| Soil salinity inversion in cotton fields based on chroma\|adaptive color correction |
|
View Fulltext View/Add Comment Download reader |
| |
| DOI：10.7606/j.issn.1000-7601.2026.02.25 |
| Key Words: cotton leaf imaging color correction machine learning linear fitting regression soil salinity inversion |
| Author Name | Affiliation | | GAO Bo | School of Water Conservancy and Hydroelectric Power, Hebei University of Engineering, Handan, Hebei 056038, China | | WANG Lishu | School of Water Conservancy and Hydroelectric Power, Hebei University of Engineering, Handan, Hebei 056038, China | | JIA Yanhui | Weifang University of Science and Technology, Key Laboratory of Facility Horticulture in Universities of Shandong Province, Shouguang, Shandong 262700, China | | CHEN Yingchao | Shandong Shengda Water\|Saving Technology Co., Ltd., Shouguang, Shandong 262700, China | | HE Shuai | Institute of Farmland Irrigation, Soil and Fertilizer, Xinjiang Academy of Agricultural and Reclamation Sciences, Shihezi, Xinjiang 832000, China |
|
| Hits: 358 |
| Download times: 12 |
| Abstract: |
| To rapidly and effectively invert soil salinity status in cotton fields using “RGB imaging technology”, an improved chromaticity\|adaptive color correction algorithm was proposed. This method utilized the extracted R， G， B values from corrected cotton leaf images to investigate the correlation patterns between phenotypic R， G， B values of cotton leaves and soil salinity gradients. The results showed that the cotton leaf images corrected by the chromaticity\|adaptive algorithm exhibit a 22.26% reduction in color difference compared to those processed by traditional correction matrix algorithms, achieving effective R， G， B value modification. In the fitting regression between R， G， B tristimulus values and Electrical Conductivity (EC) values, the Random Forest (RF) model (R2=0.869) outperformed the Multiple Linear Regression (MLR) model (R2=0.633) by more comprehensively explaining the nonlinear complex interactions between R， G， B channels and EC. The RF model showed enhanced performance across key validation metrics, R2, RMSE, and MAE, yielding improvements of 37.28%, 9.06%, and 11.52%, respectively, in optimal prediction accuracy. This confirms the explanatory power of crop phenotypic information for in\|field soil salinity assessment. Compared with conventional algorithms, the proposed method achieved higher\|precision phenotypic characterization of cotton leaves. Its enhanced image characterization capability effectively restores the true phenotypic parameters of leaves under natural physiological conditions, further validating a significant synergistic response between plant phenotypic data (based on R， G， B image color features) and soil salinity levels. |
|
|
|