基于高分辨率影像的海岸防波堤识别  

Port target recognition based on high resolution images

作　　者：陈雨欣 姜晓轶[2,3] 张文 王昶 闫仕娟[4] CHEN Yuxin;JIANG Xiaoyi;ZHANG Wen;WANG Chang;YAN Shijuan(School of Civil Engineering,University of Science and Technology Liaoning,Anshan 114051,China;National Marine Data and Information Service,Tianjin 300171,China;Technology Innovation Center of Marine Information,Tianjin 300171,China;The First Institute of Oceanography,MNR,Qingdao 266061,China)

机构地区：[1]辽宁科技大学土木工程学院,辽宁鞍山114051 [2]国家海洋信息中心,天津300171 [3]自然资源部海洋信息技术创新中心,天津300171 [4]自然资源部第一海洋研究所,山东青岛266061

出　　处：《海洋信息技术与应用》2024年第3期172-177,共6页JOURNAL OF MARINE INFORMATION TECHNOLOGY AND APPLICATION

基　　金：国家重点研发计划(2022YFC2803600);辽宁省教育厅面上项目(LJKMZ20220638)。

摘　　要：基于对遥感影像中防波堤空间位置关系特征和其相关先验知识的分析,本文从海陆分割以及防波堤提取、识别三方面进行近海岸目标检测。首先,对高分二号影像进行预处理,将经过处理后影像的二值图通过改进的区域生长法进行海陆分割;其次,根据遥感影像上防波堤的相关先验知识以及空间位置关系特征将防波堤从海岸线中提取出来;最后,利用霍夫变换对防波堤进行识别。实验分别对两种类型的防波堤进行提取,平均总体精度为96.41%,平均Kappa系数为0.9603。结果表明,该方法能够准确、稳定、高效地实现对防波堤的自动识别。Based on the analysis of the spatial relationship characteristics of breakwater in remote sensing images and its related prior knowledge,this paper conducts nearshore target detection from three aspects:sea-land segmentation,breakwater extraction and recognition.Firstly,preprocessing is applied to the high-scoring No.2 image,and sea-land segmentation is performed on the binary image of the processed image using an improved region growing method.Secondly,the breakwater is extracted from the coastline according to the prior knowledge and spatial relationship characteristics of the breakwater in remote sensing images.Finally,the breakwater is identified by Hough transform.Experiments are conducted to extract two types of breakwaters,with an average overall accuracy of 96.41%and an average Kappa coefficient of 0.9603.The results show that this method can accurately,stably,and efficiently achieve automatic recognition of breakwater.

关 键 词：防波堤 海陆分割 信息提取 边缘检测 目标识别 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置]