Synthetic Aperture Radar Image Of Buildings Damage Detection Based On Visual Cognition

Earthquake disaster,as one of the major sudden natural disasters,has a strong destructive power.Therefore,the rapid delineation of urban target damage after earthquake disaster can guide the scientific and effective post-disaster rescue work and minimize the loss.Synthetic Aperture Radar(SAR)has the ability to perform all-day and all-weather earth observation,playing an irreplaceable role in disaster emergency monitoring and disaster assessment with other remote sensing technologies.Due to the influence of special imaging mode and noise,SAR image has the problems of poor visual effect and difficult visual interpretation by human eyes,and the common change detection method is not accurate to extract damaged areas.With the continuous development of computer vision field,combining with biology and psychology,human research on visual information processing,visual cognition achieved progress in image processing,including visual significance test can let the human visual system to image the important information for rapid cognitive,significant regional features such as color,shape,texture and the surrounding area form a strong contrast,can be quickly get eyes.Therefore,in this paper,visual cognition is introduced into building damage detection after the earthquake,and different SAR data acquisition situations after the earthquake are studied.Aiming at the single phase SAR data acquired after earthquake,a visual optimization method for SAR image based on visual cognition was proposed.In order to obtain the monopolar SAR data before and after the earthquake,a fast detection method of damaged target area based on visual significance detection was proposed.Based on the polarimetric SAR data obtained before and after the earthquake,visual significance detection was introduced to quickly delineate the damaged target area by combining with the secondary scattering component in Yamaguchi polarization decomposition.This paper mainly studies in the following three aspects:(1)In view of the poor visual effect of single-phase SAR image after the earthquake and the difficulty in manual visual interpretation,an adaptive negative exponential transform method was proposed to optimize the visual effect of SAR image by effectively visualizing SAR image and obtaining images suitable for the cognition of visual system.By analyzing the statistical characteristics of SAR images,it is found that the SAR image scattering intensity value is too concentrated,and the distribution of the image scattering intensity value is quite different from the theoretical distribution,which affects the visual cognitive effect of human eyes on the image.Therefore,the negative exponential change was used to normalize the image to the range of 0-1,calculate the probability distribution mode and discriminant parameters of the pixel value after the negative exponential transformation,and automatically determine whether the negative exponential parameter correction was performed on the SAR images of scenes in different post-earthquake ranges,and finally mapped to the range of 0-255,so as to obtain the final visual optimization effect.In this experiment,the SAR image of a wide range of scenes was firstly used for visual optimization to obtain the urban building area,and then the visual effect of small range scenes in the urban building area was optimized adaptively.This method improves the effect of building damage detection using single phase SAR image after earthquake and reduces the difficulty of visual interpretation.(2)In view of the acquisition of unipolar SAR data before and after the earthquake,a fast detection method of damaged target area in SAR images based on visual significance was proposed.This method firstly uses the Z factor combining the intensity difference value and the value of the phase relation to detect the initial change of the damaged area of the building,introduces the Itti significance model to calculate the significance of the preliminary change map,and finally uses the image segmentation to obtain the scope of the damaged target area.High resolution ALOS 2 data before and after the Kumamoto earthquake in Japan and low resolution sentinel-1 data before and after the earthquake in Indonesia were selected for experiments.The results showed that this method was suitable for the delineation of damaged target areas of SAR data with different resolutions with high accuracy and less time consumption.(3)To obtain the polarization SAR data before and after the earthquake,due to the polarization of the SAR image record object target scattering measurement matrix,obtain more abundant features information,so through the analysis of intact and damaged buildings polarization scattering characteristics,selected the Yamaguchi decomposition of secondary scattering component as building characterization,difference for the secondary scattering component is calculated and Itti significant computation,significant figure damage and damage to a target zone is delineated.The ALOS PALSAR data from the east coast of Japan earthquake on March 11,2011 were selected for the experimental study and analysis,and the results showed that this method could accurately delineate the damaged target area with less time consumption.