Study On Rapid And Intelligent Extraction Of Typical Seismic Secondary Geological Disasters Using High Resolution Remote Sensing Imagery

Seismic secondary geological disasters have characteristics of short formation time,strong burstiness,great destructiveness and strong induction,which intensify the earthquake effect and seriously hinder the post-disaster emergency rescue.It is of great significance to acquire seismic secondary geological disasters(SSGDs)information quickly.With the advantages of quickness,extensiveness and economy,remote sensing technology has become one of the most important methods of earth observation from space.Traditionally,application of remote sensing to seismic secondary disasters investigation is based on visual interpretation of remote sensing data and geographic data.With the rapid development of remote sensing technology,space technology and computer technology,rapid identifying and detecting seismic secondary disasters are developing to automatic and intelligent gradually.Interaction methods combined with remote sensing technology such as fine classification,change detection and feature analysis have become main means of earthquake emergency and are hot spots both domestic and overseas in recent years,using high resolution remote sensing imagery after earthquake.The spectral,shape,texture,and spatial distribution of the seismic secondary geological disasters in high resolution remote sensing imagery are the foundation for extraction of remote sensing information.These features are important to use the characteristics of the secondary geological disasters,select reasonable feature parameters and acquire the disaster information quickly and intelligently.In addition,it is one of the hot spots in the field of computer vision to detect the saliency regions and extract the specific targets based on the theory of visual saliency.The SSGDs can be effectively identified and detected in remote sensing imagery based on the analysis and application of the visual attention models.Therefore,hierarchical analysis methods based on multi-feature and visual saliency detection methods are applied to acquire seismic secondary geological disasters rapidly and intelligently in this paper.The main achievements of this research are as follows:(1)A new hierarchical analysis method for automatic extraction of seismic secondary disasters was proposed based on several characteristics,including spectral,shape,texture and spatial distribution of the disasters in high resolution remote sensing imagery.Using this method,the SSGDs can be acquired quickly and intelligently by selecting reasonable feature parameters,computing thresholds automatically and using expert knowledges to establish effective extraction rules.This fast and intelligent method is of great significance for earthquake disaster emergency rescue.(2)Two rapid extraction methods for seismic secondary disasters were proposed based on saliency analysis and superpixel segmentation.Both methods were based on the saliency detection,and the sizes of the targets were super pixels instead of single pixels.In both methods,statistical learning and similarity measurement were applied to extract saliency targets successfully.Some landslides and debris flows images were tested in experiments,which were aimed to solve the problem of insufficient detection ability in the complex and multi-targets remote sensing imagry.The experiment results showed that the proposed algorithms more extract saliency regions for significant goals with obvious and consistent type or little difference in remote sensing imagry efficiently.Usually,water areas have lower saliency than other objects,so these methods are not suitable for extraction of barrier lakes after earthquake.(3)A new fast and intelligent extraction model for seismic secondary disaster information was proposed by summarizing the visual saliency method and the multifeature analysis method.The main innovation of this paper are as below:1)A new hierarchical multi-feature analysis method for extraction of seismic secondary disasters was proposed;2)The feature thresholds can be computed automatically based on the improved Otsu method;3)Two rapid extraction methods were proposed based on saliency analysis and superpixel segmentation,and successfully applied to the extraction of seismic secondary disasters in high resolution remote sensing imagery;4)Some general methods and key technologies of remote sensing information extraction can be summarized.