Development of Facial Recognition System for Attendance Monitoring and Cashless Transactions

Abstract

The purpose of this research is to produce an integrated model of a smart attendance system and cashless payment system for advanced processing, serving access control, intelligence security, and customer service. Manual attendance tracking is time-consuming and lacks mechanisms for compound monitoring, integration, or physical verification in large, spread-out environments. This issue is addressed through automated surveillance extending to facial recognition cashless payment (FRCP). The materials include a hardware setup comprising a Raspberry Pi 4B, a Raspberry Pi NOIR camera, a touch screen LCD, and Nvidia GPU-enabled host system. Software frameworks and libraries such as OpenCV, NumPy, Scikit-Learn, Keras, TensorFlow, Flask, Flutter, and MySQL were utilized. Using an Agile approach, the methodology combines Support Vector Machines (SVM) and Local Binary Pattern Histogram (LBPH) algorithms for face detection (via Histogram of Oriented Gradients) and classification across a self-made dataset of 48 users and 341 training images. System functions incorporate anti-spoofing, 128-D face embedding extraction, liveness detection, and two-factor authentication. Results demonstrate that the proposed model achieved an accuracy between 82% and 98.8% for identical real-time images and 77% to 88% for unidentical/spoofed images. SVM exhibited superior stability (near 100% frame recognition) compared to LBPH, though performance dropped to 68%–70% accuracy under heavy facial occlusion (e.g., masks and sunglasses). Future research and development should focus on enhancing System Requirements Specifications (SRS), resolving payment gateway transaction speeds, improving masked-face recognition accuracy, and expanding stakeholder-driven deployments across education, retail, and healthcare sectors.