Tracking People Using Face Recognition

Abstract

In recent years, face recognition has emerged as an active field within computer vision, driven by potential applications and the parallel development of algorithmic techniques alongside the increasing availability of cost-effective computers with sufficient computational power to support these algorithms. As a biometric information process, a face recognition system offers wider applicability and operational range compared to other biometric methods such as fingerprinting, iris scanning, or signature recognition. The system employs a blend of methodologies across two primary domains: face detection and recognition. Face detection is executed on real-time images without a specific application context in mind. Processes involved in the system encompass white balance correction, segmentation of skin-like regions, extraction of facial features, and the extraction of face images from a face candidate. The integration of a face classification method utilizing a Feedforward Neural Network is a key component of the system. The system's performance is evaluated using a database containing 10 individuals with 20 to 30 photos of each person.