Uas-based 3D point cloud phenotyping for crop canopy analysis in agronomic field trials

Overview

This solution enables crop scientists to implement field phenomics at developmentally-relevant scales using unmanned aerial system (UAS) RGB imagery processed into 3D point clouds. The approach allows researchers to describe multiple canopy features accurately across crop development stages without requiring high-performance computing infrastructure. The technology addresses gaps in field phenotyping implementation, modeling, and genetics of longitudinal traits, making advanced crop measurement accessible to breeding programs and agronomic research teams.

The system has been validated for predicting soybean above-ground biomass, height, lodging, and uniformity in breeding pipelines. By leveraging overlapping RGB imagery collected plot-by-plot, the solution delivers phenotypic data that supports quantitative genetic analyses and genomic prediction of traits over time. This capability is particularly valuable for plant breeders, seed companies, and crop protection researchers seeking to enhance data-driven decision-making in variety trials.

Technical specifications

Key features:

• Plot-by-plot 3D modeling from overlapping UAS RGB imagery without high-performance computing requirements
• Prediction of above-ground biomass with high accuracy (R-squared of 0.94 in validation trials)
• Quantitative genetic analysis support via random regression models for longitudinal trait assessment
• Genomic breeding value prediction with high accuracies and low biases across specific time intervals
• Detection of time-varying genetic markers associated with biomass and other canopy traits
• Applications for herbicide tolerance assessment, off-type detection, yield prediction in early generation materials, individual plant selection, and genomic prediction of longitudinal traits
• Compatibility with dense or bushy crop growth habits beyond soybeans

The platform processes aerial imagery to generate 3D point clouds that capture canopy structure throughout crop development from approximately 27 to 83 days after planting. The approach enables researchers to track trait expression dynamically rather than relying on single time-point measurements.

Technology readiness level

The technology has progressed through substantial validation. In a subset of the SoyNAM multi-environment trial, destructive sampling of above-ground biomass combined with UAS RGB imaging demonstrated biomass prediction accuracy of R-squared = 0.94. Narrow-sense heritability estimates ranged from 0.02 at 44 days after planting to 0.28 at 33 days after planting, indicating moderate genetic signal capture for longitudinal biomass traits.

The 2020 validation expanded ground-based observations to thousands of measurements for height and biomass across soybean field plots, supplemented by hundreds of UAS flights. Future validation plans include applying statistical and mathematical models to predict ground-based metrics from 3D point cloud data, expanding field experiments to additional crops with dense growth habits, and commercializing the underlying intellectual property for rapid deployment of automated field phenotyping from UAS platforms.