Reconhecimento e contagem de afídeos em tempo real usando ESP32-CAM

Abstract

This work proposes the development of a fast system for recognizing and counting aphids, pests that significantly harm agricultural productivity. Using an ESP32-CAM module, the project aims to demonstrate the ability of a simple solution to capture images and identify and count aphids in real time, for preliminary decision-making, through the AphidCV computer vision method and an artificial intelligence model of the YOLO architecture. Despite the hardware limitations, which made it impossible to process images and infer the model directly on the ESP32-CAM, the system was adapted to delegate these tasks to a personal computer, while the camera acts in the capture and transmission of images. The project showed the potential and limitations of low-cost devices, and how they can be scaled, to be used in insect monitoring applications.