Real-Time Student Attendance via OpenCV Face Recognition and Blink Liveness Verification

Abstract

Accurate, low-cost face recognition is increasingly important for automated attendance, access control, and campus security systems. This paper presents a lightweight, real-time face recognition framework built entirely with the OpenCV library aimed at university and educational environments. The system acquires live images via standard webcams, applies image preprocessing (grayscale conversion and normalization), and detects face regions using Haar cascade classifiers. For identity encoding and matching, the framework evaluates classical approaches—local binary patterns histograms (LBPH), Eigenfaces (PCA), and Fisherfaces (LDA)—and implements a modular pipeline that allows easy swapping of encoders and classifiers. Recognized identities are logged automatically into an attendance database. The implementation puts strong emphasis on resource efficiency and reproducibility, allowing its deployment on commodity hardware without GPU acceleration. Prototype deployments under controlled classroom conditions demonstrate the feasibility of the approach and its robustness to modest lighting and pose variations. We discuss limitations, ethical considerations due to privacy and consent, and propose extensions including benchmark evaluation on public datasets, anti-spoofing integration, and lightweight deep-learning backbones for improved accuracy.