Determination of the severity of Septoria leaf spot in tomato by using digital images

Abstract

The aim of this study is to determine the severity of the disease Septoria leaf spot in tomato plants, through computational analysis of digital images of leaves affected. We collected and obtained digital images of tomato leaves with absence and presence of the disease with varying degrees of severity. From a script written in R with the EBImage package, the image was decomposed into three levels of color (RGB) and, through the process of thresholding the image segmentation, was performed separating sheet and injuries in relation to the background, determining the percentage of damaged area. Statistical properties were extracted from the original images and, from them and the severity quantified by software, was realized the process of correlation and regression analysis to indicate a template that determines the percentage of damaged area through the properties of the images. Subsequently, these models were tested, with a new image bank, from the RMSE error measures. The methodology described, was able to identify and quantify the damaged areas of the leaves with symptoms of diseases, extract the statistical properties of the images as allowed to predict mathematical models with acceptable potential and quality for indirect determination of the percentage of injured area through the properties of the images.