Digital phenotyping of winter squash fruits

Abstract

Winter squash (Cucurbita moschata) has great importance as a food. Brazil has a wide genetic variability of squash; most of this is conserved in germplasm banks. The Vegetable Germplasm Bank of the Federal University of Viçosa (BGH-UFV) includes more than 350 accessions of squash; however, this germplasm is still little used. Characterization of accessions requires time, labor, and financial resources. Image-based, high-quality and large-scale phenotyping is a promising alternative tool. We propose digital phenotyping of C. moschata germplasm fruit. To achieve this, we evaluated 466 fruits from 148 accessions of squash from BGH-UFV and four checks. After longitudinal cutting, the fruits were evaluated on the basis of their length, diameter, and internal cavity dimensions. An image of every fruit was also obtained. Digital measurements were made using the software FENOM. The comparison between manual and digital forms of fruit evaluation was carried out with the software GENES. The comparisons were based on the analyses of simple linear regression, bias, the coefficient of Pearson correlation, the index of concordance, the index of performance, the efficiency of the method, the absolute average error, and the absolute maximum error. The evaluations based on images had high concordance (>0.93), almost perfect correlation (>0.99), and a performance classified as excellent (>0.92), in the evaluation of all the descriptors, when compared to manual measurements. We conclude that phenotyping of winter squash fruits based on digital images is promising for the characterization of C. moschata accessions, resulting in an efficient evaluation.