Research On Object Oriented Classification Technology And Application In Information Extraction Of Post Earthquake

At present,the high-resolution remote sensing images have been widely applied in various fields,with the increase of application demands and needs of the user,the image classification and information extraction methods now are constantly in-depth studying.Compared with the low resolution remote sensing images,high resolution remote sensing image feature contains more abundant information,the traditional pixel based classification method in some aspects cannot fully mining and analysis based on image multi feature information,which reduced the accuracy of score class.Therefore,this paper focused on the research of object oriented image classification technology and applied it to the extraction of damage information in the post earthquake building.Based on the "accurate,efficient and rapid analysis of disaster damage to housing"as the goal,according to the current research results based on object-oriented classification technology,based on the existing methods,this paper made a set of high resolution remote sensing image technology,object oriented information extraction process of buildings afterdisaster,through different levels of coordination and improvement the key technology of realization of the disaster buildings effective extraction and quantitative classification of disaster degree.In the experiment,a high-resolution remote sensing image with some characteristic information was selected as the experimental data.After the human-machine interaction interpretation of the post disaster image information,the disaster information in the study area was selected selectively.The experiment takes Ecognition and ArcGIS and ENVI software as image processing platform,combines MATLAB,Excel and other software as data analysis platform,and carries out object-oriented classification of experimental data.(1)Thispaper,taking the earthquake as a typical earthquake disaster,using high resolution remote sensing image,the principle of image object segmentation produced by internal homogeneity and heterogeneity of image object comprehensive evaluation based on the research of image segmentation scale,exploring the optimal scale in the process of segmentation.(2)Based on the improved SEaTH algorithm,the feature library is optimized andthe classification threshold of the selected feature is calculated.(3)Object oriented classification method is applied to classify the damaged buildings in the study area by using multiple features combined with preferred features and thresholds.(4)On the basis of this research,this paper compared the results extracted by the method with those obtained by the traditional classification method,and evaluated the accuracy with the same accuracy,so as to illustrate the practicability of the method and technology adopted in this paper.Through the study on object oriented classification technology research,it was found that the remote sensing image classification of object oriented method in extracting object information,extracting the target object is not only a certain degree of distortion and improved the classification accuracy,the final classification result can effectively avoid the "Pepper phenomenon",the classification results and the real world can be consistent well in shape and attributes.In the process of information extraction after disaster,this method can extract disaster information quickly and effectively,providing corresponding disaster relief information for disaster relief area in real time,and providing geographic services for disaster relief and disaster relief.