Plant phenotyping involves collecting information on the physical characteristics of plants. The information collected assists breeders and biologists to improve desired traits in crops. It is crucial to understand the behavior of crop plants in controlled settings so that genetic differences can be observed. In this period of increasing energy demand, renewable and carbon-neutral energy sources have become the subject of more research. One crop that is a possible biomass-energy source is energy sorghum, which does not compete as food source and is efficient at accumulating biomass. The stalk-thickness and height of energy sorghum are the main phenotypic parameters of interest, because 70-80 percent of the biomass is stored in the stalk. Measuring the stalk of energy sorghum can enable estimation of biomass yield. However, a phenotyping system dedicated to high-throughput data collection in energy sorghum in a greenhouse has yet to be developed. The research presented herein details the design, construction and testing of a semi-automated phenotyping system for energy sorghum plants in a greenhouse. Image collection, processing and analysis are evaluated as a potential method for measuring plant stalk thickness. The system proved capable of collecting digital images of 288 energy sorghum plants - a representative number for the greenhouse in the study - in 10.5 hours. Images were collected with 75% overlap and were stitched together manually with the GIMP software package to obtain a complete image of an individual plant. K-means segmentation was used to separate plant matter from background in the images, and a stalk-measurement algorithm was developed. Results of these image analysis techniques provided an average of 16% error as compared to measurements obtained with a caliper. The results of this research suggest that this phenotyping method is viable, with high-throughput and mainly limited by image stitching.
