2019-2021年三都澳近海岸海面目标可见光遥感图像数据集

A dataset of the visible light remote sensing images for offshore maritime targets in Sanduao from 2019 to 2021

利用计算机视觉技术对近海岸海面目标进行智能检测,可为海洋行政管理、海洋环境监督管理和海洋环境保护政策的制定提供科学依据,为经济的稳健发展提供有力的环境信息参考。本数据集采集区域为中国南部福建省宁德市东南部三都澳港湾,数据来源为谷歌地球,数据的时间跨度为2019年至2021年。本数据集包含不同季节、背景、光照条件下获取的1761张可见光遥感图像及对应的水平目标检测标签、旋转目标检测标签、语义分割标签,涵盖3种近海岸和海面目标类型,分别为船舶、鱼排网箱养殖区、筏式养殖区。通过筛选校正后,采用深度学习方法进行了严格的数据质量控制,能满足目前主流深度学习模型训练需要。本数据集可为近海岸海面目标图像的语义分割、水平目标检测、旋转目标检测等深度学习研究领域提供基础数据。

The intelligent detection of offshore maritime targets using computer vision technology can provide a scientific basis for marine administrative management,marine environmental supervision and management as well as the formulation of marine environmental protection policies,providing a powerful environmental information reference for the steady development of the economy.The dataset includes the data collected from Sanduao Harbor in the southeast of Ningde City,Fujian Province,China,with Google Earth serving as the primary data source and a time span from 2019 to 2021.This dataset comprises 1,761 visible light remote sensing images acquired under different seasons,backgrounds and illumination conditions,and corresponding horizontal object detection labels,rotational object detection labels and semantic segmentation labels,covering three types of offshore maritime targets,namely ships,fish row cage culture areas,and raft culture areas.After screening and correction,it can meet the current mainstream deep learning model training needs.This dataset can provide basic data for the semantic segmentation,horizontal object detection,rotational object detection and other research fields of offshore maritime target images.