Intelligent electronic passworded locker with unique and personalized security barriers for home security

Facial recognition is important concept when it comes to identification and security. The traditional ways of securing homes like the use of locks and keys are inefficient. Developing a security system using artificial intelligence (AI) that will monitor the surroundings and act in case of emergencies is vital. This paper proposes a smart home security monitoring system that can make decisions based on facial recognition technology. It is implemented using Mediapipe for face detection and FaceNet model for facial feature extraction. The proposed face recognition model is 80.55% accurate. An android application is developed which allows user to interact with the system even from remote distances. A door opening mechanism is implemented with the help of ESP8266. With the aid of the feedback from the reed switch, an alarm sounds when someone tries to break into the residence. The developed system provides a whole new security approach by discarding the need for traditional methods of security. The accuracy of the system is 80.55%. 

Smart Home Security Monitoring System based on Face Recognition and Android Application

This study presents the design and implementation of an IoT-based class attendance system with face recognition for the high school. The system includes the Open-Computer-Vision (OpenCV) library, Python programming language, and a Raspberry Pi as the main processing unit. The system employs a combination of Haar-Cascades for face detection and Eigenfaces, Fisher faces, and Local Binary Pattern Histograms for face recognition. The methodology for the system is described in detail, including flowcharts for each stage of the system. The experiment results of the system are analyzed and presented, including plots and screenshots. We discussed the challenges encountered during the project, the system's potential applications, and future developments. The system was developed to automate the attendance-taking process, increase the accuracy and security of attendance records, and provide data to improve student performance and progress by recording class attendance in Google Sheets. 

Class Attendance System using Unimodal Face Recognition System based on Internet of Educational Things

The high-speed development of the technology behind facial recognition systems has enabled its usage in the application to record attendance without human interference. The advantage of the software lies in its security, authentication, and identification features. The paper proposes one such attendance system for educators. The five phases of the attendance system are database creation of students, face detection, liveliness detection, face recognition, and attendance marking. The database is created by the images of the students in class. One of its distinguishing features is the liveliness detection algorithm, which solves the proxy attendance problem. Liveliness detection is performed using blinking detection and CNN classification between real and spoof images. The accuracy obtained from the system is 93%. Further, face recognition is performed using FaceNet. A web-based application is developed to make attendance marking more convenient for educators.

Face Recognition for Attendance System using CNN based Liveliness Detection

A safe box with a security system protects personal and valuable items from being stolen, tampered or damaged. The most used security mechanism is using keys and cards, or the use of passwords, personal identification numbers (PINS), and fingerprints for personal identification. Face recognition is one of the most effective security confirmation techniques for a safe box management system. This project proposed a comprehensive face-authentication-based security system that includes software, hardware, and the Internet of Things (IoT). In the software system, the OpenCV built-in library and face recognition algorithm written in Python were implemented. A total of 150 images (i.e., 50 images per subject from 3 individuals) were registered in the database for training and testing the developed face recognition system. The control and working of the hardware are performed by a Raspberry Pi, to which a Pi camera, an infrared sensor, and a solenoid lock are connected. The IoT enables the communication between the Raspberry Pi and Telegram cloud, allowing real-time monitoring of the safe box and alerting the owner by capturing images of approaches. The developed system shows an accuracy ranging from 0.79 to 0.91, the precision range of 0.571–1, the recall range of 0.64–0.88, and the F1-score ranging from 0.6 to 0.81. 

A Safe Box Safety System with Facial Authentication for Multiple Users

Computer vision has become a highly evolving field in recent years, dealing with methods for obtaining, processing, examining, and understanding digital images. Human face recognition in computer vision has a vital role to play in security and surveillance, and the mechanisms for increasing the security levels are strengthening day by day. The existing human face recognition system has been enhanced by introducing an anti-spoofing mechanism which will help to stop a nefarious person to intentionally get around with the system. This article focuses on an approach to detect a human face using texture analysis which includes computing a Histogram of Gradients (HOG) over a region of the face and then uses Support Vector Machines (SVMs) to recognize a face. A blink detection mechanism used in this article ensures the liveliness of the person, making the system more reliable. A Raspberry Pi module is used in implementing the work involved in this paper and the programming is done in Python using libraries like OpenCV and NumPy. This model can achieve a maximum accuracy of 92.68% and achieves optimal results during the afternoon, taking a total of 9.89 seconds for face recognition and blink detection.

Romit Ganjoo

Papers

Anti-Spoofing Door Lock Using Face Recognition and Blink Detection