面向对象的损毁建筑物提取
|
摘要
地震灾害已经成为一种频发的自然灾害之一,在震后的灾害评估中,建筑物的倒塌情况是很重要的一项指标。本文用GeoEye影像提取海地地震中的损毁建筑物。由于海地地震时发生倒塌的房子大多处于比较老旧的地区,房屋比较密集,而且房顶结构也比较复杂,我们采取了一种基于规则集的方法通过分类将倒塌建筑物逐步与其他各种地物区分开来。即在影像多尺度分割的基础上,结合纹理特征及几何特征,通过规则集的方法构成分类树提取损毁房屋。另外在进行植被剔除时本文根据影像的特点提出了一种新的植被指数geo-NDVI,最后根据目视解译的结果对分类的结果进行了对比。
The earthquake disaster has become one of the frequent natural disasters.The collapse of the building is important in the disaster assessment of the earthquake.This study is using GeoEye images in the Haiti earthquake to extract the damage buildings.When the earthquake occurred in Haiti,the collapsed houses mostly are comparatively old ones,and the houses are densely situated and the roof structure is dense and more complicated.We took a rule-based method in classification to distinguish the collapsed houses with other kinds of residential buildings gradually.We extract the damaged buildings with a rule of a classification tree on the base of image multi-scale segmentation,combining with the texture characteristics and geometry.The texture characteristics in the rule set are generated by the gray level co-occurrence matrix,including homogeneity,dissimilarity,entropy,energy,mean,etc.Geometric features include border area,length,width,number of pixels and etc.In addition,it is proposed a new vegetation index geo-NDVI in eliminating of the vegetation based on the characteristics of image.Finally,the results were compared with the ones of classification by visual interpretation.
The earthquake disaster has become one of the frequent natural disasters.The collapse of the building is important in the disaster assessment of the earthquake.This study is using GeoEye images in the Haiti earthquake to extract the damage buildings.When the earthquake occurred in Haiti,the collapsed houses mostly are comparatively old ones,and the houses are densely situated and the roof structure is dense and more complicated.We took a rule-based method in classification to distinguish the collapsed houses with other kinds of residential buildings gradually.We extract the damaged buildings with a rule of a classification tree on the base of image multi-scale segmentation,combining with the texture characteristics and geometry.The texture characteristics in the rule set are generated by the gray level co-occurrence matrix,including homogeneity,dissimilarity,entropy,energy,mean,etc.Geometric features include border area,length,width,number of pixels and etc.In addition,it is proposed a new vegetation index geo-NDVI in eliminating of the vegetation based on the characteristics of image.Finally,the results were compared with the ones of classification by visual interpretation.
引文
[1]Baatz M,Scha.pe A.Multi-resolution Segmentation:An Optimization Approach for High Quality Multiscale Image Seg-mentation.Angewandte Geogr[R].Information Sverarbeitung,ed.by Strobl J,Blaschke T,Wichmann,Heidel-berg,2000(Ⅶ):12-23.
[2]周春艳,王萍,张振勇,齐成涛,王莹.面向对象的高分辨率遥感影像信息提取技术[C]//SPIE.2006.
[3]黄慧萍.面向对象影像分析中的尺度问题研究[D].北京:中科院遥感应用研究所,2003.
[4]Jing Zhang,Gui-li Li,Seok-wun He.Texture-based image retrieval by edge detection matching GLCM[C]//The 10thIEEE International Conference on High Performance Computing and Communications.2008.
[5]Haralick,R.M.,Shanmugam K.,Dinstein I..Textural features for image classification[J].IEEE Transactions on Sys-tems,Man and Cybernetics,1973,3(6):610-620.
[6]Bradley A,Jackway P,Love llB.Classification in scale-space:Applications to texture analysis[C]//Proc.Of DICTA-95.
[7]Walker R F,Jackeay P T,Longstaff ID.Recent developments in the use of the co-occurrence matrix for texture recogni-tion[C]//Proc of 13Internation Conference on Digital Image Processing.1997.
[8]郭德军,宋哲存.基于灰度共生矩阵的纹理图像分析研究[D].东北林业大学机电学院,2005.
[9]贾永红.计算机图像处理与分析[M].武汉:武汉大学出版社,2001.
[10]杨淑莹,胡军,曹作良.基于图像纹理分析的目标物体识别方法[J].天津理工学院报,2001,17(4):31-35.
[11]冯建辉,杨玉静.基于灰度共生矩阵提取纹理特征图像的研究[J].北京测绘报,2007(3):19-22.
[12]Ohanian P P,Dubes R C.Performance evaluation for four class of texture features[J].Pattern Recognition,1992.
[13]刘明慧.基于图像分类的图像分割算法研究与实现[D].武汉:武汉理工大学,2005.
[14]C.S.Lu,P.C.Chung,C.F.Chen.Unsupervised texture segmentation via wavelet transform[J].Pattern Recognition,1997,30(5).
[15]eCognition Developer 8Reference Book[Z].
[2]周春艳,王萍,张振勇,齐成涛,王莹.面向对象的高分辨率遥感影像信息提取技术[C]//SPIE.2006.
[3]黄慧萍.面向对象影像分析中的尺度问题研究[D].北京:中科院遥感应用研究所,2003.
[4]Jing Zhang,Gui-li Li,Seok-wun He.Texture-based image retrieval by edge detection matching GLCM[C]//The 10thIEEE International Conference on High Performance Computing and Communications.2008.
[5]Haralick,R.M.,Shanmugam K.,Dinstein I..Textural features for image classification[J].IEEE Transactions on Sys-tems,Man and Cybernetics,1973,3(6):610-620.
[6]Bradley A,Jackway P,Love llB.Classification in scale-space:Applications to texture analysis[C]//Proc.Of DICTA-95.
[7]Walker R F,Jackeay P T,Longstaff ID.Recent developments in the use of the co-occurrence matrix for texture recogni-tion[C]//Proc of 13Internation Conference on Digital Image Processing.1997.
[8]郭德军,宋哲存.基于灰度共生矩阵的纹理图像分析研究[D].东北林业大学机电学院,2005.
[9]贾永红.计算机图像处理与分析[M].武汉:武汉大学出版社,2001.
[10]杨淑莹,胡军,曹作良.基于图像纹理分析的目标物体识别方法[J].天津理工学院报,2001,17(4):31-35.
[11]冯建辉,杨玉静.基于灰度共生矩阵提取纹理特征图像的研究[J].北京测绘报,2007(3):19-22.
[12]Ohanian P P,Dubes R C.Performance evaluation for four class of texture features[J].Pattern Recognition,1992.
[13]刘明慧.基于图像分类的图像分割算法研究与实现[D].武汉:武汉理工大学,2005.
[14]C.S.Lu,P.C.Chung,C.F.Chen.Unsupervised texture segmentation via wavelet transform[J].Pattern Recognition,1997,30(5).
[15]eCognition Developer 8Reference Book[Z].
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心