Computer vision for assessment the seed coat color of carioca common beans

Abstract

Consumer acceptance of common beans (Phaseolus vulgaris L.) belonging to the Carioca commercial group depends on the color of the seed. Therefore, producers seek bean cultivars that have a light seed coat after storage. This trait is very important for common bean breeding programs dedicated to produce a high market demand. Therefore, the objective was to propose and assess the use of a computer vision-based methodology for assessing common bean color at harvest and after storage. A total of 70 carioca bean cultivars were visually assessed using a grading scale and computer vision after harvest and 90 days after the first assessment. The images allowed the cultivars to be discriminated according to the seed coat color. The accuracies with both assessment methodologies were >0.90. In addition, the correlations between these methodologies were ≤−0.72. The coefficients of variation for computer vision were lower than 6.50, while for the visual assessment, they were >10.08. Therefore, computer vision applied to assess the seed coat color of carioca bean grains is precise and accurate and allows for better discrimination than the visual assessment. Therefore, image analysis will assist in selecting better cultivars in breeding programs.