Maoguang Chen
Research Subjects
Maoguang Chen has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
Qiuxiang Tang
4 shared publicationsTao Lin
4 shared publicationsCaixia Yin
2 shared publicationsDi Qiao
2 shared publicationsFengquan Wu
2 shared publicationsGuangyun Sun
2 shared publications
Latest Publications
- Open Access
- 21 August 2026
- Published by Elsevier in Computers and Electronics in Agriculture
- Vol. 255
From visible to yield-contributing opened bolls: An interpretable PAI-based correction framework for cotton boll estimation under canopy occlusion in UAV imagery
- 21 August 2026
- Published by Elsevier in Computers and Electronics in Agriculture
- Vol. 255
Open AccessFrom visible to yield-contributing opened bolls: An interpretable PAI-based correction framework for cotton boll estimation under canopy occlusion in UAV imagery
- Open Access
- 18 July 2026
- Published by Elsevier in Agricultural Water Management
- Vol. 333
Estimation of cotton plant moisture content using UAV multimodal data and machine learning
- 18 July 2026
- Published by Elsevier in Agricultural Water Management
- Vol. 333
Open AccessEstimation of cotton plant moisture content using UAV multimodal data and machine learning
- Open Access
- 7 July 2026
- Published by Elsevier in Smart Agricultural Technology
- Vol. 14
Drones Empower Intelligent Water Management: Depth dependency Estimation of Soil Moisture in Cotton Fields Based on Multimodal Fusion and Machine Learning
- 7 July 2026
- Published by Elsevier in Smart Agricultural Technology
- Vol. 14
Open AccessDrones Empower Intelligent Water Management: Depth dependency Estimation of Soil Moisture in Cotton Fields Based on Multimodal Fusion and Machine Learning
- Open Access
- 25 May 2026
- Published by Elsevier in SSRN Electronic Journal
Drones Empower Intelligent Water Management: Depth dependency Estimation of Soil Moisture in Cotton Fields Based on Multimodal Fusion and Machine Learning
Accurate and efficient field SMC monitoring is key to implementing precision irrigation and optimizing agricultural water management. The drone platform provides an economically effective means for large-scale measurement. However, existing research still lacks systematic solutions in dealing with high-dimensional feature redundancy, remote sensing signal attenuation with soil depth, and verifying the robustness of models under complex water and nitrogen management conditions. Therefore, this study conducted water nitrogen coupling experiments in cotton fields in Xinjiang, integrating unmanned aerial vehicle RGB, multispectral, and thermal infrared data. This study systematically compared the performance of three feature selection methods (PCC, RFECV, ReliefF) with four machine learning models (PLSR, KNNR, RFR, BPNN) to establish the optimal estimation strategy. The results showed that: (i) The ReliefF RFR combination performed the best, with validation set R2 of 0.67, 0.62, and 0.52 in the 0-20 cm, 20-40 cm, and 40-60 cm soil layers, respectively; (ii) Multi source sensor fusion can effectively improve model capability, but its optimal combination varies depending on the depth of the target soil layer. Among them, RGB+MS+TIR has the best estimation effect on the 0-20 cm surface soil, while MS+TIR is more suitable for SMC inversion in the 20-60 cm deep layer; (iii) Analysis based on SHAP reveals that multispectral vegetation index and thermal infrared features are the core variables driving model prediction; (iv) The model maintains stable performance under different nitrogen fertilizer treatments, demonstrating its strong adaptability to complex water and nitrogen management conditions. This study proposes a systematic framework that integrates multimodal data, hierarchical feature selection, and machine learning. It not only achieves high-precision estimation of cotton field SMC, but also provides a scalable methodological reference for solving related inversion problems in agricultural remote sensing. Through water nitrogen coupling verification, it lays a solid foundation for its practical application in smart irrigation.
- 25 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessDrones Empower Intelligent Water Management: Depth dependency Estimation of Soil Moisture in Cotton Fields Based on Multimodal Fusion and Machine Learning
Accurate and efficient field SMC monitoring is key to implementing precision irrigation and optimizing agricultural water management. The drone platform provides an economically effective means for large-scale measurement. However, existing research still lacks systematic solutions in dealing with high-dimensional feature redundancy, remote sensing signal attenuation with soil depth, and verifying the robustness of models under complex water and nitrogen management conditions. Therefore, this study conducted water nitrogen coupling experiments in cotton fields in Xinjiang, integrating unmanned aerial vehicle RGB, multispectral, and thermal infrared data. This study systematically compared the performance of three feature selection methods (PCC, RFECV, ReliefF) with four machine learning models (PLSR, KNNR, RFR, BPNN) to establish the optimal estimation strategy. The results showed that: (i) The ReliefF RFR combination performed the best, with validation set R2 of 0.67, 0.62, and 0.52 in the 0-20 cm, 20-40 cm, and 40-60 cm soil layers, respectively; (ii) Multi source sensor fusion can effectively improve model capability, but its optimal combination varies depending on the depth of the target soil layer. Among them, RGB+MS+TIR has the best estimation effect on the 0-20 cm surface soil, while MS+TIR is more suitable for SMC inversion in the 20-60 cm deep layer; (iii) Analysis based on SHAP reveals that multispectral vegetation index and thermal infrared features are the core variables driving model prediction; (iv) The model maintains stable performance under different nitrogen fertilizer treatments, demonstrating its strong adaptability to complex water and nitrogen management conditions. This study proposes a systematic framework that integrates multimodal data, hierarchical feature selection, and machine learning. It not only achieves high-precision estimation of cotton field SMC, but also provides a scalable methodological reference for solving related inversion problems in agricultural remote sensing. Through water nitrogen coupling verification, it lays a solid foundation for its practical application in smart irrigation.