Research On Key Technologies Of Road Disaster Monitoring Based On High-resolution Remote Sensing Imagery

Roads are vulnerable to damage in natural disasters,especially,China is a country with frequent natural disasters.Roads play a very important role as the infrastructure of people's production and the lifeline of disaster relief.At present,the road disaster information acquisition is mainly accomplished by investigators,but this mode needs a long period,and the safety of post-disaster investigators can not be guaranteed.With the rapid development of remote sensing technology and national policy support for integrated military and civilian development,the application of high-resolution remote sensing images in practice is becoming more popular.Compared with the middle and low resolution remote sensing images,the highresolution remote sensing has the characteristics of wide coverage and time efficient.It can obtain very comprehend information of ground objects without the need for people to visit the locale.Therefore,high-resolution remote sensing technology can be used as an important means of road disaster information acquisition and post-disaster assessment.How to effectively use high-resolution remote sensing images for road disaster monitoring has important research significance and application value.Firstly,this dissertation studies the storage technology of massive high-resolution remote sensing images,and on this basis,studies the key algorithms of road disaster monitoring,namely change detection.It uses multi-temporal high-resolution remote sensing images to monitor roads and surrounding facilities,and achieves the purpose of road disaster early warning and evaluation.In this process,change detection algorithm is the key technology of road disaster information acquisition.Current remote sensing image change detection algorithms can be divided into pixel-based method and object-oriented method.When the pixelbased change detection method is applied to change detection of high-resolution remote sensing images,the preprocessing errors of multi-temporal remote sensing images and other noise interference lead to the non-correspondence of temporal pixels,and then the accuracy of image change detection.Object-oriented change detection algorithms firstly divide multi-temporal remote sensing images into objects or patches,and then uses objects as the main unit for change detection.However,this kind of methods need to consider the spectral inconsistency between multi-temporal remote sensing images.Based on Multi-temporal high-resolution remote sensing imagery,this dissertation studies the key technology of road disaster monitoring and develops an program of road disaster monitoring.The work of this dissertation includes the following parts:(1)This dissertation investigates the storage technology of massive high-resolution remote sensing images,distributed caching method is introduced to improve the storage performance of remote sensing data,this can reduces the running time of remote sensing data processing process.The technology foundation of data storage and fast retrieval is prepared for the application of massive high resolution remote sensing image in road disaster monitoring.(2)The application of ICA in change detection is studied,and the damped Newton method is introduced into the fast ICA algorithm to solve the problem that the algorithm may not converge when the initial point is far from the optimal solution.This method adds additional searches in the direction of Newton iteration,which ensures the convergence of the algorithm.On the basis of ICA change image,the final change result is obtained by threshold segmentation.The result comparison shows that the improved method can detect change information more accurately.(3)Aiming at the problem that high-resolution remote sensing image change detection is sensitive to pre-processing errors such as registration and other noise,adaptive space is introduced,and high-resolution remote sensing image change detection technology combined with adaptive space and conditional random field is studied,which can reduce the impact of registration errors between multi-temporal and high-resolution remote sensing images on change detection to a certain extent.(4)In order to solve the problem that the change detection accuracy is not high due to the inconsistency of spectral information between multi-temporal images,the technology of superpixel segmentation for high-resolution remote sensing images and change detection based on conditional random field model is studied.Experiments show that this method can well adapt to the inconsistency of spectral information between multi-temporal images.For remote sensing images of the same region in different seasons,it can also detect changes well.At the same time,this method can also better protect the boundary.(5)The detection and assessment model of road disasters based on high-resolution remote sensing image is studied followed with the change detection algorithm.Road disasters identification is guided by basic geography data,and road damage grade is classified according to the degree of severity.On this basis,a road disaster monitoring system is developed,which can automatically detects the changes of roads and affiliated areas according to remote sensing images and generates reports and thematic maps.