摘要
利用国产高分六号数据构建针对尾矿库内部要素分割的Rel-Copypaste数据增强方法,实现数据集扩充;基于深度学习语义分割SE-ResUNet网络,实现京津冀地区尾矿库内部要素轮廓提取.研究结果表明:该模型平均提取精度高达85.06%,分别高出无数据增强方法和Copypaste数据增强方法2.53%和1.77%,可实现京津冀地区尾矿库的面积统计、尾矿库的分类和头顶库的筛查等.
A Rel-Copypaste data augmentation method for the segmentation of internal elements of tailings ponds is constructed by using the domestic high-definition No.6 data to realize data set expansion.Based on deep learning semantic segmentation SE-ResUNet network to achieve contour extraction of internal elements of tailings pond in Beijing-Tianjin-Hebei region.The average extraction accuracy of the model is as high as 85.06%,which is 2.53%and 1.77%higher than the no data augmentation method and the Copypaste Data augmentation method,respectively.The model realizes the area statistics of tailings ponds in the Beijing-Tianjin-Hebei region,the classification of tailings ponds,and the screening of overhead ponds.
作者
王浩洋
张小咏
吴凯
陈正超
WANG Haoyang;ZHANG Xiaoyong;WU Kai;CHEN Zhengchao(Beijing Key Laboratory of High Dynamic Navigation Technology,Beijing Information Science Technology University,Beijing 100101,China;Inner Mongolia Big Data Center,Huhhot,Inner Mongolia 010098,China;Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100094,China)
出处
《福州大学学报（自然科学版）》
CAS
北大核心
2023年第4期525-531,共7页
Journal of Fuzhou University(Natural Science Edition)
基金
国家自然科学基金资助项目(42071407)
高分辨率对地观测系统重大专项资助项目(03-Y30F03-9001-20/22)。
关键词
尾矿库
京津冀地区
要素提取
深度学习
数据增强
语义分割
tailing pond
Beijing-Tianjin-Hebei region
element extraction
deep learning
data augmentation
semantic segmentation
