二次近邻稀疏重构法及人脸识别

发布时间：2018-10-26 19:20

【摘要】：基于整个数据集的稀疏表示(sparse representation classification,SRC)用于人脸识别在很大程度上影响了运行效率。如何利用较少样本稀疏表示在保证计算效率的同时,识别率也有一定提升,尤其是面对光照、角度、姿态等非受控环境,目前仍是一个问题。考虑到协同表示(collaborative representation classification,CRC)基于l2范数稀疏求解的优势,为进一步提升CRC的整体分类性能,引入类内近邻,提出一种二次近邻稀疏重构表示法。该方法首先在原始训练集上选择各类训练样本中与待测样本距离相近的若干样本组成近邻样本集,并协同表示,接着分别用各类近邻样本重构待测样本,再次选择与待测样本相近的若干重构样本协同表示,最终实现模式分类。在ORL和FERET数据库上的仿真实验表明,相比现有的一些CRC算法,该方法在一定程度上缩短了运行时间,并使识别更精确。
[Abstract]:Sparse representation (sparse representation classification,SRC) based on the entire dataset affects the efficiency of face recognition to a great extent. How to use the sparse representation of less samples to ensure the computational efficiency and the recognition rate is also improved, especially in the face of illumination, angle, attitude and other uncontrolled environment, is still a problem. Considering the advantage of (collaborative representation classification,CRC based on L 2 norm sparse solution, in order to further improve the overall classification performance of CRC, a quadratic nearest neighbor sparse reconstruction representation is proposed by introducing intra-class nearest neighbors. In this method, first of all, some samples from all kinds of training samples, which are close to the distance of the samples to be tested, are selected on the original training set to form the nearest neighbor sample set, and then each kind of nearest neighbor sample is used to reconstruct the sample to be tested. Then, several reconstructed samples which are close to the sample to be tested are selected again, and finally the pattern classification is realized. The simulation results on ORL and FERET database show that compared with some existing CRC algorithms, this method can shorten the running time and make the recognition more accurate.
【作者单位】：南通航运职业技术学院教育信息化管理中心;南通航运职业技术学院管理信息系;
【基金】：南通航运学院科技基金重点资助项目(HYKJ/2016A02)~~
【分类号】：TP391.41

【参考文献】