Quantifying Early Vigour and Ground Cover using Digital Image Analysis

Abstract

Early vigour, or the fast leaf area development by an establishing crop is important in many breeding programs and is currently evaluated by subjective rating. The objective of this study was to test whether analysis of digital images can quantify ground cover in a fast and repeatable way in two trials where ground cover evaluations were important. Both in a greenhouse trial comparing the early vigour of different varieties of tall fescue (Festuca arundinacea Schreb.) as in a field trial comparing different varieties of rye (Secale cereale L.), Italian ryegrass (Lolium multiflorum L.) and lopsided oat (Avena strigosa Schreb.) for use as cover crops, we took pictures on regular intervals. In both trials, parameters that allowed accurate discrimination between pixels that represented bare soil and pixels that represented soil covered with plants were easily found. The Hue dimension of the Hue Saturation Brightness colour space was the parameter with the largest discriminating power between ground and plant covered pixels. Both in the field as in the greenhouse, there were significant differences in ground cover between the varieties. We found a good regression between ground cover and biomass production in the early growth stage of the cover crops in the field trial, until the vegetation reached a soil cover of ca. 50 % and a corresponding biomass of ca. 500 kg dry matter/ha (R2 = 88 %, p value = < 0.001). In later stages, correlation between ground cover and biomass was weak due to the presence of erect growing genotypes, with few tillers and a lower ground cover but with a good aboveground biomass production. We conclude that image analysis has a good potential to quantify early ground cover and early aboveground biomass production as it can work both fast and accurate in the field.