联合光谱角与组合特征参数的高光谱影像分类  

Hyperspectral Image Classification with Combination of Spectral Angle and Combined Characteristic Parameters

作　　者：占燕婷 吴柯[1] 徐宏根[2] 刘慧泽 ZHAN Yanting;WU Ke;XU Honggen;LIU Huize(Institute of Geophysics and Geomatics,China University of Geoscience(Wuhan),Wuhan 430074,China;Wuhan Geological Survey Center of China Geological Survey,Wuhan 430205,China)

机构地区：[1]中国地质大学(武汉)地球物理与空间信息学院,武汉430074 [2]中国地质调查局武汉地质调查中心,武汉430205

出　　处：《遥感信息》2021年第6期140-146,共7页Remote Sensing Information

基　　金：“全球变化与海气相互作用(二期)”专项(GASI-01-DLYG-WIND01);国防科工局民用“十三五”航天预先研究项目(D040104);国家自然科学基金项目(62071438、61701452)。

摘　　要：光谱角匹配分类以光谱整体相似度作为分类准则,却无法充分考虑光谱的局部细节特征,导致高光谱遥感影像的分类结果存在着较大的误差。针对此问题,提出一种联合光谱角与组合特征参数(spectral angle mapping-combination characteristic parameter,SAM-CCP)的新型高光谱影像分类方法。该方法在光谱角距离的基础上,引入光谱特征参数,有效突出光谱信息的局部特征,从而提高分类精度。首先,将地物反射光谱的整体特征和典型的吸收谷特征相结合,计算参数向量的欧式距离,并调节其开方系数;然后,自动选择最佳光谱特征参数组合,构建稳健的匹配模型,对影像进行逐像元分类。分别对Indian Pines地区和Cuprite矿区的高光谱遥感影像进行了分类实验。结果表明,相比传统的匹配方法,SAM-CCP方法能够有效地改善分类精度,具有较好的适用性。Spectral angle mapping uses the overall similarity of the spectrum as the classification criterion,but cannot fully consider the local details of the spectrum,which leads to large errors in the classification results of hyperspectral remote sensing images.To solve this problem,this paper proposes a new type of hyperspectral image classification method combining spectral angle and combined characteristic parameters(SAM-CCP).Based on spectral angular distance,this method introduces spectral characteristic parameters to effectively highlight the local characteristics of spectral information,thus improving the classification accuracy.First,combine the overall characteristics of surface reflection spectrum with the typical absorption valley characteristics,calculate the Euclidean distance of the parameter vector,and adjust the square root coefficient.Then,automatically select the best combination of spectral feature parameters to construct a robust matching model.The images are classified pixel by pixel.We applied this method to perform classification experiments on hyperspectral remote sensing images of Indian Pines and Cuprite.The results show that:compared with the traditional matching method,the SAM-CCP method can effectively improve the classification accuracy and has better applicability.

关 键 词：SAM-CCP 高光谱影像分类 光谱相似性测度 光谱局部特征 特征参数组合 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]