PhenoFrame: Phenotyping Plant Descriptor and Software Suite

Abstract Exchanging three-dimensional plant reconstructions between measurement and modeling workflows presents significant challenges. PhenoFrame is introduced as a shared descriptor and software suite that offers both a skeleton mode, utilizing observed leaf centerlines, and a procedural mode, in which each centerline is replaced by fitted parameters. A total of 324 sorghum plants and 3,317 repeated scans of 80 maize plants were converted into both modes. For sorghum, the median descriptor size was reduced from 9.6 to 4.5 kB per plant, with a median mode-to-mode Chamfer distance of 0.65 cm (0.46 cm for maize). For 9 of 13 sorghum traits, procedural-mode estimates reached a broad-sense heritability of at least 0.20 (10 of 13 for voxel-derived traits), as did all 13 maize traits. A descriptor-based stochastic generator approximated the marginal trait distributions, overstating the mean absolute between-node leaf-angle correlation (0.74 versus 0.25 in actual plants). Genome-wide association analysis of the descriptor azimuth traits identified several candidate genes associated with auxin signaling, which is expected to regulate phyllotaxy.