利用融合纹理与形态特征进行地震倒塌房屋信息自动提取

¹中国科学院对地观测与数字地球科学中心,北京市中关村北一条9号100190;²中国科学院研究生院,北京市玉泉路甲19号,100049;³中国科学院遥感应用研究所,北京市大屯路甲20号100101

基金项目: 中国科学院知识创新工程重要方向资助项目(kzcx2-yw-313-3)；国家863计划资助项目(2007AA12Z157)；中国科学院知识创新工程青年人才领域前沿专项资助项目(O8S01100CX)

详细信息

Extracting Damaged Buildings Information Automatically Based on Textural and Morphological Features

¹Center for Earth Observation and Digital Earth,CAS,9 Beiyitiao,Zhongguancun,Beijing 100190,China;²Institute of Remote Sensing Applications,CAS,A20 Datun Road,Beijing 100101,China;³School of Graduate,Chinese Academy of Sciences,A19 Yuquan Road,Beijing 100049,China

Funds: 中国科学院知识创新工程重要方向资助项目(kzcx2-yw-313-3)；国家863计划资助项目(2007AA12Z157)；中国科学院知识创新工程青年人才领域前沿专项资助项目(O8S01100CX)

摘要: 提出了一种以震后单一时相高空间分辨率光学遥感影像为基础,融合纹理特征和形态特征的地震倒塌房屋自动提取方法,研究了不同尺度纹理特征和形态特征在倒塌房屋提取中的作用和表现。以5.12汶川地震作为研究实例,结果表明,本方法能够有效提取地震倒塌房屋。倒塌房屋产品精度和用户精度分别为86.65%和86.35%,Kappa系数为0.790 6。
- 高空间分辨率 /
- 纹理特征 /
- 形态特征 /
- 特征选择
Abstract: Extracting damaged buildings accurately and quickly through remote sensing has an important meaning to damage evaluation and relief after earthquake.A method,which based only on high spatial resolution optical remotely sensed image acquired after earthquake and fused textural features and morphological features,was proposed to extract damaged buildings automatically.Besides,the use and behaving of different scales textural features and morphological features in damaged buildings extraction were investigated.A study case,Wenchuan Earthquake,was studied.Experiment results showed that this method could extract damaged buildings effectively and reach a good result that the product accuracy and the user accuracy of damaged buildings,and Kappa coefficient were 86.65%,86.35% and 0.790 6 respectively.
- high spatial resolution /
- textural feature /
- morphological feature /
- feature selection

1.	李钦，游雄，李科，汤奋，王玮琦. 局部物体块匹配的图像匹配算法. 武汉大学学报(信息科学版). 2022(03): 419-427 . 百度学术
2.	张乾，杨玉成，邵定琴，岳诗琴. 广义局部二值模式在彩色图像检索中的仿真. 计算机仿真. 2020(06): 457-461 . 百度学术