Research On Traffic Disaster Information Extraction Based On Single HR Spaceborne SAR Image

Remote sensing plays an increasingly important role in the post-disaster emergency rescue.However,the rainy weather accompanied by the disaster greatly obstructs the optical remote sensing’s work.Due to the high-resolution,spaceborne Synthetic Aperture Radar(SAR)images have the characteristics of penetrating clouds;it can continuously monitor the disaster area.With the improvement of SAR image resolution,it is possible to monitor the important traffic-related disasteraffected objects such as bridges,airports,landslides,dammed lakes and collapsed buildings,and the monitoring results can be used for traffic emergency decision.In recent years,high-resolution,spaceborne SAR image-based information extraction is a hot topic,but mostly based on interferometric SAR image.But from the practical point of view,the acquisition of interferometric image is full of difficulties.It is of significant influence on improving traffic-related post-disaster information extraction from the single high-resolution SAR image for the entire traffic emergency response level.To solve the problems encountered in practical application,this paper carry out the following work:(1)An adaptive SAR speckle noise suppression method is studied based on the support vector machine nonlinear-fitting.The correlation between the support vector regression and the noise type parameters is analyzed in the logarithmic domain of the image.Thereby a noise reduction filter suitable for the distribution of each image itself is presented.(2)Research on SAR image classification of support vector machine based on two kinds of texture feature spaces of Gray Level Co-occurrence Matrix and Gabor filter is studied.Then the GAC model is introduced to obtain SAR image classification which can take the global optimal and local precision into account.(3)Two categories of targets are extracted and analysis,based on the two-stage improved GAC model,combined with various geometric and structural features.The joint target refers bridges and airports;and the surface target refers landslide,dammed lake,and collapsed building area.Finally,the accuracy of the extraction is verified.