SAR Image Change Detection Based On Self-supervised Learning Algorithm

Remote sensing image change detection is a process of obtaining geographic information change by analyzing the differences in pixels,texture,structure and other aspects between remote sensing images taken at different times in the same place.Synthetic aperture radar(SAR)has the advantages of high shooting accuracy and strong penetration.Its imaging process is not affected by factors such as weather,light intensity and so on.It has the characteristics of all-weather operation.Therefore,SAR image is one of the important sources of remote sensing data.In the past decade,remote sensing image change detection technology has been widely concerned by relevant industry researchers because of its application value.Among them,SAR image change technology is widely used in many fields,such as environmental change monitoring,land cover data acquisition,urban planning,and strategic research,Focusing on how to improve the accuracy of change detection,the following work has been completed.In the SAR change detection algorithm based on self-supervised learning,speckle noise reduces the difference image quality.Therefore,the contrast of the difference image is low,and its change area is not significant.Moreover,the pre-classification algorithm with the poor robustness makes the classification results of the low-quality difference image inaccurate.When the wrong labels are sent into the classification network,the accuracy of the final detection results is reduced.First,to improve the quality of the initial difference image,we design an adaptive gamma correction algorithm that adjusts the contrast according to the mean value of the initial difference image and the variation coefficient.The contrast of the new difference image generated by this algorithm is higher.Furthermore,to suppress the noise,we adopt a new algorithm based on popular ranking to obtain the saliency map of the new difference image.Combining the initial difference image with this saliency map,a high-quality difference image with a low noise level is obtained.After that,we introduce the structure tensor into the fuzzy local information Cmeans clustering algorithm to classify the difference image more accurately.The new clustering algorithm improves the accuracy of pre-classification,especially its hierarchical version.Besides,we use the structure tensor to generate the structure maps of the original images.Finally,according to the prior information obtained from the preclassification,we use a convolution wavelet neural network(CWNN)to supervise and train the structure maps of the original images.Experimental results show that compared with other methods,the difference map generated by the proposed algorithm is closer to the ground truth.Similarly,our preclassification algorithm performs better.The detection accuracy of this algorithm for strong noisy SAR images is higher than some advanced change detection algorithms,Taking the Coastline dataset as an example,the Kappa value has increased by 21.8%compared with the latest algorithm,Which shows the effectiveness of the algorithm proposed in this paper.