Skip to content Skip to sidebar Skip to footer

Short-range photogrammetry applied to botany

Machine learning-based digital innovations for phenotyping in the field

Cost-effective phenotyping methods are urgently needed to advance crop genetics to meet the food, fuel and fibre demands of the coming decades despite increasing environmental pressures and the great need to reduce agricultural inputs. During the last decade, unmanned aerial sensing (UAS) became a prominent tool in plant phenotyping. First, we considered the capability of ML (Machine Learning) to make grain yield predictions in soybean by combining data from different optical sensors on board UAS: multispectral images to analyse the spectral response and RGB images to reconstruct the study area in 3D, assessing the dynamic growth physiology. As a second approach, we introduced canopy roughness as a new trait that can be efficiently calculated from UAS image data as a biomass indicator.

Root phenotyping by X-ray imaging

Root system architecture (RSA) establishes the connection between plants and the soil environment, being the critical link for the extraction of water and nutrients from soils. Determining the contribution of various root structural features to crop yields is therefore vital to overcome climate change, environmental degradation and food insecurity. In this context, we developed a spatio-temporal model of root architecture from digital twins obtained by X-ray computed tomography. This new approach is optimised for high performance, repeatability and robustness.