一种PCNN和边缘检测协同的SAR图像分割算法  

SAR Image Segmentation Algorithm Based on PCNN and Edge Detection

作　　者：王彬[1,2] 王国宇[1] WANG Bin;WANG Guo-yu(Ocean University of China,College of Information Science & Engineering,Shandong Qingdao 266100,China;Qingdao University of Science & Technology,School of Information Science & Technology,Shandong Qingdao 266061,China)

机构地区：[1]中国海洋大学信息科学与工程学院,山东青岛266100 [2]青岛科技大学信息科学技术学院,山东青岛266061

出　　处：《现代防御技术》2021年第3期92-97,122,共7页Modern Defence Technology

摘　　要：针对高分辨率SAR(synthetic aperture radar)图像目标复杂,且具有严重的相干斑噪声,灰度出现剧烈起伏导致目标边缘模糊难以实现精准分割的问题,提出了一种分数阶傅里叶变换(fractional Fourier transform,FrFT)域中里兹分数导数(Riesz fractional derivative,RFD)边缘检测和脉冲耦合神经网络(pulse coupled neural network,PCNN)协同的高分辨率SAR图像分割算法。该算法首先将原始图像经过边缘检测处理以保留较好的边缘信息,再由PCNN模型进行图像分割,最后通过形态学进一步去除相干斑点。将所提算法应用到不同区域的高分辨SAR图像分割中,实验结果表明该方法能够有效抑制相干斑噪声和灰度边界模糊的影响,获得精准的分割效果。Aiming at the problem that the target in high-resolution SAR image is complex,has serious coherent speckle noise,and the gray level fluctuates violently leading to blurred edges of the target for accurate segmentation,a high-resolution SAR image segmentation algorithm based on Riesz fractional Derived(RFD)edge detection and pulse coupled neural network(PCNN)in fractional Fourier transform(FrFT)domain is proposed.The original image is processed by edge detection to retain better edge information,and then the image is segmented by the PCNN model.The coherent spots are further removed by morphology.The proposed algorithm is applied to high-resolution SAR image segmentation in different regions,and the experimental results show that it can effectively suppress the influence of speckle noise and gray boundary-blurring,and achieve an accurate segmentation effect.

关 键 词：高分辨率SAR图像 边缘检测 图像分割 分数阶傅里叶变换 里兹分数导数 脉冲耦合神经网络 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置] TJ76[自动化与计算机技术—控制科学与工程]