基于多尺度信息融合的SAR图像边缘检测-电路与系统专业论文.docxVIP

AbstractAbstract Abstract Abstract The study of edge detection for Synthetic Aperture Radar(SAR)has been an important subject on processing and interpretation of SAR images：However SAR images are affected by strong correlative speckle noise，SO the result of traditional edge detection algorithms is not satisfactory．It makes edge detection of SAR images to be a challenging problem． Considering the advantages and disadvantages of Nonsubsampled Contourlet transform(NSCT)and Ratio of average algorithm in edge detection of SAR images， this paper proposes a new edge fusion detection method based on Dempster-Shafer(D-S) evidence theory．We firstly analyze theoretically that NSCT has a s仃ong ability of weak edge detection and Ratio of average algorithm can effectively suppress the influence of speckle noise．Then three mass functions are constructed by using amplitude variation， direction and angle information of coefficients of NSCT and Ratio．Finally D．S evidence theory and its modified method ar e introduced to solve the problem of evidence conflict．The proposed method combines the advantages of these two algorithms．Experiment results demonstrate that our algorithm Can accurately detect edges while effectively suppressing the influence of speckle noise，especially for weak edges．The fusion edges detected of our algorithm are complete and accurate． Based on the ideas of information fusion and image segmentation，Fuzzy C—means clustering(FCM)is used to fuse Ratio and wavelet transform to detect SAR image edges．According to the characteristics of SAR image edges，two characteristics values are defined．Then coe髓cients of Ratio and wavelet transfonll with a reasonable treatment were also used as the characteristics of SAR．So a four-dimensional feature vector of image feature Can be obtained．Clustering was carried out by FCM based on the collection of feature vectors，then the information of image edges iS obtained． Finally，the lacks of edge line are connected and the short lines