Monitoramento de Pedestres nas Dependências da UFAM Utilizando Reconhecimento Facial Visando a Segurança da Comunidade Acadêmica

Resumo

Resumo

At the Federal University of Amazonas (UFAM), security is an increasing challenge, with the frequent occurrence of robberies highlighting the need for efficient technologies to protect the academic community. In this context, this scientific initiation research presents the development and implementation of an intelligent monitoring system based on facial recognition, utilizing Computer Vision (CV) and Artificial Intelligence (AI) techniques. Initially, the study conducts a comparative investigation of the Convolutional Neural Network (CNN), Haar Cascade (HC), and You Only Live Once (YOLO) methods, determining the fastest and most accurate approach for real-time identification. Following the selection of this method, the research is further refined through model variations and preprocessing actions to identify the best classifier. The proposed system integrates a web application for continuous monitoring, identification of suspicious individuals, and automatic recording of movements on campus. The results demonstrate that the YOLOv11x method, using color images with a resolution of 200x200 pixels, achieves the highest reliability and efficiency for the application. Tests conducted in a real environment validate the robustness of the solution, highlighting its feasibility for risk mitigation and security enhancement in the academic setting.

palavras-chave

Palavras-chave

pedestrian monitoringface recognitioncampus safetyuniversity securityYOLOv11convolutional neural networksHaar cascadereal-time object detectionvideo surveillancemotion recordingFederal University of Amazonas

links

Links e arquivos