综合ULBP与Gabor纹理特征的SAR影像建筑区提取方法研究  

Research on Building Area Extraction Method of SAR Image Integrating ULBP and Gabor Texture Features

作　　者：邵乐 王晓青[1] 翟玮 丁香[1] Shao Le;Wang Xiaoqing;Zhai Wei;Ding Xiang(Institute of Earthquake Science,China Earthquake Administration,Beijing 100036,China;Earthquake Administration of Gansu Province,Lanzhou 730000,China)

机构地区：[1]中国地震局地震预测研究所,北京100036 [2]甘肃省地震局,兰州730000

出　　处：《震灾防御技术》2023年第1期186-193,共8页Technology for Earthquake Disaster Prevention

基　　金：高分辨对地观测重大专项(民用部分)科研项目(31-Y30F09-9001-20/22-12)。

摘　　要：建筑区提取对于地震灾情快速评估和地震灾害风险识别至关重要,合成孔径雷达(SAR)影像的纹理特征研究可在建筑区提取方面发挥重要作用。利用全极化SAR影像,提出综合使用ULBP纹理特征和Gabor纹理特征的建筑区提取方法。在SAR数据预处理的基础上,首先对基于Gabor滤波的纹理特征影像进行主成分分析,保留前2个最优主成分纹理特征影像;然后与ULBP纹理特征进行波段组合;最后利用支持向量机监督分类方法对组合后的影像进行分类,获得建筑区。研究结果表明,综合使用ULBP纹理特征和Gabor纹理特征可得到更高的建筑区提取精度,总体分类精度达90%,Kappa系数为0.78。Building area extraction is very important for the rapid assessment of earthquake disasters and the identification of earthquake disaster risks.Research on texture features of synthetic aperture radar(SAR)images can play an important role in building area extraction.Using fully polarimetric SAR images,a building area extraction method using ULBP texture features and Gabor texture features is proposed.On the basis of SAR data preprocessing,the texture feature image based on the Gabor filter is analyzed by principal component analysis,the first two optimal principal component texture feature images are retained.Then,band combination is performed with ULBP texture features.Finally,the support vector machine supervised classification method is used to classify the combined image to obtain the building area.The results show that the comprehensive use of the ULBP texture feature and Gabor texture features can obtain higher building area extraction accuracy,the overall classification accuracy is 90%,and the Kappa coefficient is 0.78.

关 键 词：SAR影像 ULBP纹理特征 GABOR 纹理特征 建筑区提取 

分 类 号：P237[天文地球—摄影测量与遥感] TN957.52[天文地球—测绘科学与技术]