Automatic morphological trait characterization for corn plants via 3D holographic reconstruction
Plant breeding is an extremely important route to genetic improvements that can increase yield and plant adaptability. Genetic improvement requires careful measurement of plant phenotypes or plant trait characteristics, but phenotype measurement is a tedious and error-prone task for humans to perform. High-throughput phenotyping aims to eliminate the problems of manual phenotype measurement. In this paper, we propose and demonstrate the efficacy of an automatic corn plant phenotyping system based on 3D holographic reconstruction. Point cloud image data were acquired from a time-of-flight 3D camera, which was integrated with a plant rotating table to form a screening station. Our method has five main steps: point cloud data filtering and merging, stem segmentation, leaf segmentation, phenotypic data extraction, and 3D holographic visualization. In an experimental study with five corn plants at their early growth stage (V3), we obtained promising results with accurate 3D holographic reconstruction. The average measurement error rate for stem major axis, stem minor axis, stem height, leaf area, leaf length and leaf angle were at 7.92%, 15.20%, 7.45%, 21.89%, 10.25% and 11.09%, respectively. The most challenging trait to measure was leaf area due to partial occlusions and rolling of some leaves. In future work, we plan to extend and evaluate the usability of the system in an industrial plant breeding setting.
This is a manuscript of an article published as Chaivivatrakul, Supawadee, Lie Tang, Matthew N. Dailey, and Akash D. Nakarmi. "Automatic morphological trait characterization for corn plants via 3D holographic reconstruction." Computers and Electronics in Agriculture 109 (2014): 109-123. DOI: 10.1016/j.compag.2014.09.005. Posted with permission.