High-resolution measurement of moisture filed at soil surface with interfered image processing method and machine learning techniques
文献类型：期刊论文
| 作者 | Liu, Guanshi3; Tian, Shengkui2,3; Wang, Qiong1,2; Wang, Huazhe3; Kong, Lingwei3 |
| 刊名 | JOURNAL OF HYDROLOGY
 |
| 出版日期 | 2025-05-01 |
| 卷号 | 652页码:18 |
| 关键词 | Soil moisture field Machine learning Image fusion features Soil surface Image size effects |
| ISSN号 | 0022-1694 |
| DOI | 10.1016/j.jhydrol.2024.132623 |
| 英文摘要 | The spatial and temporal distribution of soil surface water content (SSWC) is a crucial indicator in fields such as hydrology, geotechnical engineering, and agriculture. Current methods for testing water content often rely on destructive, small-scale measurements, making it challenging to capture high-resolution soil moisture fields (SMF) at larger scales. Despite some progress in using machine vision and machine learning (ML) to analyze soil properties has been made, challenges persist in extracting effective features from interfered soil images, developing robust relationships between soil images and SSWC, and achieving high-resolution SMF mapping. In this study, the response of soil images to SSWC and the effect of interference on feature extraction were explored using an innovative image acquisition device. To mitigate these interferences, a simple and effective color feature extraction method, multi-Gaussian fitting (MGF), was introduced. A multi-source matrix of image features including color, texture, and ratio features was fused to uncover the potential relationships between image features and SSWC. Subsequently, five optimized ML models were constructed for SWC prediction, incorporating image color features, texture features, and fusion features, respectively. Based on these findings, a nondestructive testing system for SSWC with cm-level accounting for image size effects was proposed, and complementary software was developed. Finally, the performance of the proposed SMF system was validated through two typical model experiments. The results indicate that surface interferences significantly affect soil color feature extraction. The anti-interference ability (AIA) of color features extracted using MGF improved by 30 %-50 % compared to traditional image segmentation algorithms, enhancing the robustness of the relationship between image features and SWC. The ML models developed using image fusion features, generally outperformed traditional models and those constructed using color and texture features separately (p < 0.01). Among these, the Gaussian Process Regression (GPR) model, coupled with principal component analysis of image fusion features, emerged as the most suitable SSWC prediction model due to its superior accuracy and robustness (RMSE < 0.90 %, R-2 > 0.98). Its accuracy has been improved by 2-10 times compared to traditional models and also by 3-15 times compared to previously published studies. The reconstructed SMF field using the proposed system exhibits a more realistic distribution and higher signal-to-noise ratio, enabling the accurate capture of critical spatial and temporal characteristics during capillary water migration and drying tests. However, the SMF system is currently limited to obtaining SWCC. This research provides a new method for non-contact and efficient monitoring of SSWC, potentially offering technical support for studies on seepage in porous media and water migration in unsaturated soils. |
| 资助项目 | National Natural Science Foundation of China[52179115] ; National Natural Science Foundation of China[42172298] ; National Natural Science Foundation of China[42002289] ; National Natural Science Foundation of China[41907231] ; Fundamental Research Funds for the Central Universities[22120230229] |
| WOS研究方向 | Engineering ; Geology ; Water Resources |
| 语种 | 英语 |
| WOS记录号 | WOS:001398461300001 |
| 出版者 | ELSEVIER |
| 源URL | [http://119.78.100.198/handle/2S6PX9GI/37174]  |
| 专题 | 中科院武汉岩土力学所 |
| 通讯作者 | Tian, Shengkui |
| 作者单位 | 1.Minist Educ, United Res Ctr Urban Environm & Sustainable Dev, Shanghai 200092, Peoples R China 2.Tongji Univ, Key Lab Geotech & Underground Engn, Minist Educ, Shanghai 200092, Peoples R China 3.Chinese Acad Sci, Inst Rock & Soil Mech, State Key Lab Geomech & Geotech Engn, Wuhan 430071, Peoples R China |
| 推荐引用方式 GB/T 7714 | Liu, Guanshi,Tian, Shengkui,Wang, Qiong,et al. High-resolution measurement of moisture filed at soil surface with interfered image processing method and machine learning techniques[J]. JOURNAL OF HYDROLOGY,2025,652:18. |
| APA | Liu, Guanshi,Tian, Shengkui,Wang, Qiong,Wang, Huazhe,&Kong, Lingwei.(2025).High-resolution measurement of moisture filed at soil surface with interfered image processing method and machine learning techniques.JOURNAL OF HYDROLOGY,652,18. |
| MLA | Liu, Guanshi,et al."High-resolution measurement of moisture filed at soil surface with interfered image processing method and machine learning techniques".JOURNAL OF HYDROLOGY 652(2025):18. |
入库方式： OAI收割
来源：武汉岩土力学研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。