电子学报 ›› 2018, Vol. 46 ›› Issue (1): 175-184.DOI: 10.3969/j.issn.0372-2112.2018.01.024

• 学术论文 • 上一篇    下一篇

联合核稀疏多元逻辑回归和TV-L1错误剔除的高光谱图像分类算法

徐金环1, 沈煜1, 刘鹏飞2, 肖亮1   

  1. 1. 南京理工大学计算机科学与工程学院, 江苏南京 210094;
    2. 南京邮电大学计算机学院、软件学院, 江苏南京 210023
  • 收稿日期:2016-05-27 修回日期:2017-01-03 出版日期:2018-01-25
    • 通讯作者:
    • 肖亮
    • 作者简介:
    • 徐金环,女,1992年9月出生,山东济南人.2014年毕业于南京理工大学计算机科学与技术系,现为博士研究生,从事高光谱遥感图像分类方面的研究.E-mail:jinhuan_2014@163.com;肖亮(通讯作者),男,1976年出生于湖南长沙,南京理工大学计算机科学与工程学院教授、博士生导师,江苏省"光谱成像与智能感知"重点实验室和教育部"高维信息智能感知与系统"重点实验室副主任.主要研究领域为:图像处理与计算机视觉、机器学习与模式识别.E-mail:xiaoliang@mail.njust.edu.cn
    • 基金资助:
    • 国家自然科学基金 (No.61571230); 国家重点研发计划 (No.2016YFF0103604); 江苏省自然科学基金 (No.BK20161500); 江苏省333工程 (No.BRA2015345)

Hyperspectral Image Classification Combining Kernel Sparse Multinomial Logistic Regression and TV-L1 Error Rejection

XU Jin-huan1, SHEN Yu1, LIU Peng-fei2, XIAO Liang1   

  1. 1. Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing, Jiangsu 210094, China;
    2. School of Computer Science, School of Software, Nanjing University of Posts and Telecommunications, Nanjing, Jiangsu 210023, China
  • Received:2016-05-27 Revised:2017-01-03 Online:2018-01-25 Published:2018-01-25
    • Corresponding author:
    • XIAO Liang
    • Supported by:
    • National Natural Science Foundation of China (No.61571230); National Key Research and Development Program of China (No.2016YFF0103604); Natural Science Foundation of Jiangsu Province,  China (No.BK20161500); 333 High-level Talents Cultivation Project in Jiangsu Province (No.BRA2015345)

摘要: 稀疏多元逻辑回归(SMLR)是高光谱监督分类中的重要方法,然而仅仅利用光谱信息的SMLR忽略了影像本身的空间特征,在少量监督样本下的分类精度和算法的鲁棒性仍明显不足;虽然通过引入核技巧,核稀疏多元逻辑回归(KSMLR)可以部分克服上述缺点,其分类错误仍然有待进一步降低.本文基于核稀疏多元逻辑回归分类误差的统计建模分析,提出一种联合核稀疏多元逻辑回归和正则化错误剔除的高光谱图像分类模型.提出的模型通过引入隐概率场,采取L1范数度量KSMLR分类误差的重尾特性建立数据保真项;利用全变差(Total Variation,TV)正则化度量隐概率场的局部空间光滑性.由Indian Pines和University of Pavia数据集等实测数据应用表明,该方法可以得到更鲁棒和更高的分类精度.

关键词: 高光谱, 图像分类, 核稀疏多元逻辑回归, 错误剔除

Abstract: Sparse multinomial logistic regression (SMLR) is an important supervised classification method for hyperspectral images (HSI). However, because the traditional SMLR based pixel-wise classifiers only use the spectral signatures, the good robustness and high classification accuracy are hardly achieved with a small number of samples without considering the spatial information of HSI. By using the kernel tricks, the kernel sparse multinomial logistic regression (KSMLR) method can partly overcome this limitation, however the resulted misclassification errors are still expected to be further reduced. According to the statistical analysis of classification errors resulted in KSMLR, we propose a novel two stage framework which combines KSMLR and error rejection for HSI classification. The proposed model, named KSMRL-TVL1, adopts the L1 norm to measure the heavy-tailed property of the classification errors so as to build the data fidelity term, and uses the total variation (TV) regularization term to measure the local spatial smoothness of the hidden probability field. The experiments on Indian Pines dataset and University of Pavia dataset show that the proposed method can better improve the robustness and classification accuracy.

Key words: hyperspectral image, image classification, kernel sparse multinomial logistic regression, error rejection

中图分类号: