ESTIMATIVA DE FITOTOXICIDADE POR GLYPHOSATE EM Digitaria insularis (L.) Fedde (DIGGIN) VIA PROCESSAMENTO DE IMAGENS

Abstract

Glyphosate, combined with the technology of plants resistant to its molecule, facilitated the management of weeds, but selected resistant biotypes. Among these species, sourgrass (Digitaria insularis) stands out, due to its wide geographical distribution in the Brazilian territory. Phytotoxicity, which indicates the degree of action of a chemical molecule in plant tissues, is usually estimated visually. The aim of the present study was to adjust models for the estimation of phytotoxicity in sourgrass by means of proximal digital images, in order to reduce subjectivity. The study was carried out at the premises of IF Goiano - Urutaí campus, where plants of sourgrass grown in pots received different doses of glyphosate (0%, 25%, 50%, 75%, 100%, 150%, 200% and 300% ) in which 100% equals 1,440 g ea ha-1, generating a wide gradient of phytotoxicity. Visual assessments and digital image capture in the visible spectrum (RGB) were performed at 7, 14, 21 and 28 days after application (DAA). The images were processed in software R, with plant / background segmentation by applying the Otsu method to the hue values, obtained after conversion from RGB to HSV. The variation in visual phytotoxicity was analyzed as a function of the median hue of plant pixels, adjusting regression models. The plants showed a significant increase in phytotoxicity, with a maximum of 77% and a minimum of 3% at 14 DAA. However, there was a recovery of plant tissues in all treatments at 21 and 28 DAA. Nevertheless, it is possible to model phytotoxicity using the hue median, with a degree of adjustment greater than 60%.