Clasificación de manzanas mediante visión por computador e inteligencia artificial para detectar anomalías

Abstract

This work presents an automated system for apple classification using computer vision and artificial intelligence, aiming to improve the efficiency and accuracy of fruit selection processes in the agricultural sector. To achieve this, two YOLOv11 neural network models were trained: one for general apple detection and another for identifying anomalies such as open wounds and rot. The system architecture integrates an ESP32 microcontroller that communicates via WiFi with a server implemented in Python using the Flask framework. This server is responsible for executing real-time detection models and coordinating the actions of the physical mechanism. The mechanical design includes a sectioned conveyor belt driven by a stepper motor, allowing the sequential entry of apples into the system. Additionally, a leadscrew mechanism enables rotation of the axis and displacement of the fruit for a complete analysis, while a servomotor directs each apple to a specific container based on its classification. This system is expected to contribute to the automation of agricultural tasks using accessible and open-source technologies, promoting their adoption in small and medium-sized enterprises in the sector.