IoT-Based Security With Facial Recognition Smart Lock System
Tri-Nhut Do, Cong-Lap Le, Minh-Son Nguyen
Abstract
In the era of technology development along with intelligent algorithms to make people’s lives more and more convenient, face recognition smart lock is a system that uses one of the mentioned algorithms and addresses a security aspect in smart home technologies. It can be placed at doors for monitoring home, workplaces, and campuses. The problem with existing face recognition smart lock systems is that they are not fast and accurate enough yet. In this paper, a smart lock system is designed by employing a deep learning algorithm (called Yolo) on Jetson nano board. The proposed approach is able to collect multiple images of new person in short time, automatically transfer data to server then using auto label for training and addresses several methods to increase speed and accuracy such as employing Yolo run in background, automatic adjusting brightness by built-in light in camera, asking people put their faces in appropriate distance range as hand’s length. Users can open or close the lock by face recognition and by installing the developed android application in devices like tablets, smartphones, laptops, etc. by providing the login credentials like username and password which is verified in the database over the internet as well. The system is trained on several people (members and strangers) with thousands of captured images for each person. It is also tested by many experiments with results of accuracy reaching to 99% and of speed to recognize in 0.6 second in some cases of good conditions of nice light and distance.