Author
![]() |
SCOTT, JASON - Tasmanian Institute Of Agricultural Research |
![]() |
Gent, David - Dave |
![]() |
HAY, FRANK - Tasmanian Institute Of Agricultural Research |
![]() |
PETHYBRIDGE, SARAH - Cornell University |
Submitted to: HortTechnology
Publication Type: Peer Reviewed Journal Publication Acceptance Date: 8/12/2015 Publication Date: 11/5/2015 Citation: Scott, J.B., Gent, D.H., Hay, F.S., Pethybridge, S.J. 2015. Estimation of pyrethrum flower number using digital imagery. HortTechnology. 25(5):617-624. Interpretive Summary: Estimating yield in pyrethrum is laborious and prone to measurement errors due to variability within plots. In this research we developed a simple approach to automate estimation of the number of flowers in a plot, with is a good proxy for yield. Digital images were processed and used to to count the number of objects with the correct color and shape for a pyrethrum flower disc floret. There was a good correlation between the image analysis and manual counts of flowers, although errors increased when the crop canopy was lodged or shaded. Only 8 images were needed to estimate the number of flowers in a standard sized plot. With this technology, automated image analysis can serve as a reasonable substitute for manual flower counts in pyrethrum, which will greatly enhance the efficiency of yield estimation. Technical Abstract: Flower number is a major determinant of pyrethrin yield in pyrethrum [Tanacetum cineariifolium (Trefi.) Sch. Bip.] production. Traditional estimates of flower numbers utilize physical harvesting of flowers which is time consuming and destructive. Physical harvest is complicated by constraints such as labor availability, and the precision of yield measurements may be highly influenced by spatial heterogeneity of plant density and vigor. Here we examine the potential for digital image analysis to enable rapid, non-destructive assessment of flower number to quantify treatment differences in agronomic trials. In brief, the technique involved estimating flower numbers using a combination of color thresh-holding to remove pixels with color profiles not typical of the disc florets of pyrethrum. Particle counting was then performed using defined size and shape parameters. Automated estimates were correlated with manual counts of flowers, representing approximately 32% of total flower numbers present within a sampling unit. This relationship was consistent across all flower densities observed. System accuracy was improved by censoring data sets for occurrences of crop lodging and over-mature flower canopies. Shadows also adversely affect estimates. Pyrethrum flowers were found to be spatially aggregated within fields, with the degree of aggregation greatest at the lowest flower densities. Based on modelled flower distributions, 8 sampling units (0.49 m2) were sufficient to achieve a coefficient of variation of 0.1 in a 600 m2 plot area for all fields except those with the lowest flower densities. Given the relative speed of image capture, oversampling of plot areas to ensure the threshold accuracy is achieved across all plant densities appears the most practical option for implementing this system. Automated image analysis can serve as a reasonable proxy for manual flower counts in pyrethrum, which will greatly enhance the efficiency of yield estimation. |