基于边缘信息和区域特征的水平集侧扫声呐图像分割研究

项目名称： 基于边缘信息和区域特征的水平集侧扫声呐图像分割研究

项目编号： No.41306089

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 天文学、地球科学

项目作者： 霍冠英

作者单位： 河海大学

项目金额： 24万元

中文摘要： 目标及其阴影的准确分割是基于侧扫声呐图像进行水下目标探测和识别的关键。然而，侧扫声呐图像噪声强烈、灰度及几何畸变严重，如何获得准确而鲁棒的分割结果，亟待解决。为此，在畸变校正的基础上，研究边缘信息和区域特征相结合的水平集侧扫声呐图像分割方法。首先探索消除灰度不均、突出目标边缘特征的灰度校正方法以降低分割难度；并尽可能消除斜距成像、偏航和航速变化等引起的几何畸变，以使目标形态更加真实。研究基于多尺度几何变换的边缘信息精确提取方法，并同时探索噪声鲁棒的区域特征提取方法。综合提取出的边缘信息和区域特征，并考虑先验约束构造分割能量函数，进而用水平集方法求解得到边缘定位准确、噪声干扰较少的分割结果。本课题的研究成果将有助于侧扫声呐图像的准确解译，更好地发挥侧扫声呐在水下目标识别和海洋科学研究等领域的重要作用。

中文关键词： 侧扫声呐图像分割；斑点噪声；灰度畸变；边缘信息；水平集

英文摘要： For underwater target detection and recognition based on side-scan sonar images, correct segmentation of targets and their shadows is critical. But side-scan sonar images have strong noise, serious gray distortion and geometric distortion, and hence how to obtain accurate and robust segmentation results is still a problem to be solved. To solve this problem, after proper distortion correction, level set side-scan sonar image segmentation based on the combination of edge information and region features is studied. Gray correction which can eliminate uneven gray distribution and highlight target edge is proposed to reduce segmentation difficulty, and geometric distortion caused by slant range imaging, yaw and speed change is corrected to make target shape more real. Accurate edge information extraction method based on multi-scale geometric transform and noise-robust region features extraction method are both studied. Segmentation energy functional with edge information, region features and a priori constraint is constructed, and solved by the level set method to get the final segmentation results with edge location accuracy and noise robustness. The research results will help to correctly interpret side-scan sonar images, and therefore can provide better support for side-scan sonar to play an important role in fie

英文关键词： Sidescan sonar image segmentation；speckle noise；gray distortion；edge infromation；level set