基于低秩约束的极限学习机高效人脸识别算法

Low-rank Constrained Extreme Learning Machine for Efficient Face Recognition

复杂应用场景中,光照变化、遮挡和噪声等干扰使得将像素特征作为相似性度量的识别算法的图像类内差大于类间差,降低了人脸识别性能。针对这一问题,提出了一种低秩约束的极限学习机鲁棒性人脸识别算法,提升了复杂场景下的识别性能。首先,利用人脸图像分布的子空间线性假设,将待识别图像聚类到相对应的样本子空间;其次,将像素域分解为低秩特征子空间和稀疏误差子空间,依据图像子空间的低秩性对噪声鲁棒的原理,提取人脸图像的低秩结构特征训练极限学习机的前向网络;最后,实现对噪声干扰鲁棒的极限学习机人脸识别算法。实验结果表明,相比前沿的人脸识别算法,所提方法不仅识别精度高、算法时间复杂度低,且具有较好的实用性。

In complex scenarios,illumination change,occlusion and noise make the image intra-variance of recognition algorithm(taking pixel feature as similarity measure)greater than the between-class variance,and reduce the performance of face recognition.To solve this problem,this paper proposed an low-rank supported extreme learning machine for robust face recognition to improve recognition performance.Firstly,the subspace linear assumption of face image distribution is used to make the image waiting to be recognized cluster to the corresponding sample subspace.Secondly,the pixel domain is resolved into low-rank feature subspace and sparse error subspace,and the forward network of lowrank structure characteristic of face image for training extreme learning machine is extracted,according to the low-rank principal of the image subspace for noise robustness.Finally,the extreme learning machine face recognition algorithm for noise robustness is realized.Experimental results show that,compared with the state-of-the-art face recognition algorithm,the proposed method not only has high recognition accuracy,but also has lower time complexity and better practicability.