Research On Object-oriented Remote Sensing Image Road Flood Disaster Extraction Method

Road is the lifeline of the country.Once the road is damaged and damaged after the disaster,it will seriously affect the flow of people and material transportation.Therefore,it is very important for the reconstruction and emergency rescue to accurately obtain the road disaster information and determine the road damage.In the case of road damage in the disaster area,the traditional ground field investigation method is difficult to carry out,the speed of road disaster point investigation is slow,the investigation area is small,the emergency response is not timely,and timeconsuming.Remote sensing technology is applied to highway disaster survey and disaster point extraction,which can make rapid response to the detection and emergency of the disaster point,and accurately obtain the highway flood damage situation on a macro scale and in a large range.Therefore,using high-resolution remote sensing image to detect the road damage in disaster area is of great significance.The main purpose of this paper is to study the flood disaster caused by the rising water level.How to use the remote sensing method to extract the flood information effectively.With the continuous improvement of remote sensing image resolution,the traditional pixel classification method only uses spectral information for information extraction,but ignores other effective information,which can not meet the requirements of information extraction from high-resolution remote sensing image.In recent years,object-oriented classification method is a hot spot in remote sensing image classification research,especially high-resolution remote sensing image.However,at present,in the research of object-oriented remote sensing image classification,spectral information and attribute information feature difference are often used to analyze and classify.However,in the object-oriented remote sensing image classification method,the determination of multi-scale segmentation scale parameters and feature selection are often based on empirical method,and there are subjective factors.To solve the problem of determining the optimal scale parameters in the process of object-oriented classification,this paper uses the method of qualitative comparison and exclusion one by one to determine the optimal segmentation scale.For the image with only visible light band,in the absence of near-infrared band and other bands,the user-defined band feature BSCC combined with a variety of image features is used to interpret the remote sensing image with a variety of methods to solve the road flood damage information Extract with interest.The specific research contents are as follows:(1)Based on the high-resolution two remote sensing image with only three visible bands of RGB,preprocessing is carried out.It mainly includes geometric correction,image splicing and image enhancement.(2)Research on multiscale image segmentation based on ecognition software.Under the support of ecognition platform,the influence of shape factor,spectral factor,compactness and smoothness parameter setting on multiscale segmentation of remote sensing image is studied,and the effectiveness of multiscale segmentation based on research area is determined.The optimal scale parameter value is determined by referring to ESP model plug-in.(3)Using the object-oriented method,the image is classified and extracted.Based on the multi-scale segmentation algorithm,the spectral features,shape features,texture features,context features and custom features of image features are fully mined and analyzed,and the appropriate rule set is created by combining threshold analysis with support Support vector machine(SVM)is used to extract the pre disaster and post disaster image,and the seed growth algorithm is used to extract the post disaster water body,interpret the image,classify the image in the study area,and verify the accuracy and kappa coefficient.(4)Analyze the affected roads.The vector data of highway before disaster,highway after disaster and water body after disaster are derived.Combined with Arc GIS software,at the same time,the vector data of pre disaster road is mapped to the vector data of post disaster Road,and the buffer zone of pre disaster road is analyzed,and the result and the water body after disaster are overlapped to get the information of road being destroyed by flood.This method can effectively identify the terrain information of only visible light(RGB)band image,and extract the information of road flood damage effectively.