DETECÇÃO E QUANTIFICAÇÃO DE FERRUGEM ASIÁTICA (Phakopsora pachyrhizi) NA CULTURA DA SOJA POR IMAGENS MULTIESPECTRAIS

Abstract

Brazil is one of the largest agricultural producers in the world, especially soybean. Currently, Asian rust (Phakopsora pachyrhizi) is the most severe soybean disease, which can cause significant losses and is considered by many to be the greatest current challenge for national soybean. Correctly and efficiently identifying and quantifying the severity of this disease is essential for choosing the appropriate anagement strategy. In search of more updated information and more efficient tools, this work aimed to apply current proximal and remote sensing tools for the detection and quantification of rust in soybeans. For this, images taken in the visible and near infrared spectrum were processed, obtained with digital cameras coupled or not, to an unmanned aerial vehicle (UAV). An experiment was carried out at the FMC Experimental Field in Primavera do Leste - Mato Grosso, in the 19/20 harvest. A randomized block design with four replications, using the following treatments: T1 - Azoxystrobin + Cyproconazol; T2 - Picoxystrobin + Ciproconazole; T3 - Protioconazole + Trifloxystrobin; T4 - Azoxystrobin + Benzovindiflupir; T5 - Picoxystrobin + Benzovindiflupir; T6 - Untreated Check. Three sequential leaf applications were performed at 14-day intervals, the first of which was carried out when the soybean crop was at the R2 stage (full flowering). The treatments were chosen with the objective of simulating different levels of severity in the field, presenting a gradient of control levels, from low to greater efficiency. Visual notes of rust symptoms (% of affected leaf area) were taken in the three strata of the plant - low, medium and upper, using a diagrammatic scale. Images of lateral leaflets of the plots and aerial images were processed in order to determine vegetation indexes based on near infrared and specific injury indexes for rust, based on the “Red, Green, Blue” (RGB) color system. The indexes adopted were: LCI (Lesion Color Index), RSI (Rust Severity Index) and RIA Lesion Color Index). The mean values of these indices were correlated with data on visual severity, area under the disease progress curve (AUDPC), percentage of defoliation and productivity. The R software as used in image processing and for statistical analysis. An R-language application was developed to utomate evaluations. The results obtained with this work confirm that the vegetation indexes showed a high correlation with the visual evaluation, both in RGB and RGN (Red, Green, NIR), and this one was able to detect minor variations, thus demonstrating that the proximal and remote sensing for the detection and quantification of rust in soybeans is feasible.