Electrical & Electronics
Attendo: An IoT-based Smart Attendance System Using Biometric Edge-to-Cloud Architecture and LBPH Textural Analysis
The administrative workflow for student attendance tracking in higher educational institutions has historically relied on manual roll calls and physical sign-in sheets, methods that are inherently inefficient, time-consuming, and prone to the pervasive security vulnerability of “proxy attendance” (buddy-punching). This project presents “Attendo,” an integrated IoT-enabled smart attendance system designed to automate the authentication process using contactless facial recognition. The research addresses the “Security Usability-Cost Trilemma” by synthesizing low-cost edge-computing hardware with enterprise-grade cloud-native software architectures. The hardware architecture centers on a custom-fabricated Printed Circuit Board (PCB) integrating a standard ESP32-CAM microcontroller, a Passive Infrared (PIR) sensor for energy-efficient presence detection, dual LED status indicators, and a dedicated push-button triggered biometric capture pipeline. By leveraging a hardware-interrupt driven state machine, the edge node minimizes baseline power consumption to approximately 2.5mA in deep sleep while ensuring deterministic acquisition of VGA-resolution (640×480) JPEG matrices via the onboard OV2640 CMOS sensor. The software ecosystem is engineered on a distributed PERN (PostgreSQL, Express, React, Node.js) stack, utilizing AWS S3 buckets for scalable and persistent image blob storage. To accommodate the resource constraints of edge environments, computationally intensive bio metric processing is offloaded to a dedicated Python-based vision microservice. This microservice employs Haar Cascade classifiers for precise facial localization and the Local Binary Pattern Histogram (LBPH) algorithm for texture based feature extraction. By projecting micro textural patterns into concatenated spatial histograms, identity verification is executed efficiently using a strict Chi-Square (χ2) distance metric against pre-computed vectors securely stored in the PostgreSQL database.