Discrimination of Urochloa seed genotypes through image analysis: Morphological features

Abstract

The tropical forage species seed market, including signalgrass (Urochloa), is expanding. There is demand for fast, automated methods that can be implemented for quality control of the seeds produced, with the aim of evaluating physical purity. In this respect, seed morphological features obtained by processing radiographs with the Tomato Analyzer software and of red-green-blue obtained and processed on the Groundeye device were used to test differentiation of materials of the Urochloa genus. Seeds of Urochloa brizantha (A.Rich.) R.D.Webster cvv. BRS Piatã, Marandu, and Xaraés MG-5; Urochloa ruziziensis (R.Germ. & C.M.Evrard) Crins cv. Ruziziensis; and Urochloa decumbens (Stapf) R.D.Webster cv. Basilisk were evaluated. Morphological features obtained by Tomato Analyzer allowed differentiation of Urochloaseeds at an accuracy level greater than 80% for all the materials evaluated. Seed area was one of the features that allowed this differentiation. The Groundeye system also showed high efficiency in distinguishing Urochloa seeds, except for Piatã and Xaraés, which exhibited morphological similarity, and lower accuracy levels in distinguishing their seeds (<50%). Additional studies should be conducted to evaluate Urochloa seed lots from other origins to validate the use of image analysis techniques for the purpose of obtaining morphological features and to assist in determining the physical purity of the materials. INTRODUCTIONCorrect identification of species and even cultivars in the seed analysis laboratory is a time-consuming operation (Chaugule & Mali, 2016) because of the similarity among seeds of the same genus, or even as a result of the absence of specific morphological descriptors of the materials under Abbreviations: LDA, linear discriminant analysis; RGB, red-green-blue; WMIH, width at mid-height.