A Real-Time Smart Door Access Control System Using Haar Cascade Classifier and Embedded Vision

Abstract

Traditional access control methods, such as mechanical keys and PIN-based systems, are ever more vulnerable to theft, copying, and unauthorized access, while high-fidelity biometric systems require intensive computing resources not suitable for low-power embedded systems. This research addresses the difficulty of achieving a real-time contactless face recognition system that compromises between security, performance, and expense in the resource-constrained embedded IoT setting. We show the design and test of an intelligent door access system using a Haar Cascade classifier on the ESP32-CAM module with cloud analytics for enhanced control, monitoring, and data logging. The system is developed to perform well on resource-constrained hardware while providing secure and remote access control. Experimental trials under varying light conditions, facial angles, and image resolutions show that the system detects faces with 97% accuracy, with a mean detection time of 151.25 milliseconds and CPU utilization of 40.5%. By adapting conventional machine learning for embedded vision, this project bridges the gap between high-fidelity biometric security and real-world IoT deployment in practice, offering an affordable and scalable solution for office and home access control in contemporary times.