地球信息科学学报 ›› 2012, Vol. 14 ›› Issue (4): 531-539.doi: 10.3724/SP.J.1047.2012.00531

• 遥感技术与应用 • 上一篇    下一篇

基于特征概率函数的双阈值分割海面溢油检测

李煜1,2, 张渊智*1,3, 陈杰2   

  1. 1. 香港中文大学太空与地球信息科学研究所, 香港 沙田;
    2. 北京航空航天大学电子信息工程学院, 北京 100037;
    3. 香港中文大学圆玄卫星遥感研究中心, 香港 沙田
  • 收稿日期:2011-09-15 修回日期:2012-07-14 出版日期:2012-08-25 发布日期:2012-08-22
  • 通讯作者: 张渊智(1964-),男,教授,主要研究方向为遥感信息机理与应用。E-mail:yuanzhizhang@cuhk.edu.hk E-mail:yuanzhizhang@cuhk.edu.hk
  • 作者简介:李煜(1986-),男,博士研究生,主要研究方向为遥感图像处理与应用。E-mail:liyu_buaa@126.com

Dual-threshold Oil Spills Detection Based on Characteristic Possibility Function

LI Yu1,2, ZHANG Yuanzhi*1,3, Chen Jie2   

  1. 1. Institute of Space and Earth Information Science, The Chinese University of Hong Kong, Shatin, Hong Kong, China;
    2. School of Electronic and Information Engineering, Beijing University of Aeronautics and Astronautics, Beijing 100037, China;
    3. Yuen Yuen Research Centre for Satellite Remote Sensing, The Chinese University of Hong Kong, Shatin, Hong Kong, China
  • Received:2011-09-15 Revised:2012-07-14 Online:2012-08-25 Published:2012-08-22

摘要:

海面溢油对生态环境造成了严重危害,故及早发现和尽快处理对降低事故影响和经济损失起着至关重要的作用。合成孔径雷达(SAR)是观测海面溢油、快速检测和事故态势分析判断的有效技术途径。本文针对SAR图像的海面溢油检测,提出了一种特征概率函数的双阈值分割方法。首先,通过高低阈值分割提取不同层次的灰度信息,再利用密度估计提取灰度的空间分布信息,然后,通过构建概率函数对油膜和类油膜区域进行形态学分类,最后,结合辅助信息,获得最终的海面溢油检测结果。本文利用香港中文大学卫星地面站接收的ENVISAT ASAR图像开展实验,结果表明,本文提出的方法能够准确地排除由风场或者水流场导致的低散射区域,有效地检测和识别生成不久的中型油膜,从而有助于溢油事故的早期预警与处置。

关键词: 分类, 溢油检测, SAR图像, 黑斑提取

Abstract:

Oil spills can cause huge damage to the ecology of marine environment and its detection and clean-up plays a very important role in the reduction of economic and environmental losses. As an advanced technology, synthetic aperture radar (SAR) has the advantage of day-night all weather observation capability. Besides, SAR has relatively wide swath width and high resolution, which all helps a lot in the early warning and damage analysis of oil spills accidents. Due to the special imaging mechanism of SAR, oil spills can be found as dark spots in SAR images. However, there still remain a lot of difficulties in the related detection and classification algorithms. In this paper, a double-threshold oil spills detection based on characteristic possibility function was proposed for taking the best advantages of backscatter information contained in different grayscale levels. Both high and low levels of grayscale information were extracted from the backscatter image obtained from SAR signal. Then the density of pixels was evaluated by Gauss kernel to enhance the stability of the segmentation. By using high level segmentation result, look-alikes with large area are classified from oil spills by basic morphological analysis. By taking advantage of low threshold grayscale information, other look-alikes were distinguished from oil spills by means of probability likelihood function derived from morphological characters such as complexity, length to width ratio, Euler number, etc.. Finally, the detected spills were obtained by fusing classification result of different level and other auxiliary information. The proposed method was implemented on EVISAT ASAR images of coastal region around Hong Kong received by the satellite ground station, CUHK. Experimental results demonstrated that real oil spills and look-alikes generated by other natural phenomena such as low wind speed and internal water turbulence could be distinguished accurately and effectively. This method can be further developed and has potential use in the surveillance and early alarm of marine and coastal oil leak accidents.

Key words: SAR imagery, dark spot extraction, oil spill detection, classification