Sistema Inteligente de Verificação de Pedidos com Visão Computacional e Aprendizado de Máquina para Expedição Industrial 4.0

Abstract

This dissertation presents the development and evaluation of the PRS (Poligonal Reconnaissance System), designed for automated order verification in industrial environments within the context of Industry 4.0. The main objective was to develop an intelligent order verification system based on computer vision and machine learning, integrating image analysis, depth simulation, and hybrid validation (fixed rules and AI), with a focus on applications in automated shipment processes. The adopted methodology included the use of computer vision techniques to identify and calculate packaging areas in simulated images, depth simulation to correct perspective distortions, and a supervised Random Forest model trained with historical weight and area data. The resulting hybrid system demonstrated accuracy above 94% in detecting volume and weight inconsistencies, significantly reducing verification time and operational rework .