Plant Disease Diagnosis in the Visible Spectrum

Lili Guadarrama,Carlos Paredes,Omar Mercado

doi:10.3390/app12042199

Abstract

A simple and robust methodology for plant disease diagnosis using images in the visible spectrum of plants, even in uncontrolled environments, is presented for possible use in mobile applications. This strategy is divided into two main parts: on the one hand, the segmentation of the plant, and on the other hand, the identification of color associated with diseases. Gaussian mixture models and probabilistic saliency segmentation are used to accurately segment the plant from the background of an image, and HSV thresholds are used in order to achieve the identification and quantification of the colors associated with the diseases. Proper identification of the colors associated with diseases of interest combined with adequate segmentation of the plant and the background produces a robust diagnosis in a wide range of scenarios.

Highlights

And reliable diagnoses are important for the adequate treatment of any of the diseases that are present in plants
It can be seen that the Gaussian mixture models (GMM) method is more efficient on images with a controlled background, while the probabilistic saliency (PS) method is more efficient on images with high background noise
A simple proposal for the problem generating a robust diagnosis of the presence of chlorosis, necrosis, and white spots was presented; in order to do so, the task was divided in the segmentation of the plant and subsequent quantification of colors

Summary

Introduction

And reliable diagnoses are important for the adequate treatment of any of the diseases that are present in plants. A correct diagnosis of diseases in plants has the potential to be very important for environmental conservation and for agricultural efficiency; both are aspects that are crucial for the general well-being of the population. FAO estimates that annually, up to 40% of global crop production is lost to pests. Plant diseases cost the global economy over $220 billion. Necrosis, and white spots are the diseases considered for the automatic diagnosis generation in this work, because they have a strong association with the coloration of the leaves of the plants when they have these diseases. A brief description of the mentioned diseases is presented

Methods

Results

Conclusion