基于改进CV模型的SAR图像水域检测方法  

Water Detection Method from SAR Images Based on Improved CV Model

作　　者：周小健[1] 于子桓 吴诗婳 ZHOU Xiaojian;YU Zihuan;WU Shihua(The 28th Research Institute of China Electronics Technology Group Corporation,Nanjing 210023,China)

机构地区：[1]中国电子科技集团公司第二十八研究所,南京210023

出　　处：《指挥信息系统与技术》2022年第1期64-68,73,共6页Command Information System and Technology

基　　金：装备发展部“十三五”预研课题(41412030901)资助项目。

摘　　要：为了能更快速、准确地将河流从背景中提取出来,提出了一种基于改进无边缘主动轮廓(Chan-Vese,CV)模型的合成孔径雷达(SAR)图像水域检测方法。首先,采用基于优化的人工蜂群(ABC)算法的倒数交叉熵多阈值选取方法提取初始水域;然后,在检测结果中引入移动因子改进传统CV模型,加快收敛速度的同时提高水域检测精度;最后,大量试验结果表明,该方法解决了单一阈值分割法分割精度不够高和传统CV模型收敛速度低、对初始条件敏感的问题,能更快速准确地分割出水域。To rapidly and accurately split rivers from the background,a water detection method from synthetic aperture radar(SAR)images based on improved Chan-Vese(CV)model is proposed.Firstly,the multi-threshold selection method with the reciprocal cross entropy based on the optimized artificial bee colony(ABC)algorithm is adopted to make a coarse segmentation for water area.Then,the motion factors are imported to improve the traditional CV model in the coarse segmentation results.Thus,the convergence is accelerated and the accuracy of the water detection is enhanced.Finally,a large number of experimental results show that the segmentation accuracy of the single threshold segmentation method is not high enough,the convergence speed of the traditional CV model is low,and the traditional CV model is sensitive to the initial condition.The method can solve the above problems and can segment the water area more quickly and accurately.

关 键 词：水域检测 SAR图像 倒数交叉熵 优化的ABC算法 带移动因子CV模型 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]