Research On Edge Detector For SAR Image

With the development of SAR,edge detection in SAR images has become an increasing research topic,where the most representative ratio-based detector won the wide attention of researchers for its good direction.However,the ratio-based detector has some problems: it uses the same weight to calculate local mean,which leads to high frequency noise in the detection results;when suppressing the noise and false edges in the detection process,low-contrast edge will also be suppressed,resulting in the loss of the edge or edge fracture problems;the threshold selection is realized by adjusting parameters manually through repeated experiments,in this situation,the obtained threshold is probably not the best one,so as to affect the detection results.In this thesis,two edge detection algorithms are proposed for the above mentioned problems in SAR image edge detection.One is Gaussian-Gamma-Shaped bi-windows and Edge Compensation(GGSEC)algorithm,the other is ROEWA-based adaptive threshold selection edge detection(DROEWA_IOtsu)algorithm.These two algorithms are described as follows:(1)The GGSEC algorithm uses the change weight of GGS window to calculate the local mean value,reducing the high frequency noise in the detection result;and also introduces the edge compensation method,solving edges loss and edge crack problems effectively in the ratio-based detector for SAR image edge detection.The process of GGSEC algorithm is as follows: firstly,get the candidate edge pixels set using GGS detection method;secondly,calculate the edge strength ratio to get weak candidate edge pixels set;then,use edge compensation algorithm to enhance the weak candidate edges;finally,the nonmaxima suppression and the hysteresis threshold is used to extract edge.GGSEC algorithm use edge compensation method to enhance the weak candidate edge,to solve the problem of edge loss and edge crack effectively,especially the loss of weak edge.The ROC curve is used to verify the validity of the GGSEC method,and several experiments are compared to verify the effectiveness of GGSEC algorithm in both synthetic SAR and real SAR image.(2)The DROEWA_IOtsu algorithm uses change weight of ROEWA operator to calculate the local mean,reducing the high frequency noise;and presents the calculation method of the edge direction,solving the problem that ROEWA operator cannot determine the edge direction;and also calculates the optimal threshold using the adaptive threshold method to avoid selecting threshold manually.The main features of this algorithm are that it does not need the range template to estimate edge direction,and does not adjust the threshold manually.The main work of DROEWA_IOtsu algorithm is as follows: the ratio of the ROEWA is redefined with the theory verification and the DROEWA operator is proposed,and also gives the calculation method of the edge direction,solving the problem that the ROEWA operator cannot determine the edge direction in edge detection process;combined with the characteristics of the optimal threshold to improve the Otsu method,the IOtsu threshold selection method is proposed,which gives the calculation method of the optimal threshold,and solves the manual selection threshold in SAR edge detection process.The ROC curve is used to verify the validity of the DROEWA_IOtsu method,and several experiments are compared to verify the effectiveness of DROEWA_IOtsu algorithm in both synthetic SAR and real SAR image.