多源高分辨率卫星遥感海岸线数据集  

Multi-source High-resolution Satellite Remote Sensing Coastline Dataset

作　　者：曹恒硕 丁宁[2] 钱建军 王厚军[2] 李广宇 CAO Hengshuo;DING Ning;QIAN Jianjun;WANG Houjun;LI Guangyu(Key Lab of Intelligent Perception and Systems for High Dimensional Information of Ministry of Education,Nanjing University of Science and Technology,Nanjing 210014;National Ocean Technology Center,Tianjin 300112)

机构地区：[1]南京理工大学高维信息智能感知与系统教育部重点实验室,南京210014 [2]国家海洋技术中心,天津300112

出　　处：《计算机与数字工程》2025年第1期202-208,共7页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:62006119);江苏省基础研究计划项目(编号:BK20190444)资助。

摘　　要：岸线研究可为海岸带科学规划利用、海洋生态保护以及岸线高效管理等方面提供巨大支持,已成为当前研究热点,海岸线数据集在其中扮演着重要角色。论文构建了高分辨率卫星遥感海岸线数据集,该数据集具有数据源多样化、岸线覆盖范围广、图像分辨率高、标注实例数目大等优势。首先,基于四颗卫星,论文收集海量多源海岸线原始遥感图像,并对其进行数据预处理以消除畸变;其次,提出基于实地测绘和目视解释融合的类别标签提取方法,以获取多源卫星遥感图像类别标签;再次,基于设计的自动化裁剪方法,对整幅带标签遥感图像进行智能化裁剪,得到多源高分辨率遥感海岸线数据集MHRSCD(Multi-source High-resolution Remote Sensing Coastline Dataset);最后,利用六种经典深度学习网络模型,论文开展一系列验证实验,以证明构建数据集的有效性和实用性。As a hot academic topic,the coastline studies are able to provide wide supports to scientific planning and utiliza⁃tion of coast zones,marine conservations,efficient coastline managements and so on,in which coastline datasets play an important role.The paper establishes an available coastline dataset by means of high-resolution satellite remote sensing images,and this data⁃set indicates various advantages,namely different types of data sources,large coastline cover range,high-resolution remote sensing pictures,enormous labelled instance number and so forth.Specifically,first of all,based on four remote sensing satellites,this pa⁃per collects a large number of original multi-source coastline images,which are then pretreated to remove redundant picture distor⁃tions.In addition,by combining field surveying and visual interpretation,an efficient classification label extraction algorithm is pro⁃posed to obtain multisource satellite remote sensing image categories.Furthermore,by means of the designed automatic cropping method,this paper implements intelligent cutting operations to divide the whole labelled images into several small pictures,to ob⁃tain the useful coastline dataset called MHRSCD.Finally,a series of experiments are carried out to prove the effectiveness and prac⁃ticability of the established dataset by virtue of six classical deep learning network models.

关 键 词：卫星遥感 多源 海岸线 数据集 深度学习 

分 类 号：P748[天文地球—海洋科学]