Research On Damage Information Extraction For UAV Image Data

China is one of the countries with frequent natural disasters.In order to improve the efficiency of disaster relief,it is necessary to quickly obtain the information of disaster after disaster.UAV aerial remote sensing system can reflect the information of disaster in time.The traditional image analysis which is based on image classification approaches can not fully exploit the presence of the image spectrum,geometry,texture and context information classification accuracy which is low,slow and o ther limitations,and can not meet the rapidly extraction yield damage information needs.However,object-oriented High Resolution Remote sensing disaster information extraction provides a new way of thinking.In this paper,the object-oriented image analysis method is used to extract the water,road and building damage information of the UAV image.The main content and result of the research will be presented as follows:(1)The paper is based on the analysis of the existing main remote sensing image segmentation algorithm,and it focuses on the study of object-oriented image analysis,multi-scale segmentation algorithm and the flow,and in-depth analysis choice basis and the principle of the band weight,color,shape,compactness,smoothness,segmentation scale parameter.(2)This paper introduces the classification method that's based on the object-oriented method and based on pixel,and compares the advantages and disadvantages of the method.According to the experimental results that's based on the support vector machine method,the classification result has a large error,so it is not suitable for the extraction of the current disaster information.(3)It Obtained township slope damage information extracting thematic map based on the object-oriented and summarized properties characteristics of damaged river,damaged roads and damaged buildings.And the object-oriented classification technology is used to extraction damage information experiments of the Cao Po township UAV image.According to the characteristics of different ground objects,drainage,roads,and the structure of the optimal segmentation scale of image segmentation,it will build its multi-scale hierarchical network structure.The fuzzy rule knowledge base can be established with spectral features of the image objects,geometric feature,texture feature and spatial topology relationship information,and extracting the different objects in the optimal segmentation scale layered,obtaining results to make thematic maps.(4)Statistical analysis of the results of disaster information extraction.The kappa coefficient is used to evaluate the accuracy of damage information extraction results.Not only the use of the spectral information of ground objects,and considering the geometry,texture and context and other information for object classification method,which the classification accuracy is higher,and the overall accuracy of the experimental reaches 80.55%.The area of damaged river,length of damaged road and number of damaged buildings are analyzed,and it found that the main damaged areas are the buildings on both sides of the bank river and the road.