基于GA-WNN的极化SAR海洋溢油检测方法研究

Ocean oil-spill detection using Pol-SAR data based on GA-WNN

海洋溢油对海洋生态和人类生活带来严重的影响。由于合成孔径雷达(Synthetic Aperture Radar,SAR)具有全天时全天候的工作能力,在海洋溢油检测中发挥重要作用。目前,极化SAR是SAR探测技术的先进手段。本文利用6个极化特征进行溢油检测,通过对比分析这些特征对不同溢油的检测能力,得出单一极化特征在溢油检测中存在不足。通过J-M特征优选方法,提取出溢油检测识别度较高的特征影像,并利用遗传算法优化的小波神经网络(Genetic Algorithm-Wavelet Neural Network,GA-WNN)进行溢油检测。利用2套Radarsat-2全极化数据进行了方法验证,结果表明,该方法优于其他检测方法,溢油检测精度分别达到90.31%和95.42%。

Ocean oil spills seriously threaten both the marine environment and human activity. Synthetic aperture radar （SAR） plays an important role in ocean oil-spill detection due to its all-weather and day-and-night capabilities. Polarimetric SAR （Pol-SAR） is an advanced SAR detection technology that makes full use of the backscattering characteristics between SAR channels and has demonstrated obvious advantages in ocean oil-spill detection. We conducted experiments to investigate six polarimetric characteristics, based on the fact that a single characteristic can be inadequate in oil-spill detection with respect to the analysis of different features. Using the J-M distance index method to perform feature selection, we then used the genetic-algorithm-optimized wavelet neural network （GA-WNN） to detect oil spills. The experimental results from two sets of Radarsat-2 data confirm the superior accuracy of the proposed method with regard to oil-spill detection, i.e., 90.31% and 95.42%, respectively.