北京航空航天大学学报 ›› 2021, Vol. 47 ›› Issue (3): 549-557.doi: 10.13700/j.bh.1001-5965.2020.0498

• 论文 • 上一篇    下一篇

基于记忆关联学习的小样本高光谱图像分类方法

王聪1,2, 张锦阳2, 张磊2, 魏巍1,2, 张艳宁2   

  1. 1. 西北工业大学 深圳研究院, 深圳 518057;
    2. 西北工业大学 计算机学院, 西安 710129
  • 收稿日期:2020-09-07 发布日期:2021-04-08
  • 通讯作者: 魏巍 E-mail:weiweinwpu@nwpu.edu.cn
  • 作者简介:王聪,女,博士研究生。主要研究方向:高光谱图像分类;张锦阳,男,硕士。主要研究方向:高光谱图像分类;张磊,男,博士,教授。主要研究方向:图像处理及机器学习;魏巍,男,博士,副教授,博士生导师。主要研究方向:计算机视觉、图像处理;张艳宁,女,博士,教授,博士生导师。主要研究方向:图像处理与计算机视觉、模式识别与人工智能、机器学习等。
  • 基金资助:
    深圳市科技创新委员会基金(JCYJ20190806160210899);国家自然科学基金(61671385,U19B2037)

Small sample hyperspectral image classification method based on memory association learning

WANG Cong1,2, ZHAGN Jinyang2, ZHANG Lei2, WEI Wei1,2, ZHANG Yanning2   

  1. 1. Research & Development Institute of Northwestern Polytechnical University in Shenzhen, Shenzhen 518057, China;
    2. School of Computer Science, Northwestern Polytechnical University, Xi'an 710129, China
  • Received:2020-09-07 Published:2021-04-08
  • Supported by:
    Foundation of Science, Technology and Innovation Commission of Shenzhen Manicipality (JCYJ20190806160210899); National Natural Science Foundation of China (61671385,U19B2037)

摘要: 高光谱图像(HSI)分类是遥感领域的基础应用之一。该任务旨在根据部分带类别标签的像素样本训练分类器,预测图像中剩余像素对应的类别标签。在实际应用中,由于人工标记样本成本过高,只能获得少量带标签的样本。针对少量样本无法准确描述数据分布从而导致训练过程过拟合的问题,提出一种基于记忆关联学习的小样本高光谱图像分类方法。考虑到无标签样本中包含大量与数据分布相关的信息,构建基于有标签样本记忆模块,并根据样本间的特征关联,利用不断更新的记忆模块学习无标签样本的潜在类别分布,构建无监督分类模型,并与传统的有监督分类模型进行联合学习。在多个高光谱图像分类数据集上的实验结果表明,所提方法能有效提升小样本高光谱图像分类的准确性。

关键词: 记忆关联学习, 半监督, 小样本, 高光谱图像(HSI), 分类

Abstract: Hyperspectral Image (HSI) classification is one of the fundamental applications in remote sensing domain. Due to the expensive cost of manual labeling in HSIs, in real applications, only small labeled samples can be obtained. However, limited samples cannot accurately describe the data distribution and often cause the training of classifiers to be overfitting. To address this problem, we present a small sample hyperspectral image classification method based on memory association learning. First, considering that the unlabeled samples also contain a lot of information related to the data distribution, we construct a memory module based on the labeled samples. Then, according to the feature association among labeled and unlabeled samples, we learn the label distribution of the unlabeled sample with the continuously updated memory module. Finally, we build an unsupervised classifier model and a supervised classifier model, and jointly learn these two models. Extensive experimental results on multiple hyperspectral image classification datasets demonstrate that the proposed method can effectively improve the accuracy of small sample HSI classification.

Key words: memory association learning, semi-supervised, small sample, Hyperspectral Image (HSI), classification

中图分类号: 


版权所有 © 《北京航空航天大学学报》编辑部
通讯地址:北京市海淀区学院路37号 北京航空航天大学学报编辑部 邮编:100191 E-mail:jbuaa@buaa.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发