期刊文献+

基于U-Net的珊瑚礁遥感影像自动分类 被引量:2

U-Net model for automatic classification of coral reef remote sensing images
下载PDF
导出
摘要 珊瑚礁遥感影像分类是珊瑚礁遥感监测的关键性基础技术,对珊瑚礁生态保护与制图应用起着重要的支撑作用。提出一种新的基于U-Net模型的珊瑚礁遥感影像自动分类方法,该方法使用小样本珊瑚礁影像训练即可得到分类精度较高的模型,克服了一般深度学习模型需要海量样本数据训练的缺陷。基于LandsatTM影像,对南海珊瑚礁进行遥感分类,其准确度潟湖坡为78%,向海坡为85%,珊瑚礁、海洋、陆地均大于95%,所有类别的边界轮廓指数大于92%。因此,这种自动分类方法比传统的珊瑚礁遥感影像分类方法精度更高,分类速度更快。 The massive amounts of remote sensing image data of coral reefs provide challenges to traditional remote sensing image data processing methods,especially for coral reef remote sensing image classification methods.In recent years,the deep learning technology that has emerged has been widely used in natural language processing and computer vision,especially convolutional neural networks,which plays an increasingly important role in image classification fields including remote sensing image classification.The article attempts to apply a convolutional neural network—U-Net to the automatic classification of coral reef remote sensing images.Based on LandsatTM images of Zhongjian Island,the classification precision of lagoon slope is 78%;seaward slope is 85%;coral reef and sea and land are no less than 95%.The boundary contour index of the all classifications are no less than 92%.Experiments show that it is feasible to use the U-Net model for automatic classification of coral reef remote sensing images,and it has greater advantages than traditional classification methods in terms of classification accuracy and automation.
作者 王桓 吴迪 左秀玲 王浩 WANG Huai;WU Di;ZUO XiuHng;WANG Hao(Department of Military Oceanography and Hydrography,Dalian Naval Academy,Dalian 116018,China;PLA Key Laboratory of Hydrography and Cartography Engineering,Dalian 116018,China;School of Marine Sciences,Guangxi University,Nanning 530004,China)
出处 《海洋测绘》 CSCD 北大核心 2023年第1期63-67,共5页 Hydrographic Surveying and Charting
基金 国家自然科学基金青年基金(41801341)。
关键词 珊瑚礁遥感 影像自动分类 U型卷积神经网络(U-Net) 深度学习 Landsat-8卫星 coral reef remote sensing automatic image classification U-Net CNN model deep learning Landsat-8 images
  • 相关文献

参考文献10

二级参考文献76

共引文献117

同被引文献39

引证文献2

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部