Research On The Key Techniques Of Seismic Damage Detection Of High Resolution Remote Sensing Image

Earthquake is one of the serious natural disasters,so it is very important to accurately extract the damage information of buildings in earthquake disaster.High-resolution remote sensing image has many advantages,such as high spatial resolution,wide coverage and relatively easy data acquisition.It has become an important data source for post-earthquake emergency response and earthquake-damaged buildings detection.Currently,the main earthquake-damaged buildings detection methods can be divided into the classification method based on single temporal post-earthquake image and the change detection method based on multiple temporal pre-and post-earthquake images.The former method overcome the dependence on the post-earthquake information,but the post-earthquake scene has complex structure and spatial layout,as well as a wide variety of ground objects.Therefore,it is a key and difficult problem to identify the earthquake-damaged buildings only according to the information after earthquake.The latter method identifies the earthquake-damaged buildings by comparing the change information of pre-and post-earthquake buildings,but it usually ignores the "pseudo change" caused by the shadow and the multi-scale characteristics in the change information.To this end,this article focuses on two types of earthquake-damaged buildings detection methods,the specific content is as follows:(1)In the method of the classification method based on single temporal post-earthquake image,a new method for damaged building detection based on optimized visual dictionary from post-earthquake high-resolution remote sensing images is proposed.First,WJSEG and a set of non-building screening rules are applied to extract the potential building set.Secondly,a visual dictionary model of earthquake damage is constructed by introducing spectral,texture and geometric features,and a visual dictionary optimization strategy based on intraand inter-class penalty factors was designed.Finally,the buildings are further classified into intact buildings,partially damaged buildings and ruins by random forest classifier.In the experiments performed on the images of Wenchuan and Yushu areas,the overall accuracy of this method is above 85%,which is significantly better than the other two advanced comparison methods.(2)In the method of the change detection method based on multiple temporal pre-and post-earthquake images,an object-based change detection method for high-resolution remote sensing images combining shadow compensation and multi-scale fusion is proposed.First,the shadows in the remote sensing images are extracted.Then multi-scale change detection is conducted with shadow compensation.In the process,an objective function is constructed of mutual scale information minimization to realize the adaptive extraction of scale parameters.Finally,combined with the shadow compensation factor,a multi-scale decision-level fusion strategy built on D-S theory of evidence is designed,and the levels of change intensity are further divided.In experiments performed on images from Chongqing,Nanjing,and Shanghai,this method solves the problem of misdetection caused by shadows,and the overall accuracy is significantly better than other comparison methods,which lays a solid foundation for subsequent earthquake-damaged building identification based on change detection results.