Pyrethrum yield estimation by digital image analysis

Pyrethrum is grown for the production of insecticidal pyrethrins. Production in Australia is constrained by a number of fungal pathogens, including Stagonosporopsis ligulicola and Sclerotinia spp. To accurately assess control strategies for these and other diseases, a practical method of yield assessment is required. Previously, yield assessment has relied upon destructive, manual harvesting, which is both costly and time consuming. As an alternative, yield estimation by digital image analysis was trialled at 29 sites in summer 2010, with replication. Images were taken at a height of 0.9 m above the crop canopy, encompassing a defined 0.7 × 0.7 m quadrat, placed level with flower height. Image capture occurred once over 50% ray florets were open within a crop. With the software package ImageJ, HSB color thresholding was used to isolate the yellow centres of individual flowers, from which the particle analyser function estimated the number of flowers per quadrat. Linear regression indicated a strong relationship between automated (AC) and manual (MC) flower counts, (AC = 21.1 + 0.97*MC; R2 = 0.95, P < 0.0001) with an intercept not significantly greater than zero. A significant correlation was also obtained between automated flower counts and hand harvested dry weight of flowers (DW) from each quadrat (AC = 109.1 + 3.45*DW; R2 = 0.488, P < 0.0001). These results highlight the potential for predicting yield using non-destructive image capture.