Digital image processing techniques for enhancement and classification of SeaMARC II side scan sonar imagery
Digital image processing techniques for enhancement and classification of SeaMARC II side scan sonar imagery
复制标题
用于 SeaMARC II 侧扫声纳图像增强和分类的数字图像处理技术
DOI：
--
复制
发表时间：
1989
期刊：
影响因子：
--
通讯作者：
D. Hussong
中科院分区：
文献类型：
--
作者：
T. Reed;D. Hussong
近年来，数字侧扫声纳图像的生产率不断提高，再加上系统的海底勘探计划迅速扩大，因此需要快速和定量的海底图像处理手段。计算机辅助分析技术填补了这一需求。本文件记录了用于增强SeaMARC II（一种远程组合侧扫声纳和测深海底测绘系统）生成的图像并对其进行分类的若干数值技术。提出了三类技术：（1）预处理校正（辐射和几何），（2）特征提取，（3）图像分割和分类。介绍了“特征向量”的概念，沿着解释了基于灰度共生矩阵（GLCM）的纹理特征向量的评价方法。一种替代的先验纹理元素（纹理元素）细分的图像的区域生长和纹理分析（REGATA）的形式。这个程序提供了一个空间分辨率上级于用任意指定纹素边界可获得的空间分辨率的纹理图，并使由于任意指定纹素中两个或多个纹理的组合而产生的混合纹理信号的可能性最小化。通过灰度共生矩阵技术提取的这些纹理特征的计算机分类导致图像变换成图像纹理的地图。这些地图可以根据纹理签名和波长之间所示的理论关系来解释，或者通过纹理签名与地面实况数据的相关性来转换为地质地图。这些技术适用于SeaMARC II侧扫声纳图像从各种地质环境，包括石化和非石化沉积地层，火山和沉积泥石流，结晶玄武岩露头。上述处理步骤的应用不仅提供了用于主观和定量分析的上级图像，而且还提供了区分具有不同岩性但图像强度相似的露头的关键能力。
The recent growth in the production rate of digital side scan sonar images, coupled with the rapid expansion of systematic seafloor exploration programs, has created a need for fast and quantitative means of processing seafloor imagery. Computer-aided analytical techniques fill this need. A number of numerical techniques used to enhance and classify imagery produced by SeaMARC II, a long-range combination side scan sonar, and bathymetric seafloor mapping system are documented. Three categories of techniques are presented: (1) preprocessing corrections (radiometric and geometric), (2) feature extraction, and (3) image segmentation and classification. An introduction to the concept of “feature vectors” is provided, along with an explanation of the method of evaluation of a texture feature vector based upon gray-level co-occurrence matrices (GLCM). An alternative to the a priori texel (texture element) subdivision of images is presented in the form of region growing and texture analysis (REGATA). This routine provides a texture map of spatial resolution superior to that obtainable with arbitrarily assigned texel boundaries and minimizes the possibility of mixed texture signals due to the combination of two or more textures in an arbitrarily assigned texel. Computer classification of these textural features extracted via the GLCM technique results in transformation of images into maps of image texture. These maps may either be interpreted in terms of the theoretical relationships shown between texture signatures and wavelength or converted to geologic maps by correlation of texture signatures with ground truth data. These techniques are applied to SeaMARC II side scan sonar imagery from a variety of geologic environments, including lithified and nonlithified sedimentary formations, volcanic and sedimentary debris flows, and crystalline basaltic outcrops. Application of the above processing steps provided not only superior images for both subjective and quantitative analysis but also the critical ability to discriminate between outcrops with distinct lithologies but similar image intensity.