IDENTIFYING CORN (Zea mays L.) NUTRITIONAL STATUS by ARTIFICIAL VISION SYSTEM.

Corn nutritional status can be evaluated by visual symptoms and chemical analysis. The first depends of personal evaluator’s interpretation and on the second the sampling time occurs in advanced developmental stage and is not useful for deficiency correction in the same cycle. The artificial vision system (AVS) is a digital method able to identify and correct nutritional deficiencies in the same cycle. This study aimed to evaluate the AVS method to identify nutritional deficiencies in corn (Zea mays L.) submitted to levels of Ca, Mg and S. Three trials were carried out in greenhouse under hydroponic condition, using Hoagland and Arnon (1950) solutions for: a) Ca, b) Mg and c) S. In each experiment treatments were: 0, 33, 66 and 100% of nutrient supply) using 4 replicates. Leaf samples were collected (new leaves for Ca and S; old leaves for Mg and leaf 4) at the V4 stage and images obtained using a 9600 DPI scanner; the samples were also chemically analyzed. The SVA techniques evaluated were: Fractal Dimension Volumetric (FDV), Gabor Wavelet (GW) and Fractal Dimension Volumetric using canonical analysis (FDVCA). For each technique we analyzed the base, middle and tip of the leaves using gray and color images. Increasing concentration of nutrient solution increased the levels of Ca, Mg and S in the leaves, creating specific visual symptoms. The SVA was able to identify all levels of nutrient deficiency. In the new leaves, GW with color images accurately diagnosed Ca deficiency in the middle third (80% accuracy) and S at the base (78% accuracy) of the leaves. The FDVCA technique using color images accurately diagnosed Mg deficiency in the middle third of the leaf with 75.5% of accuracy. The AVS method was effective to identify levels of Ca, Mg and S in the corn leaves.