Crop growth analysis is used for the assessment of crop yield potential and stress tolerance. Capturing continuous plant growth has been a goal since the early 20th century; however, this requires a large number of replicates and multiple destructive measurements. The use of machine vision techniques holds promise as a fast, reliable, and non-destructive method to analyze crop growth based on surrogates for plant traits and growth parameters. We used machine vision to infer plant size along with destructive measurements at multiple time points to analyze growth parameters of spring wheat genotypes. We measured side-projected area by machine vision and RGB imaging. Three traits, i.e., biomass (BIO), leaf dry weight (LDW), and leaf area (LA), were measured using low-throughput techniques. However, RGB imaging was used to produce side projected area (SPA) as the high throughput trait. Significant effects of time point and genotype on BIO, LDW, LA, and SPA were observed. SPA was a robust predictor of leaf area, leaf dry weight, and biomass. Relative growth rate estimated using SPA was a robust predictor of the relative growth rate measured using biomass and leaf dry weight. Large numbers of entries can be assessed by this method for genetic mapping projects to produce a continuous growth curve with fewer replicates.
Keywords: digital growth analysis; machine vision; plant phenotyping; relative growth rate; wheat.