结合MCLU准则和DE算法的高光谱数据分类方法  

A hyperspectral data classification method based on the multi-class-level uncertainty criterion and differential evolution algorithm

作　　者：王立国 池辛格[2] WANG Liguo;CHI Xinge(College of Information and Communication Engineering,Dalian Minzu University,Dalian 116600,China;College of Information and Communication Engineering,Harbin Engineering University,Harbin 150001,China)

机构地区：[1]大连民族大学信息与通信工程学院,辽宁大连116600 [2]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001

出　　处：《哈尔滨工程大学学报》2021年第11期1688-1693,共6页Journal of Harbin Engineering University

基　　金：国家自然科学基金项目(62071084)。

摘　　要：针对现有高光谱数据半监督分类方法在精度和运算效率方面的不足,本文提出了一种结合多层次不确定性采样策略和自适应差分进化算法的半监督分类方法。通过多层次不确定性准则从无标签样本中选取信息量较高的样本和可以确定类别的样本,用后者和已知的有标签样本共同对前者进行标记;通过自适应差分进化算法寻优,扩充训练样本;使用新的训练样本集训练支持向量机分类器,对测试样本进行分类。采用2组真实高光谱遥感数据进行实验。结果表明:该方法能够充分利用样本信息,有效提高了小样本情况下高光谱数据的分类精度。Addressing the shortcomings in accuracy and efficiency of existing semi-supervised classification methods for hyperspectral data, a semi-supervised classification method combining multi-class-level uncertainty(MCLU) sampling strategy and a self-adaptive differential evolution(DE) algorithm is proposed in this paper. First, samples with more information and samples that can be classified are selected from the unlabeled samples through the MCLU criterion, with the latter sample and the known labeled sample used to mark the former. Next, the self-adaptive DE algorithm is used to execute the optimization of the proposed method and expand the training samples. Finally, the new training sample set is used to train the support vector machine classifier to classify the test samples. Two groups of real hyperspectral remote sensing data are used to conduct the experiment, and the results show that this method can make full use of sample information and improve the classification accuracy of hyperspectral data in the case of small samples.

关 键 词：遥感 高光谱 半监督分类 多层次不确定性准则 自适应差分进化 机器学习 支持向量机 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置] TP751[自动化与计算机技术—控制科学与工程]