Performance comparison of deep learning models to extract silt storage dams in remote sensing images to prevent water loss and soil erosion

IF 1.1 4区 工程技术 Q3 ENGINEERING, CIVIL
Jingwei Hou, Moyan Zhu, Bo Hou
{"title":"Performance comparison of deep learning models to extract silt storage dams in remote sensing images to prevent water loss and soil erosion","authors":"Jingwei Hou, Moyan Zhu, Bo Hou","doi":"10.1680/jwama.22.00094","DOIUrl":null,"url":null,"abstract":"Determining the locations and shapes of silt storage dams (SSDs) is necessary before planning and constructing new ones or maintaining old ones. Google images with a spatial resolution of 0.54 m were cropped, labelled and enhanced to establish two schemes of remote sensing images that contain SSDs with different input and batch sizes. Five deep learning models (FCN (fully connected convolutional neural network, SegNet (deep convolutional encoder–decoder architecture for image segmentation), U-Net (convolutional networks for biomedical image segmentation), PSPNet (pyramid scene parsing network) and DeepLab-V3+) were constructed to extract SSDs from the images based on the two schemes. The loss curves, accuracies and extraction results derived from the five models were compared to identify the optimal model for SSD extraction. The results show that the overall accuracies, F 1 scores and mean intersections over unions obtained from DeepLab-V3+ were, respectively, 95.29%, 70.33% and 74.13% for scheme 1 (S1) and 96.29%, 73.34% and 76.99% for scheme 2 (S2), which were better than the values for other models. PSPNet had the shortest training times (128 s/step for S1 and 348 s/step for S2). An input size of 480 × 480 pixels, a batch size of 4 and 2304 images enhanced the extraction accuracy and prevented overfitting. The results provide a reference for the planning, construction and maintenance of water and soil conservation measures.","PeriodicalId":54569,"journal":{"name":"Proceedings of the Institution of Civil Engineers-Water Management","volume":"73 21","pages":"0"},"PeriodicalIF":1.1000,"publicationDate":"2023-11-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the Institution of Civil Engineers-Water Management","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1680/jwama.22.00094","RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"ENGINEERING, CIVIL","Score":null,"Total":0}
引用次数: 0

Abstract

Determining the locations and shapes of silt storage dams (SSDs) is necessary before planning and constructing new ones or maintaining old ones. Google images with a spatial resolution of 0.54 m were cropped, labelled and enhanced to establish two schemes of remote sensing images that contain SSDs with different input and batch sizes. Five deep learning models (FCN (fully connected convolutional neural network, SegNet (deep convolutional encoder–decoder architecture for image segmentation), U-Net (convolutional networks for biomedical image segmentation), PSPNet (pyramid scene parsing network) and DeepLab-V3+) were constructed to extract SSDs from the images based on the two schemes. The loss curves, accuracies and extraction results derived from the five models were compared to identify the optimal model for SSD extraction. The results show that the overall accuracies, F 1 scores and mean intersections over unions obtained from DeepLab-V3+ were, respectively, 95.29%, 70.33% and 74.13% for scheme 1 (S1) and 96.29%, 73.34% and 76.99% for scheme 2 (S2), which were better than the values for other models. PSPNet had the shortest training times (128 s/step for S1 and 348 s/step for S2). An input size of 480 × 480 pixels, a batch size of 4 and 2304 images enhanced the extraction accuracy and prevented overfitting. The results provide a reference for the planning, construction and maintenance of water and soil conservation measures.
深度学习模型在遥感影像中提取泥沙坝防止水土流失的性能比较
在规划和建造新水坝或维修旧水坝之前,必须先确定泥沙贮存坝的位置和形状。对空间分辨率为0.54 m的Google图像进行裁剪、标记和增强,建立了包含不同输入和批量大小ssd的两种遥感图像方案。基于这两种方案构建了FCN(全连接卷积神经网络)、SegNet(用于图像分割的深度卷积编码器-解码器架构)、U-Net(用于生物医学图像分割的卷积网络)、PSPNet(金字塔场景解析网络)和DeepLab-V3+ 5个深度学习模型,从图像中提取ssd。比较了5种模型的损失曲线、准确度和提取结果,确定了SSD提取的最佳模型。结果表明,方案1 (S1)和方案2 (S2)的总体精度、f1分数和平均相交次数分别为95.29%、70.33%和74.13%和96.29%、73.34%和76.99%,优于其他模型。PSPNet的训练时间最短(S1为128 s/步,S2为348 s/步)。输入大小为480 × 480像素,批处理大小为4和2304张图像,提高了提取精度并防止了过拟合。研究结果可为水土保持措施的规划、建设和维护提供参考。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
CiteScore
2.10
自引率
0.00%
发文量
28
审稿时长
6-12 weeks
期刊介绍: Water Management publishes papers on all aspects of water treatment, water supply, river, wetland and catchment management, inland waterways and urban regeneration. Topics covered: applied fluid dynamics and water (including supply, treatment and sewerage) and river engineering; together with the increasingly important fields of wetland and catchment management, groundwater and contaminated land, waterfront development and urban regeneration. The scope also covers hydroinformatics tools, risk and uncertainty methods, as well as environmental, social and economic issues relating to sustainable development.
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信