Plant Root Phenotyping with Computer Vision and Robotics
Students: Fedor Chursin, Borislav Nachev, Wesley van Gaalen, Francisco Mansilha, Vlad Matache
Overview
Our students set out to teach a liquid-handling robot the finesse needed for delicate root inoculations. Starting with a single greyscale photograph of five Arabidopsis seedlings in a Petri dish, they built an end-to-end pipeline that pinpoints any chosen spot on each root—such as the primary root tip—and converts that pixel location into precise XYZ commands for the robot. A custom-trained U-Net first segments every root, after which graph-based root-system-architecture analysis traces the full network of primary and lateral roots.
Real images, however, are messy: segmentation masks often contain gaps because the U-Net may miss very thin root sections or because artefacts (a stray leaf, a speck of contamination) temporarily hide part of the root. To restore continuity, the team trained a second neural network that “inpaints” these missing pixels. They tackled overlaps separately—mature neighbouring plants frequently intertwine—by designing a graph algorithm that untangles crossings and assigns each root segment to the correct plant. With these challenges solved, the software converts pixel coordinates into millimetre-accurate robot movements and automatically generates the liquid-dispensing sequence, delivering a novel workflow that fuses deep learning, graph theory and robotics.
Their tireless work ethic and problem-solving mindset transformed cutting-edge AI into a practical tool now empowering plant scientists; bravo to Fedya, Borislav, Francisco, Wesley, and Vlad!
About NPEC
NPEC is a joint initiative of Wageningen University & Research and Utrecht University. This integrated, national research facility is housed by Wageningen University & Research and Utrecht University and is co-funded by The Netherlands Organisation for Scientific Research (NWO).
NPEC facilitates state-of-the-art measurement of plant phenotypes to support research on genotype-phenotype associations. Establishing these associations is critical for the development of novel climate-proof crops and cropping systems. These novel crops and systems are necessary to secure our future high-quality food production, and improve the ecological sustainability of food production.