UAV-Based LiDAR and Multispectral Imaging for Estimating Dry Bean Plant Height, Lodging and Seed Yield

Area of Science:

• Agricultural Science
• Remote Sensing
• Plant Breeding

Background:

• Climate variability impacts dry bean production, necessitating advanced breeding techniques.
• High-throughput phenotyping is crucial for developing climate-resilient dry bean cultivars.

Purpose of the Study:

• To evaluate Unmanned Aerial Vehicle (UAV)-based Light Detection and Ranging (LiDAR) and multispectral imaging (MSI) for dry bean phenotyping.
• To assess the accuracy of LiDAR and MSI in estimating plant height, lodging, and seed yield.
• To demonstrate the utility of integrated sensor data for accelerating dry bean breeding programs.

Main Methods:

• Collected LiDAR and MSI data across two dry bean field trials.
• Extracted LiDAR-derived features (canopy height, lodging, biomass) and MSI-derived indices (e.g., NDVI).
• Employed machine learning models (Gradient Boosting, Random Forest, Logistic Regression) for trait estimation and classification.

Main Results:

• LiDAR-derived canopy height strongly correlated with measured plant height (R² = 0.86) at the R6 stage.
• Lodging classification was most accurate using canopy height at the R8 stage.
• Integrated LiDAR and MSI models improved seed yield prediction (highest R² = 0.64) compared to individual datasets, with NDVI being a key spectral feature.

Conclusions:

• UAV-based LiDAR and MSI provide accurate, non-destructive methods for phenotyping dry bean traits.
• Integrating data from both sensors enhances the prediction of critical agronomic traits.
• This approach supports efficient cultivar development for improved dry bean production.