Research On Automatic Identification Methods Of Earthquake Damaged Buildings Based On High-resolution Images

The earthquake is one of the most important natural disasters facing human beings,and it is a serious threat to the lives and property of the people.China is one of the countries with the most serious earthquake disasters in the world,and more than 60% of the cities have been threatened by moderate-strong earthquakes.Buildings are an important hazard in earthquake disasters.More than 75% of casualties in earthquakes are related to the collapse of houses.The collapse of large buildings can lead to a rapid increase in the number of casualties.After the earthquake,the damage degree and spatial location information of the building is the key information for emergency rescue.At the same time,degree of disaster in buildings can be used to estimate the structural damage of urban buildings,providing a basis for further estimating casualties.With the rapid development of remote sensing technology,remote sensing,with its characteristics of convenience,rapidity,macroscopic and dynamic,can gain time for disaster relief and disaster reduction,and plays a very important role in earthquake emergency and disaster relief work.The most critical step in remote sensing seismic damage assessment is the automatic identification and extraction of remote sensing seismic damage,and its accuracy directly affects the results of the entire assessment.Therefore,improving the accuracy of automatic identification of remote sensing earthquake damage is the key to the application of remote sensing technology in earthquake emergency work.Based on the analysis of the current research status,and the single-temporal remote sensing image classification,improved the morphological building index extraction method for damaged buildings proposed by Huang xin,this paper proposes a morphological section based on multiple structural elements and a seismic damage building method based on Markov random field iterative condition model,which has a higher accuracy in information extraction results.(1)Firstly the paper analyze and organize related the knowledge of object-oriented image segmentation and classification.The concept,steps,segmentation parameters and multi-scale segmentation network structure of multi-scale segmentation are introduced.The common regional merging algorithm based on the principle of minimum heterogeneity is introduced in detail.The selection of optimal segmentation scale and the principles to be followed are expounded.The quantitative description method of image features is introduced,and the concept and theory of object-oriented fuzzy classification method are analyzed and summarized.Secondly the paper introducing the building index method,a method for extracting earthquake damage buildings based on multiple structural element morphological profiles is proposed.Construct a suitable earthquake damage building index that can be directly applied to the inspection of intact and collapsed buildings without any training or segmentation process.The method is suitable for extracting damaged building information of different sizes and materials in earthquake,and can be used for large-scale processing of high-resolution satellite data,and has the advantages of high speed,high precision,less parameter adjustment and high degree of automation.The experimental results show that the accuracy of the method can be kept above 90%,far exceeding the experimental precision and speed of the object-oriented method.(2)Lastly a method for identifying earthquake-damaged buildings based on Markov random field iterative conditional model is proposed.The method divides the damage degree of the study area into three different levels,extracting intact buildings,partially collapsed buildings,and completely collapsed buildings with high speed,high precision,and high degree of automation.The accuracy of the method can be maintained above 90%.