智能科学与技术学报 ›› 2021, Vol. 3 ›› Issue (3): 342-350.doi: 10.11959/j.issn.2096-6652.202135

• 专刊:目标智能检测与识别 • 上一篇    下一篇

鲁棒的非负监督低秩鉴别嵌入算法

姚裕1, 万鸣华1,2   

  1. 1 南京审计大学信息工程学院,江苏 南京 211815
    2 江苏省社会安全图像与视频理解重点实验室,江苏 南京 210094
  • 修回日期:2021-08-05 出版日期:2021-09-15 发布日期:2021-09-01
  • 作者简介:姚裕(1997− ),男,南京审计大学信息工程学院硕士生,主要研究方向为模式识别与特征提取
    万鸣华(1978− ),男,博士,南京审计大学信息工程学院教授,主要研究方向为模式识别与机器学习
  • 基金资助:
    国家自然科学基金资助项目(61876213);江苏省自然科学基金资助项目(BK20201397);2018年江苏省高校自然科学研究重大项目(18KJA520005);江苏省社会安全图像与视频理解重点实验室项目(J2021-4);2020年江苏省科研与实践创新计划资助项目(SJCX20_0670)

Robust non-negative supervised low-rank discriminant embedding algorithm

Yu YAO1, Minghua WAN1,2   

  1. 1 School of Information Engineering, Nanjing Audit University, Nanjing 211815, China
    2 Key Laboratory of Image and Video Understanding for Social Safety, Nanjing 210094, China
  • Revised:2021-08-05 Online:2021-09-15 Published:2021-09-01
  • Supported by:
    The National Natural Science Foundation of China(61876213);The Natural Science Fund of Jiangsu Province(BK20201397);The Key Research and Development Program Science Foundation in Colleges and Universities of Jiangshu Province(18KJA520005);Jiangsu Key Laboratory of Image and Video Understanding for Social Safety of Nanjing University of Science and Technology(J2021-4);Postgraduate Research and Practice Innovation Program of Jiangsu Province(SJCX20_0670)

摘要:

非负矩阵分解(NMF)已经得到了广泛应用。但NMF更注重数据的局部信息,忽略了数据的全局信息,而在有噪声图像的分类问题上,数据的全局信息往往比局部信息更具鲁棒性。为了提高算法的鲁棒性,结合数据的局部与全局信息,并且考虑低秩表示的特性,提出了一种新的非负监督低秩鉴别嵌入算法,此算法假设数据存在噪声,将数据分解为干净数据与噪声数据,并通过L1范数对噪声矩阵进行稀疏约束,增强对噪声的鲁棒性。此外,该算法使用低秩表示学习到一个低秩矩阵,然后通过非负分解再一次增强算法的鲁棒性。最后结合图嵌入理论,同时保留了数据的局部性和全局性。将该算法应用于各种加噪数据库中发现,相对于对比方法,该算法识别率提升了5%~15%。

关键词: 非负矩阵分解, 低秩表示, 特征提取, 图嵌入

Abstract:

Non-negative matrix factorization (NMF) has been widely used.However, NMF pays more attention to the local information of the data, it ignores the global representation of the data.In terms of image classification, the global information of the data is often more robust to noise than the local information.In order to improve the robustness of the algorithm, combined with the data of local and global representation, and considered the characteristics of low-rank representation, a non-negative supervised low-rank discriminant embedded algorithm was proposed.This algorithm assumed the existence of noise in the data, decomposed the data into clean data and noise data, and made sparse constraints on the noise matrix through the L1norm, so as to enhance the robustness to noise.In addition, the algorithm used low-rank representation to learn a low-rank matrix, then through non-negative decomposition, the robustness of the algorithm was enhanced again.Finally, combined with a study of graph embedding method, the local and global data were retained at the same time.The algorithm is applied to various noisy databases, and the recognition rate of this algorithm is improved by about 5%~15% compared with the comparison algorithm.

Key words: non-negative matrix factorization, low-rank representation, feature extraction, graph embedding

中图分类号: 

No Suggested Reading articles found!