基于不同特征的随机森林极化SAR图像分类  

Tidal Flat Classification Based on Random Forest Model Using Different Features of Polarimetric SAR

作　　者：陈媛媛 郑加柱 魏浩翰[1] 张荣春 欧翔 CHEN Yuan-Yuan;ZHENG Jia-Zhu;WEI Hao-Han;ZHANG Rong-Chun;OU Xiang(College of Civil Engineering, Nanjing Forestry University, Nanjing 210037, China;School of Geographic and Biologic Information, Nanjing University of Posts and Telecommunications, Nanjing 210023, China)

机构地区：[1]南京林业大学土木工程学院,南京210037 [2]南京邮电大学地理与生物信息学院,南京210023

出　　处：《计算机系统应用》2019年第8期183-189,共7页Computer Systems & Applications

基　　金：江苏省测绘地理信息科研项目(JSCHKY201708);江苏省自然科学基金(BK20180779);南京林业大学青年科技创新基金(CX2018015)~~

摘　　要：近些年,利用计算机对极化SAR图像进行分类逐渐成为遥感领域的一个研究热点.本文采用全极化SAR数据,利用不同的特征提取算法提取特征,并基于随机森林模型最终实现对江苏沿海滩涂的分类.首先采用H/α和Freeman两种分解算法提取极化特征参数,采用灰度共生矩阵提取纹理特征参数;然后将提取的所有特征进行不同的组合,构成不同的特征集;最后采用随机森林模型对不同特征集合进行分类和精度评估.结果表明仅用纹理特征对沿海滩涂进行分类时效果较差;利用极化分解提取出的散射特征进行分类的结果要优于矩阵元素特征的分类结果;综合了极化散射特征和纹理特征的组合方式在沿海滩涂的分类中可以取得最优的分类结果,总体精度和Kappa系数可以达到94.44%和0.9305,表明极化SAR图像中蕴含的不同方面的特征在分类中具有一定的互补性.The classification of polarimetric SAR images by computer has become a research hotspot in remote sensing. In this study, the fully polarimetric SAR data is used to extract characteristics by different algorithms, and the classification of tidal flat of Jiangsu coastal is realized. Firstly, the polarimetric scattering characteristics are extracted by H/α and Freeman decompositions, and the texture features are extracted by gray level co-occurrence matrix. Then, all the extracted features are combined to form different feature sets. Finally, the random forest model is used to classify and accurately evaluate with different feature sets. The study shows that using only texture features to classification achieves a poor performance. The classifications using the scattering features extracted by polarimetric decompositions are better than that of matrix element features. The combination of polarimetric scattering and texture characteristics can obtain best classification in coastal tidal flat, and the overall accuracy and Kappa coefficient are 94.44% and 0.9305, respectively. It indicates that the characteristics of different aspects contained in fully polarimetric SAR image have certain complementarity in the classification of coastal area.

关 键 词：极化SAR 极化分解 特征提取 随机森林 分类 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]