Change Detection In Synthetic Aperture Radar Images Base On Image Fusion And Fuzzy Clustering

In recent years, change detection in synthetic aperture radar (SAR) sensor has been attracted wide attention. In this paper, a systematic study is carried out for the SAR image change detection technique based on analysis of difference image, including the generation of difference image, the analysis of difference image and the real-time application of change detection technique that related to parallel computing. The detail of this paper can be described as follows:(1) A novel method that based on wavelet image fusion for generating difference image is proposed. In this method, wavelet multi-scale fusion is introduced to generate difference image by using complementary information from mean-ratio image and log-ratio image. Wavelet fusion rules that based on average operator and minimum local area energy are chosen to combine the advantage of both methods, it helps to inhibit the speckle noise effectively and improve the contrast between the unchanged and changed classes.(2) A reformulated classification method based on fuzzy c-means clustering is proposed for classifying changed and unchanged regions in the difference image. In this method, the problem of change detection is studied in the perspective of fuzzy clustering, so it can be transformed in image segmentation. This reformulated fuzzy clustering algorithm can incorporate the information about both local gray and spatial context in a novel fuzzy way for the purpose of reducing the effect of speckle noise and enhancing the cluster performance. In general, this method makes a balance between noise-immunity and the preservation of image detail.(3) A novel method that based on GPU parallel computing cluster for fuzzy clustering is proposed to meet the challenge of real-time processing of massive data and an extortionate time complexity with fuzzy clustering algorithm. The parallel programming model that GPU cluster applied is MPI+CUDA. Firstly, the data of difference image will be assigned to each compute node in cluster, and then the assigned data are computed on GPU with parallel computing. This system can reduce the execution time effectively. It has a very important practical significance in the application of massive image data change detection.