Enhancing maize LAI estimation accuracy using unmanned aerial vehicle remote sensing and deep learning techniques

Artificial Intelligence in Agriculture, Journal Year: 2025, Volume and Issue: unknown

Published: April 1, 2025

Language: Английский

---

Estimation Model of Corn Leaf Area Index Based on Improved CNN

Agriculture, Journal Year: 2025, Volume and Issue: 15(5), P. 481 - 481

Published: Feb. 24, 2025

In response to the issues of high complexity and low efficiency associated with current reliance on manual sampling instrumental measurement for obtaining maize leaf area index (LAI), this study constructed a image dataset comprising 624 images from three growth stages summer in Henan region, namely jointing stage, small trumpet large stage. Furthermore, LAI estimation model named LAINet, based an improved convolutional neural network (CNN), was proposed. carried out at these key stages. study, output structure ResNet architecture adapt regression tasks. The Triplet module introduced achieve feature fusion self-attention mechanisms, thereby enhancing accuracy estimation. adjusted enable integration growth-stage information, loss function accelerate convergence speed model. validated self-constructed dataset. results showed that incorporation attention improvement increased model’s R2 by 0.04, 0.15, 0.05, respectively. Among these, information led greatest improvement, increasing directly 0.54 0.69. model, achieved 0.81, which indicates it can effectively estimate maize. This provide technology support phenotypic monitoring field crops.

Language: Английский

---

Construction and Evaluation of a Cross-Regional and Cross-Year Monitoring Model for Millet Canopy Phenotype Based on UAV Multispectral Remote Sensing

Agronomy, Journal Year: 2025, Volume and Issue: 15(4), P. 789 - 789

Published: March 24, 2025

Accurate, high-throughput canopy phenotyping using UAV-based multispectral remote sensing is critically important for optimizing the management and breeding of foxtail millet in rainfed regions. This study integrated multi-temporal field measurements leaf water content, SPAD-derived chlorophyll, area index (LAI) with UAV imagery (red, green, red-edge, near-infrared bands) across two sites consecutive years (2023 2024) Shanxi Province, China. Various modeling approaches, including Random Forest, Gradient Boosting, regularized regressions (e.g., Ridge Lasso), were evaluated cross-regional cross-year extrapolation. The results showed that single-site achieved coefficients determination (R2) up to 0.95, mean relative errors 10–15% independent validations. When models transferred between sites, R2 generally remained 0.50 0.70, although SPAD estimates exhibited larger deviations under high-nitrogen conditions. Even severe drought 2024, predictions still attained values near 0.60. Among these methods, tree-based demonstrated a strong capability capturing nonlinear trait dynamics, whereas offered simplicity interpretability. Incorporating multi-site multi-year data further enhanced model robustness, increasing above 0.80 markedly reducing average prediction errors. These findings demonstrate rigorous radiometric calibration appropriate vegetation selection enable reliable diverse environments time frames. Thus, proposed approach provides technical support precision cultivar semi-arid production systems.

Language: Английский

---

Multi-Source Feature Fusion Network for LAI Estimation from UAV Multispectral Imagery

Agronomy, Journal Year: 2025, Volume and Issue: 15(4), P. 988 - 988

Published: April 20, 2025

The leaf area index (LAI) is a critical biophysical parameter that reflects crop growth conditions and the canopy photosynthetic potential, serving as cornerstone in precision agriculture dynamic monitoring. However, traditional LAI estimation methods relying on single-source remote sensing data often suffer from insufficient accuracy high-density vegetation scenarios, limiting their capacity to reflect variability comprehensively. To overcome these limitations, this study introduces an innovative multi-source feature fusion framework utilizing unmanned aerial vehicle (UAV) multispectral imagery for precise winter wheat. RGB datasets were collected across seven different stages (from regreening grain filling) 2024. Through extraction of color attributes, spatial structural information, eight representative indices (VIs), robust dataset was developed integrate diverse types. A convolutional neural network (CNN)-based backbone, paired with (MSF-FusionNet), designed effectively combine spectral information both imagery. experimental results revealed proposed method achieved superior performance compared models, R2 0.8745 RMSE 0.5461, improving by 36.67% 5.54% over VI respectively. Notably, enhanced during phases, such jointing stages. Compared machine learning techniques, exceeded XGBoost model, rising 4.51% dropping 12.24%. Furthermore, our facilitated creation distribution maps key stages, accurately depicting heterogeneity temporal dynamics field. These highlight efficacy potential integrating UAV deep wheat, offering significant insights evaluation agricultural management.

