本文提出一种基于内港岸线特征谱的港口识别方法.进行海域分割,提取海岸线及其特征点.通过分析海岸线封闭性测度矩阵得到内港岸线及其特征谱.根据特征谱的匹配相似度实现港口识别.从理论上证明了该方法的合理性,并在实际遥感图像上进行了港口识别实验,结果表明:内港岸线的特征谱具有平移、旋转、缩放不变性;文中方法能够准确分割内港岸线,对现有港口目标样本的识别准确率达到100%.

A novel port recognition algorithm based on the feature spectrum of inner port coastline is presented.The algorithm distinguishes sea from land and extracts coastline and its dominant points in the remote sensing image.A closure matrix is computed,whose element reflect the close degree of the coastline between a pair of dominant points.The inner port coastline and its feature spectrum are extracted based on the closure matrix.The unknown port is recognized based on comparing its feature spectrum to the templets which have been stored anteriorly.Experiment reveals that the feature spectrum of inner port coastline is a translation,scale and rotation invariant.It is also robust to the changes often appearing in remote sensing images.All port rcognition experiments acquire correct result.