experiments on Equinox databases show that the performance of face recognition based on infrared light is significantly higher than visible light face recognition for the variation of illumination; integrating local deformable model based face recognition can effectively improve the recognition rate, and can overcome the problems of glasses occlusion and expression change.