Language: Английский

---

Non-Destructive Monitoring of Peanut Leaf Area Index by Combing UAV Spectral and Textural Characteristics

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(12), P. 2182 - 2182

Published: June 16, 2024

The leaf area index (LAI) is a crucial metric for indicating crop development in the field, essential both research and practical implementation of precision agriculture. Unmanned aerial vehicles (UAVs) are widely used monitoring growth due to their rapid, repetitive capture ability cost-effectiveness. Therefore, we developed non-destructive method peanut LAI, combining UAV vegetation indices (VI) texture features (TF). Field experiments were conducted multispectral imagery crops. Based on these data, an optimal regression model was constructed estimate LAI. initial computation involves determining potential spectral textural characteristics. Subsequently, comprehensive correlation study between LAI using Pearson’s product component recursive feature elimination. Six models, including univariate linear regression, support vector ridge decision tree partial least squares random forest determine estimation. following results observed: (1) Vegetation exhibit greater with than (2) choice GLCM parameters impacts estimation accuracy. Generally, smaller moving window sizes higher grayscale quantization levels yield more accurate estimations. (3) SVR VI TF offers utmost precision, significantly improving accuracy (R2 = 0.867, RMSE 0.491). Combining enhances by 0.055 0.541 (TF), reducing 0.093 0.616 findings highlight significant improvement achieved integrating characteristics appropriate parameters. These insights offer valuable guidance growth.

Language: Английский

---

Enhanced Crop Leaf Area Index Estimation via Random Forest Regression: Bayesian Optimization and Feature Selection Approach

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(21), P. 3917 - 3917

Published: Oct. 22, 2024

The Leaf Area Index (LAI) is a crucial structural parameter linked to the photosynthetic capacity and biomass of crops. While integrating machine learning algorithms with spectral variables has improved LAI estimation over large areas, excessive input parameters can lead data redundancy reduced generalizability across different crop species. To address these challenges, we propose novel framework based on Bayesian-Optimized Random Forest Regression (Bayes-RFR) for enhanced estimation. This employs tree model-based feature selection method identify critical features, reducing improving model interpretability. A Gaussian process serves as prior optimize hyperparameters Regression. field experiments conducted two years maize wheat involved collecting LAI, hyperspectral, multispectral, RGB data. results indicate that outperformed traditional correlation analysis Recursive Feature Elimination (RFE). Bayes-RFR demonstrated superior validation accuracy compared standard Pso-optimized models, R2 values increasing by 27% hyperspectral data, 12% multispectral 47% These findings suggest proposed significantly enhances stability predictive capability various types, offering valuable insights precision agriculture monitoring.

Language: Английский

---

Enhancing LAI estimation using multispectral imagery and machine learning: A comparison between reflectance-based and vegetation indices-based approaches

Computers and Electronics in Agriculture, Journal Year: 2024, Volume and Issue: 230, P. 109790 - 109790

Published: Dec. 18, 2024

Language: Английский

---

Estimation of Water Interception of Winter Wheat Canopy Under Sprinkler Irrigation Using UAV Image Data

Water, Journal Year: 2024, Volume and Issue: 16(24), P. 3609 - 3609

Published: Dec. 15, 2024

Canopy water interception is a key parameter to study the hydrological cycle, utilization efficiency, and energy balance in terrestrial ecosystems. Especially sprinkler-irrigated farmlands, canopy further influences field distribution microclimate, then plant transpiration photosynthesis, finally crop yield productivity. To reduce damage increase measurement accuracy under traditional measurement, UAVs equipped with multispectral cameras were used extract situ information. Based on correlation coefficient (r), vegetative indices that are sensitive screened out develop models using linear regression (LR), random forest (RF), back propagation neural network (BPNN) methods, lastly these evaluated by root mean square error (RMSE) relative (MRE). Results show first closely related normalized difference vegetation index (R△NDVI) r of 0.76. The seven from high low R△NDVI, reflectance values blue band (Blue), near-infrared (Nir), three-band gradient (TGDVI), (DVI), red edge (NDRE), soil-adjusted (SAVI) chosen models. All developed based three (R△NDVI, Blue, NDRE), RF model, BPNN model performed well estimation (r: 0.53–0.76, RMSE: 0.18–0.27 mm, MRE: 21–27%) when less than 1.4 mm. methods underestimate 18–32% higher which could be due saturation NDVI leaf area 4.0. Because easy perform, method recommended for winter wheat. proposed R△NDVI can estimate other plants as canopy.

Language: Английский

---

Corn Yield Prediction Based on Dynamic Integrated Stacked Regression

Agriculture, Journal Year: 2024, Volume and Issue: 14(10), P. 1829 - 1829

Published: Oct. 17, 2024

This study focuses on the problem of corn yield prediction, and a novel prediction model based dynamic ensemble stacking regression algorithm is proposed. The aims to achieve more accurate in-depth exploration potential correlations in multisource multidimensional data. Data weather conditions, mechanization degree, maize Qiqihar City, Heilongjiang Province, from 1995 2022, are used. Important features determined extracted effectively by using principal component analysis indicator contribution assessment methods. Based combination an early stopping mechanism parameter grid search optimization, performance eight base models, including deep learning model, fine-tuned. theory heterogeneous learning, threshold established stack high-performing realizing employing averaging optimized weighting methods for prediction. results demonstrate that accuracy proposed significantly better as compared individual with mean squared error (MSE) being low 0.006, root (RMSE) 0.077, absolute (MAE) 0.061, high coefficient determination value 0.88. These findings not only validate effectiveness approach field but also highlight positive role data fusion enhancing models.

Language: Английский

---