Publication Details

Abstract

Traditional attendance-tracking methods oral roll calls, paper sign-in sheets, and even fingerprint or RFID-based biometric terminals remain slow, disruptive to ongoing activity, and vulnerable to proxy marking. This paper presents the design and implementation of a low-cost, embedded, real-time attendance system built around a Raspberry Pi single-board computer, a USB camera module, and the OpenCV computer-vision library. Faces are detected within the incoming video stream using the Viola–Jones algorithm, which leverages Haar-like rectangular features computed efficiently via an integral-image representation, together with a cascade of classifiers trained through AdaBoost. Once a face has been isolated, identity is established using the Local Binary Pattern Histogram (LBPH) method, selected for its relative robustness to lighting variation and its modest computational demands on resource-limited hardware. The system operates across four stages: one-time face enrolment, offline model training, live recognition with automatic attendance logging, and scheduled e-mail delivery of attendance records to the relevant faculty member. This paper describes the system architecture, the mathematical foundations of the two core algorithms, the hardware components used, and the software workflow, and positions the design relative to comparable systems in existing literature. It further discusses the well-known sensitivity of appearance-based detectors to face orientation and lighting conditions, and outlines potential improvements including deep-learning-based facial embeddings and edge-accelerated inference for enhancing robustness in future versions.