摘要
本文基于哨兵卫星Sentinel-1A双极化微波遥感数据,应用H/Alpha-Wishart、H/A/Alpha-Wishart两种非监督分类和Wishart迭代分类、支持向量机(SVM)两种监督分类方法分别对上海及长江入海口区域进行地物信息提取。对遥感数据进行预处理,利用H/A/Alpha极化目标分解提取极化特征参数;对影像进行非监督分类,对比非监督分类结果,计算分类精度;利用Wishart分类器和SVM分类器进行监督分类;参考Google earth等地图工具,对研究区的训练样本进行选取,将训练样本和极化参数带入分类器进行训练并不断调整参数,得到分类结果;进行精度分析,对比各分类方法的分类精度。由最终分类结果可知:监督分类对图像的解译效果最好,可解释性强,SVM监督分类结果最好,H/Alpha-Wishar最差,非监督分类方法中增加了A的H/A/Alpha-Wishart,非监督分类精度得到了提升。
Based on Sentinel satellite sentinel-A dual-polarization microwave remote sensing data,two unsupervised classifications H/Alpha-wishart and H/A/alpha-wishart,and two supervised classifications Wishart classification and Support Vector Machine(SVM)were used to extract ground object information in the mouth of the Changjiang River and Shanghai city.Firstly,remote sensing data were preprocessed,and polarization characteristic parameters were extracted by H/A/Alpha polarization target decomposition.The images were classified unsupervised and the results were compared.Wishart classifier and SVM classifier are used for supervised classification.With reference to Google earth and other map tools,the training samples in the research area are selected.The training samples and polarization parameters are put into the classifier for training,and the parameters and samples are adjusted to obtain the classification results.Finally,the accuracy analysis is carried out and the classification accuracy of each classification method is compared.According to the final classification results that supervised classification has the best interpretation effect on images,with strong interpretability,SVM supervised classification results were the best,while H/alpha-wishar was the worst.In the unsupervised classification method,H/A/alpha-wishart’s unsupervised classification accuracy was improved by adding A polarization parameter.
作者
赵美玲
侯成磊
ZHAO Meiling;HOU Chenglei(School of Earth Sciences and Resources,Chang’an University,Xi'an 710054,China)
关键词
极化目标分解
非监督分类
信息提取
polarization target decomposition
unsupervised classification
information extraction
