基于KFCM和改进CV模型的海面溢油SAR图像分割  

Marine spill oil SAR image segmentation based on KFCM and improved CV model

作　　者：吴一全[1,2,3,4] 郝亚冰[1] 吴诗婳[1] 张宇飞[1] 谢乾坤[1] 

机构地区：[1]南京航空航天大学电子信息工程学院,南京210016 [2]国家海洋局海洋溢油鉴别与损害评估技术重点实验室,青岛266033 [3]水声通信与海洋信息技术教育部重点实验室(厦门大学),厦门361005 [4]中国科学院海洋环流与波动重点实验室,青岛266071

出　　处：《仪器仪表学报》2012年第12期2812-2818,共7页Chinese Journal of Scientific Instrument

基　　金：国家海洋局海洋溢油鉴别与损害评估技术重点实验室开放研究基金(201112);水声通信与海洋信息技术教育部重点实验室(厦门大学)开放课题(201101);中国科学院海洋环流与波动重点实验室开放基金(KLOCAW1110);江苏高校优势学科建设工程资助项目

摘　　要：图像分割是合成孔径雷达(synthetic aperture radar,SAR)图像海面溢油检测的关键步骤之一,将核模糊C均值(kernel fuzzy C-means,KFCM)聚类方法及Chan-Vese(CV)模型应用于海面溢油SAR图像分割,为了解决单一KFCM方法分割精度不够高,及传统CV模型对初始条件敏感和收敛速度低的问题,提出了一种基于KFCM和改进CV模型的海面溢油SAR图像分割方法。首先利用KFCM算法将海面溢油SAR图像从原始样本空间映射到高维特征空间,得到聚类结果;然后将其作为CV模型的初始条件,以降低CV模型对初始条件的敏感性,并利用图像边缘强度取代传统CV模型中的Dirac函数,以提高模型的收敛速度和对不同SAR图像的适应性。大量实验结果表明,所提出的基于KFCM和改进CV模型的海面溢油SAR图像分割方法具有分割精度高、运算速度快的优点。Image segmentation is a key step in marine spill oil detection of synthetic aperture radar（SAR） images;in this paper, the kernel fuzzy C-means （KFCM） clustering method and Chan-Vese （CV） model are applied to marine spill oil SAR image segmentation. To solve the problems that the image segmentation accuracy of simple KFCM clustering method is not high enough, and traditional CV model is sensitive to the initial condition and has low convergence speed, a marine spill oil SAR image segmentation method based on KFCM clustering and improved CV model is proposed. Firstly, the marine spill oil SAR images are mapped into high dimensional feature space from original sample space using KFCM algorithm, and the clustering results are obtained. Then the clustering results are served as the in- itial condition of the CV model to reduce the sensitivity to the initial condition. The Dirac function in traditional CV model is replaced with the image edge intensity, so as to improve the convergence speed and adaptability to different SAR images. A large number of experi- mental results show that the proposed method based on KFCM clustering and improved CV model for marine spill oil SAR image segmen- tation has the advantages of high segmentation accuracy and fast computational speed.

关 键 词：海面溢油检测 SAR图像 图像分割 核模糊C均值聚类 Chan—Vese模型 

分 类 号：TN911.73[电子电信—通信与信息系统] TN391.41[电子电信—信息与通信工程]