摘要
遥感图像中不同港口的内港区域呈现出不同的形状,该文提出一种基于内港区域的港口目标识别方法。首先利用直方图和形态学算子分割海域;再利用多边形近似法提取海岸线上的特征点,根据特征点的闭合度确定内港区域,通过检测直线和角点结构减少虚警;然后计算内港区域的形状主方向;最后提取特征并进行相似性度量。实验证明,该方法检测的港口目标准确,提取的特征具有良好的不变性,满足识别的要求。
Different inside regions of harbors show different shapes from remote sensing images. A method of harbor target recognition based on inside region is proposed. Firstly, potential sea regions are segmented from image by using histogram and dilation operations. Secondly, the coastline is partitioned into a series of feature points with polygon approximation in order to compress the shape description datum. After calculating the closure from feature points, the inside regions are extracted. Straight lines and corner structures are detected to reduce false alarms. Then the principal orientation of inside region is rotated to the same right horizontally to avoid the rotation effect. Finally, the features are extracted to measure the similarity. Experiments show that harbors are detected accurately using the proposed method. Besides, the extracted features have good invariability and can meet the need of recognition.
出处
《电子与信息学报》
EI
CSCD
北大核心
2009年第6期1275-1278,共4页
Journal of Electronics & Information Technology
关键词
遥感图像
港口识别
内港区域
闭合度
Remote sensing images
Harbor recognition
Inside region
Closure
