基于改进SRC的局部遮挡人脸识别方法

Local Occlusion Face Recognition Method Based on Improved Sparse Representation Classification

针对现实场景中人脸存在较为随机的局部遮挡的情形,创建了多种遮挡类型并存的过完备字典,并提出改进的稀疏表示分类方法。此算法在稀疏表示分类(Sparse Representation Based Classification,SRC)的基础上引入噪声补偿,通过对测试图像与去噪图像间噪声差的调节来控制重构图像的生成,进而实现对重构图像与测试图像间残差的控制,以达到准确分类识别的效果。实验结果表明,改进后的算法稳定性好、识别率高,对于局部遮挡图像的识别有较好的鲁棒性。

In view of the fact that there are relatively random partial occlusions in real scenes,a complete dictionary with multiple pick types is created.,and proposed an improved sparse representation classification method.The algorithm in this paper introduces noise compensation on the basis of Sparse Representation Based Classification(SRC),which controls the generation of reconstructed images by adjusting the noise difference,and then realizes the control of residual difference between reconstructed images and test images to achieve accurate classification and recognition effect.The experimental results show that the improved algorithm has good stability,high recognition rate and good robustness for local occlusion image recognition.