Recognition Of Earthquake-caused Buildings Damage Based On UAV Orthophoto

The damage caused by earthquakes are the most serious in the natural disasters.Most earthquake will cause multiple secondary disasters,heavy casualties and property losses.The damage caused by earthquakes include geological disaster and buildings collapsed.Therefore,the extraction and identification of seismic damage information plays an important role in post earthquake assessment.The extraction and identification of seismic damage through high-resolution remote sensing images is an important means of post earthquake assessment.High resolution remote sensing data are the most important source of data for earthquake damage assessment.It is widely used to evaluate the loss after the earthquake.Especially with the rapid development of UAV technology,high resolution remote sensing data acquisition is more convenient and fast.Compared with traditional remote sensing technology,UAV remote sensing technology has higher resolution,faster acquisition speed,lower cost,and can carry sensors on demand.This paper selects UAV data from PI Shan earthquake as research data and research on damage grade information of buildings.This paper selected the UAV Orthophoto image,the image range includes the whole area of PI Xi Na township,PI Shan county.As the Orthophoto image,we can only obtain the roof information of buildings.After the earthquake,the damage level of buildings is related to the degree of roof damage.The walls on the side of the buildings with serious roof collapse or falling are often accompanied by cracks and collapse.But there are a few buildings that are damaged only by side walls,and the roof of buildings is basically intact.Therefore,it is necessary to explore whether there is a big discrepancy between UAV orthophoto information and building site information.According to the GPS trajectory information,match the photos on the spot with the UAV images.A total of 30 buildings were matched.In addition,the paper selected typical buildings of township government buildings,health centers and teacher's office for comparison of earthquake damage grades.It is found that the Orthophoto image is basically consistent with the results of the field investigation when the earthquake damage is especially serious or basically intact,and the results will be different when the damage is medium.Therefore,this paper combines the basic integrity and the minor damage of the five types of earthquake damage into one class,and combines the medium damage and the serious damage into one class.The earthquake damage grade is classified into three categories to ensure the accuracy of the results.This paper selects the UAV data of the July 2015 PI Shan earthquake to study the magnitude of building damage.The multi-scale segmentation method is used to segment the UAV image.After many attempts,we determined the segmentation scale is 300.The images after classification mainly include four types of vegetation,roads,bare land and buildings.Due to the lack of near infrared band data in UAV images,the paper uses GLI instead of NDVI index to extract vegetation.The part with GLI value greater than 0.015 is divided into vegetation types.In the remaining categories of objects,we divided the objects with more than 500 pixel values and the ratio of length to width is greater than 4 into road.The rest of the ground is bare land and buildings.The nearest neighbor classification method in fuzzy classification is used for classification.The buildings can be correctly extracted to reach 90%.According to the magnitude of earthquake damage,the buildings are divided into three categories: basically intact,destroyed and completely collapsed.We selected 218 building samples to calculate 11 basic parameters including the entropy of gray level co-occurrence matrix.According to the collinearity test,the logistic regression model is excluded.In the end,we selected ridge regression model.Based on the principle of independent variables,we selected 6 characteristic parameters as independent variables.We validated the model through 218 buildings.The accuracy of the model is 81%,the kappa coefficient is 0.67.The results show that the damage grade of buildings can be divided into three categories by using object-oriented technology and ridge regression model technology.It can meet the requirements of post earthquake assessment.It shows that UAV technology has great application prospects in the post earthquake disaster acquisition.The low altitude remote sensing technology of UAV can provide reliable data sources for earthquake emergency and seismic damage extraction.By setting up an appropriate model,we can judge the earthquake damage grade of buildings and improve the efficiency of evaluation.Therefore,the use of UAV technology to obtain seismic damage data and identify seismic damage has a good research prospect.