Skip to main content
ARS Home » Pacific West Area » Wapato, Washington » Temperate Tree Fruit and Vegetable Research » Research » Publications at this Location » Publication #395421
Research Project: Developing New Potatoes with Improved Quality, Disease Resistance, and Nutritional Content

Location: Temperate Tree Fruit and Vegetable Research

Title: TubAR: an R package for quantifying tuber shape and skin traits from images

Author
item MILLER, MICHAEL - University Of Minnesota
item SCHMITZ-CARLEY, CARI - Aardevo
item FIGUEROA, RACHEL - University Of Minnesota
item Feldman, Max
item Haagenson, Darrin
item SHANNON, LAURA - University Of Minnesota

Submitted to: American Journal of Potato Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/21/2022
Publication Date: 12/14/2022
Citation: Miller, M.D., Schmitz-Carley, C.A., Figueroa, R.A., Feldman, M.J., Haagenson, D., Shannon, L.M. 2022. TubAR: an R package for quantifying tuber shape and skin traits from images. American Journal of Potato Research. 100:52-62. https://doi.org/10.1007/s12230-022-09894-z.
DOI: https://doi.org/10.1007/s12230-022-09894-z

Interpretive Summary: The physical appearance of potato tubers is a major determinant of consumer cultivar preference. Traditionally, potato breeders have relied upon subjective scoring methods performed by human graders to evaluate these qualities. Unfortunately, such quality assessments can vary due to bias between individuals and differences in the scoring methodology used between different breeding programs. Machine vision, the measurement of features from digital images, allows researchers to uniformly acquire precise measurements of tuber appearance characteristics at low cost. To evaluate the performance of machine vision to improve quality assessment of red-skinned potato cultivars, scientists at the USDA-ARS laboratories in Prosser, WA and Grand Forks, ND in collaboration with researchers at University of Minnesota and Aardevo LLC developed a software package named TubAR (Tuber Analysis in R) that measures multiple tuber quality traits simultaneously. Machine vision measurements of skinning, roundness, and length to width ratio as measured by TubAR consistently outperformed human measurement of these same quality traits. The software package TubAR provides a toolbox for researchers to efficiently collect tuber quality data for fresh market breeding programs by measuring multiple traits simultaneously using a single phenotyping protocol.

Technical Abstract: Potato market value is heavily affected by tuber quality traits such as shape, color, and skinning. Despite this, potato breeders often rely on subjective scales that fail to precisely define many phenotypes. Individual human raters and the environments in which ratings are taken can bias visual quality ratings. Collecting quality trait data using machine vision allows for precise measurements that will remain reliable between raters and breeding programs. Here we present TubAR (Tuber Analysis in R), an image analysis program designed to collect data for multiple tuber quality traits at a low cost to breeders. To assess the efficacy of TubAR in comparison to visual scales, red potatoes were evaluated using both methods. Broad sense heritability was consistently higher for skinning, roundness, and length to width ratio using TubAR. TubAR collects essential data for fresh market potato breeding programs while maintaining efficiency by measuring multiple traits through one phenotyping protocol.