%0 Journal Article %T 结合层次聚类的海洋内波SAR图像特征自动检测
Automatic Feature Detection of Ocean Internal Wave SAR Image Based on Hierarchical Clustering %A 许素芹 %A 陈捷 %A 李婷婷 %A 陶荣华 %J Computer Science and Application %P 1118-1125 %@ 2161-881X %D 2019 %I Hans Publishing %R 10.12677/CSA.2019.96126 %X
海洋内波是一种典型的海洋现象,其特征检测和参数反演研究对海洋生态系统、海洋渔业、海洋工程和军事等具有重要意义。海洋内波在合成孔径雷达(SAR)图像上的明暗条纹特征,使SAR图像成为检测海洋内波特征的重要数据源。针对复杂的多簇海洋内波SAR图像,本文首先利用高斯滤波、Canny算子对SAR图像中多簇海洋内波特征进行检测,通过非内波边缘剔除、特征曲线搜索,结合层次聚类分析,进行了海洋内波特征的自动提取和多个内波波包的识别与分离。基于计算机视觉库Opencv,利用RADARSAT卫星SAR图像进行了海洋内波特征检测,检测结果表明该方法有效解决了内波特征自动检测问题,为海量SAR图像海洋内波特征参数自动反演和星上快速处理奠定了基础。
Ocean internal wave is a typical ocean phenomenon, and its characteristic detection and parameter inversion are of great significance to marine ecosystem, marine fishery, marine engineering and military affairs. The characteristics of light and dark stripes of ocean internal waves on synthetic aperture radar (SAR) images make SAR images become an important data source for detecting ocean internal wave features. In this paper, for complex multi-cluster SAR images of ocean internal waves, Gauss filter and Canny operator are used for the feature detection of multi-cluster ocean internal wave in SAR image, and automatic extraction of ocean internal wave features and identification and separation of multiple internal wave packets are realized by eliminating the edge of non-internal wave, searching the feature curve and combing the method of hierarchical clustering analysis. Based on computer vision library Opencv, ocean internal wave feature detection is carried out based on the SAR image of RADARSAT. The detecting results show that this method can effectively solve the problem of internal wave feature automatic detection, which lays a foundation for automatic inversion of ocean internal wave feature parameters and onboard fast processing of massive SAR images.
%K SAR图像,海洋内波特征,Canny算子,层次聚类,自动检测
SAR Image %K Ocean Internal Wave Characteristics %K Canny Operator %K Hierarchical Clustering %K Auto-matic Withdrawal %U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=30898