Location: Sugarbeet and Potato Research
Project Number: 3060-21430-008-04-S
Project Type: Non-Assistance Cooperative Agreement
Start Date: Jun 1, 2020
End Date: Jun 30, 2022
Develop phenotypic quantification photo-image analysis software for quantifying 1) fresh market storage quality attributes (bruising and skin color) and 2) processing fry quality (sugar end defects and fry color). A user friendly R package for quantifying appearance based phenotypes using image analysis will be developed and provided to stakeholders.
Images and phenotype reference values will be collected at harvest and throughout storage among 1) fresh market and 2) processing variety trials. Both ARS and Cooperator staff will collect images in a Photosimile 200 lightbox, equipped with a Canon Rebel T6i camera using a 24mm lens, ISO 100, 1/30 sec shutter speed and aperture f/5.6, and a blue background for uniform comparison. For fresh market storage quality software development: skin blemishes, color, and bruising (bruise incidence and severity) will be documented at the time of image collection. Samples will include fresh market clones grown in MN and from entries included in National Processing Trials hosted by USDA-ARS in East Grand Forks, MN. For development of fry quality analysis software, fry images will be collected immediately after frying potato planks (5/16 x 7/8'') for 3.5 minutes at 365°F. Potato plank photovolt % reflectance, Munsell color scores and % sugar ends will be recorded at the time of image collection. To obtain variability in fry processing quality, 40 advanced breeding clones representing six public breeding programs will be analyzed throughout 6 months of storage at contrasting storage temperatures (42 and 48°F). The cooperator will use these images to develop thresholds for assessing tuber quality (color, bruising), fry color, and sugar end defects. Image software development methodology relies on the EBImage package2 in R. Specifically, an RGB is converted to an Lab color scale and then a threshold is applied to create a binary filter. In order for the current series of scripts to be useful to a wider audience they will be converted into a R package for wider distribution. The package will be free, open source, and will be available through the Comprehensive R Archive Network.