Uso de algoritmos basados en machine learning para la detección de enfermedades en plantas frutales de piso climático templado utilizando procesamiento de imágenes caso práctico manzano.

Abstract

The current research work aimed to detect the apple plant diseases through the Artificial Neural Networks app based on Machine Learning with a suitable processing environment. An uncrewed aerial vehicle (UAV) was used to extract the samples in the field under study, and several videos of the apple plants were captured. Later, the videos were processed with Matlab software, and the samples extracted from the videos were stored in a folder. The neural network training process was carried out using Matlab software. With the train Cascade Object Detector tool, the study area was selected in each image. Once this was done, the training was supported with three learning algorithms LBP, HOG, HAAR; each one has unique characteristics, but each one needs a specific processing time. Once the training of each of the apple tree diseases had been carried out, the respective tests were accepted to check the artificial neural network's validity. After the results were found, all the information obtained was uploaded to an IoT ThingSpeak platform, where each positive detection was marked for a better analysis. Besides, three windows were added to the IoT platform, each with the corresponding disease. A window with the study site's real-time location and two windows allow a comparison between each of the diseases under study. Additionally, a blog was created in which information on apple tree diseases was uploaded. The link of this blog is linked to the IoT platform.