基于Gabor纹理的高光谱影像空谱特征分类  

Spectral-spatial feature classification of hyperspectral image based on Gabor textures

作　　者：杨帆 张雪松 杨其淼 谭熊[1] 秦进春 YANG Fan;ZHANG Xuesong;YANG Qimiao;TAN Xiong;QIN Jinchun(Information Engineering University,Zhengzhou 450001,China;Unit 32035,Xi'an 710600,China;Unit 32023,Dalian 116000,China;Xi'an Research Institute of Surveying and Mapping,Xi'an 710054,China;State Key Laboratory of Geo-Information Engineering,Xi'an 710054,China)

机构地区：[1]信息工程大学,河南郑州450001 [2]32035部队,陕西西安710600 [3]32023部队,辽宁大连116000 [4]西安测绘研究所,陕西西安710054

出　　处：《测绘科学与工程》2019年第4期36-40,共5页Geomatics Science and Engineering

摘　　要：在高光谱影像分类中,为增加传统光谱分类精度,现有的多数方法在光谱特征的基础上加入了基于空间信息的形态学剖面特征,虽取得一定的精度提升,但较少顾及高分辨率影像中非常有用的纹理结构信息,对高分辨率高光谱影像蕴含的丰富信息利用不够,特征提取的广度有待提高。对此,本文在分类中加入了Gabor滤波器提取的影像纹理信息,进一步提高了高光谱影像的分类精度。首先,从高光谱影像的前三个主分量中提取Gabor纹理特征;然后,与光谱特征进行融合,合并为高维特征;最后,在支持向量机(Support Vector Machine,SVM)中进行分类。实验结果显示与传统光谱分类方法和空间形态学剖面方法相比,Gabor纹理特征在表达空间信息和提高分类精度方面作用明显。In hyperspectral image classification.currently most methods will add morphological profile features based on spatial information to spectral features.Although the classification accuracy has been improved to some extent,those methods seldom utilize the useful texture structure information in hyperspectral image,leading to the insufficient use of abundant information contained in high accuracy hyperspectral images,and the breadth of feature extraction needs to be improved.In order to solve these problems,the image texture information obtained by Gabor filter is added to the classification.Gabor texture features are extracted from tlie first three principal coinpneiits of hyperspeclral images,ilien they are fused with spectral features to merge into high-diniensional features,and finally classified using the support vector machine(SVM)algorithm.The experimental results show that the proposed method with Gabor texture features is more effective in expressing spatial iiifonnation anti improving classification accuracy compared with traditional spectral classification methods.

关 键 词：高光谱影像分类 主成分分析 GABOR纹理特征 特征融合 支持向量机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]