Research On SAR Intensity Image Change Detection

Synthetic Aperture Rader(SAR)image change detection is a technology of detecting change in multiple images of the same scene taken at different times.The technology has been more widely applied in such fields as damage assessment,monitoring and analysis of vegetation changes,monitoring military objective and obtain military intelligence,civil infrastructure.At present.,SAR image change detection has been widely studied,these approaches to change detection generally fall into three categories: straightforward approaches on the basis of single-channel intensities,coherent interferometric change detection,and incoherent change on the basis of polarimetric signatures.This paper research on SAR image amplitude detection approaches based on difference image classification,focusing on obtaining difference image,analyzing the difference image and postprocessing.In general,using pixel and statistics of pixel patch to obtain difference image and using threshold method to partition the difference image into changed area and unchanged area.Traditional threshold methods are k-means,Otsu,etc.In order to improve the performance,this paper proposed some novel algorithm based on researched.The main research contents and work are as follows:1?First,the properties of SAR image been introduced,for example,geometrical characteristics,statistical distribution characteristics,properties of speckle noise.Then,concentrate on introducing the general procedure and the most fundamental approaches of SAR image amplitude change detection based on pixel and based on the pixel patch.2 Analyzing the properties of this algorithm on difference image based on traditional SAR image approaches based on intensities,such as the approaches of image subtraction,image ratio and image log ratio.Focusing on studing the approaches of using combined difference image and k-means clustering for SAR image change detection and using principal component analysis and k-means clustering for change detection.This paper proposed the SAR image detection based on combined difference image and block-based classification.We have tested our method on real spaceborne SAR images with reference image and airborne X-SAR image,the result of experiment demonstrate the effectiveness of our method,comparing with typical methods,our method reduce the false alarms,miss alarms and evaluate the kappa index.Finally,we analyze the size of patch impact on performance.3?Researching on the typical methods based on first-order statistics of pixel patches and higher order statistics of pixel patches,in order to against the poor performance of the mean-ratio detector,the log-ratio detector,and against the limitation of the Gaussian-based Kullback-Leibler divergence,what's more,in order to avoid using a large-size dense neighborhood around each pixel to measure its change level.Therefore,we propose change detection between SAR images using key points and graph theory.Here,some experiment study on real spaceborne SAR images with reference image and airborne X-SAR image is carried out,the result validates the stability of the proposed approach,comparing with typical methods,our method evaluate the percentage of overall accuracy,and analyzing the effects of crucial parameters using the receiver operating characteristic curve(ROC).