基于线不变矩和封闭性的遥感图像港口识别

Port Recognition in Remote Sensing Images Based on Invariant Linear-moment and Closure

针对遥感图像中的港口目标识别问题,本文提出了一种基于海岸线不变线矩特征与港口海岸线封闭特征结合的识别算法。首先采用边界寻找和阈值分割的方法进行海陆分割和海岸线的提取,然后通过封闭性检测确定港口区域和属于港口区域部分的海岸线,将封闭度和线不变矩特征组成的特征谱作为识别港口的依据,最终实现了对港口的识别。实验结果表明,该算法能够准确地识别遥感图像中的港口目标。

Aimed at the characteristic of the port target revealed in remote sensing image, a port target detection algorithm based on the invariant linear-moment and the closure of the port region is proposed. Firstly, edge-search method and thresholding are used to realize the segmentation and obtain the binary image of sea and land, Then, a closure is computed, and through the closure of the coastline the port region, its inner coastline is confirmed. The invariant linear-moment of the port＇s inner coastline is computed, a feature spectrum consists of the invariant linear-moment and closure is got. The unknown port is recognized based on comparison against feature spectrum to the templets which have been stored anteriorly. Experimental results show that the proposed algorithm can correctly recognize the port target.