Use of Digital Images to Identify Zinc Levels in Maize.

The identification of the nutritional status of plants based on chemical analysis of leaf tissue requires an advanced phenological stage of plant development. The Artificial Vision System (AVS) is a technique to overcome such disadvantage by using a set of algorithms to interpret images of plant leaves, allowing the identification of nutrient deficiency at the beginning of plant development. The objective was to organize a data base of images of maize leaves (Zea mays L.) grown in nutrient solutions with Zinc (Zn) levels, and evaluate its use. The study was performed in the Universidade of São Paulo/Pirassununga-SP/Brazil. The maize hybrid (DKB 390) was grown in a greenhouse using a hydroponic system (Hoagland and Arnon, 1950). The treatments were: D1=0.0(0%); D2=0.034(20%); D3=0.068(40%); and D4=0.170(100%) mg.L^-1 Zn, and evaluated at three development stages: V4, V7 and R1, with four replications (Tukey 5%). The medium third of the index leaf (FI) – leaf ‘four’, ‘seven’ and ‘opposite and below the ear’, respectively, during the V4, V7 and R1 was scanned (1200 dpi), and submitted both to the AVS and chemical analysis. The application of Wavelets Gabor multispectral and color fractal (WGmfc), resulted in a global rights percent (GRP) and Kappa index (K). Visual symptoms of Zn deficiency were observed in the cases of small concentration of such element in solution. The Zn concentration was significant (p<0.01) along the increase in Zn in solution in V4, adjusting to a quadratic model (y=-2240.7x²+416.1x+46.2). The AVS identified the levels of Zn deficiency, resulting in a GRP of 48.9% in V4, 60.3% in V7 and 70.4% in R1, and K= 0.61 (reliable), 0.47 and 0.32 (moderately reliable), respectively. The WGmfc technique was efficient to identify Zn deficiency using IL in maize at stage V4, being reliable. However, for the V7 and R1 stages, it was moderately reliable.