Reconocimiento facial mediante visión artificial para el control de acceso a áreas restringidas

Abstract

Abstract: This project focuses on facial recognition through computer vision. Computer vision enables the processing and analysis of images in real time. According to González and Woods (2018), image recognition allows the extraction of relevant features to classify or identify objects within an image, making it an essential tool for multiple applications, including automated security systems. One of the main objectives of this project is to develop a facial recognition model using computer vision. The system is designed to manage access control to restricted areas, authorizing or denying entry, while simultaneously recording each access in real time. For the system development, facial recognition algorithms were implemented using the InsightFace framework. Specifically, the Buffalo_L package was employed, which integrates the RetinaFace detector for facial detection and alignment, and the ArcFace model for feature extraction. Classification was performed using the K-Nearest Neighbors (KNN) algorithm, complemented by an HMI interface to display recognition results. Testing was conducted with the VGGFace2 dataset, along with an additional set of images collected in a workplace environment. The results demonstrated outstanding performance, achieving 99% accuracy, precision, and recall, with a response time of less than 500 ms. These findings confirm the feasibility of the system for security and access control applications.