基于小波模极大值的声纳图像特征提取与描述研究  

Sonar Image Feature and Description Based on Wavelet Transform Modulus Maximum

作　　者：石守东[1] 徐德民[1] 李俊[1] 

机构地区：[1]西北工业大学航海学院,陕西西安710072

出　　处：《中国造船》2009年第2期171-176,共6页Shipbuilding of China

摘　　要：海底地图特征提取,是水下航行器(AUV)在未知环境中同时定位与地图创建(SLAM)的关键技术之一。本文针对连续的多波束声纳图像,提出了一种基于小波模极大值法的图像特征提取与描述算法。运用小波模极大值法获取轮廓及模方位,并在轮廓图像中寻找极大区域;以极大区域中心点作为图像特征点,且将区域模方位、特征点区域图像重心坐标组合生成特征点特征向量。实验结果表明:对连续图像,该算法能提取稳定有效轮廓,在边缘上获得图像的特征点一致很好。The feature extraction of sonar image is one of the key technologies for automobile underwater vehicle＇s SLAM navigation. In this paper, we describe a method of acquiring sonar image feature and description based on Wavelet Transform Modulus Maximum. In allusion to the sequence of sonar images from automobile Underwater vehicle, we will first get image modulus and direction matrix of modulus. Then obtain the modulus maximum region and calculate the center point of maximum region as image feature point. Feature vectors of this feature point will be created by the combination of region direction of modulus and barycentric coordinate of the region. The experimental results show that the method is valid and stable for the profile and is good consistent for the feature point between two sequences of sonar images.

关 键 词：船舶 舰船工程 声纳图像 模极大值 特征 水下航行器 

分 类 号：U666.7[交通运输工程—船舶及航道工程]