一种改进增量非负矩阵分解人脸识别算法研究

Research on an Improved Incremental Non-Negative Matrix Factorization Face Recognition Algorithm

针对人脸面部图像辨识易受非均匀光照因素的影响,从而降低识别率问题,提出一种改进的人脸识别算法。首先,通过使用中心对称局部二阶微分模式提取人脸图像的二阶微分表征向量;其次,利用改进的增量非负矩阵分解(IINMF)算法来训练人脸图像样本的类别信息进而提取人脸图像表征向量,同时提高了算法迭代优化求解时的收敛机能;然后使用典型相关性分析(CCA)合并CS-LDP和IINMF提取的人脸面部图像表征向量获得融合后的人脸图像表征向量,最后利用最近邻分类器进行分类,获得最后的辨识结果。实验结果表明,提出的算法对光照下的识别具有较高的识别率,拥有良好的实时性和鲁棒性。

An improved face recognition algorithm is proposed to reduce the recognition rate of facial image recognition which is susceptible to non-uniform illumination.Firstly,the second order differential representation vectors of face images are extracted by using the central symmetric local second order differential mode(CS-LDP).Secondly,the improved incremental non-negative matrix decomposition(IINMF)algorithm is used to train the category information of face image samples and then extract the face image representation vector.Meanwhile,the convergence function of iterative optimization algorithm is improved.Then,canonical correlation analysis(CCA)was used to combine the facial image representation vectors extracted by CS-LDP and IINMF to obtain the fused facial image representation vectors.Finally,the nearest neighbor classifier was used for classification to obtain the final recognition results.Simulation results show that the proposed algorithm has high recognition rate,good real-time performance and robustness for illumination recognition.