Development of an artificial vision algorithm to detect the Huanglonbing disease in the citrus lemon plant of the “Fundo Amada” (#1351)

Authors

Santos Gonzales, Cesar Enrique

Delgado Rodríguez, Jholy Jhanela

Blas Varas, Stefany Anabeth

Abstract

In Peru, between August and October 2023, lemon production faced a crisis due to exceptional weather conditions associated with the El Niño Phenomenon. This situation led to the proliferation of pests and diseases, generating a 500% increase in the price of lemon, and reducing its availability, which led to a decrease in its consumption. The Amada Farm was also affected by pests such as penicillium, exocortis and cottony mealybug. The Huanglonbing (HLB) disease, known as "yellow dragon", stood out as the most destructive and represented a serious threat to the farm. Currently, the control of this disease is carried out mainly through the use of chemicals, but there is growing interest in solutions based on biological control to reduce dependence on chemical substances. Given the difficulty of controlling the disease in susceptible varieties and in areas with years of presence of the pathogen, it was proposed to develop an artificial vision algorithm. This algorithm aims to detect the disease in lemon plants, identifying both diseased leaves and those that, at first glance, appear healthy, but already have the disease in its initial stage. To achieve this, convolutional neural networks and the Python programming language were used. 119 images of leaves were captured in various conditions, including diseased and healthy, with variations in lighting, position, and quantity. The purpose of this set of images was to evaluate the effectiveness of the algorithm using a confusion matrix. The results showed that, of the 119 samples taken, only 4 errors were recorded, confirming that the algorithm achieved an efficiency of 96.64%.