A smart door prototype with a face recognition capability

Abstract

This research aimed to integrate a face recognition capability in a smart door prototype. By using a camera-based face recognition, the house owner does not need to make physical contact to open the door. Avoid physical contact is important due to the coronavirus disease 2019 (COVID19) pandemic. Raspberry Pi 3B was used as the main controller, while a servo motor was utilized as a locking door actuator. The program was developed using Node-RED, Blynk, and message queue telemetry transport (MQTT) platforms which are very powerful for developing internet of things (IoT) devices. All of the programs were coded using Python. Haar cascade and local binary pattern histogram methods were implemented on the face recognition stage. Google Assistant integration was done by using Dialogflow and Firebase as Google Cloud services. Integration of face recognition and the smart door was successful. The smart door was unlocked if faces were recognized (average threshold=60%). If a face was not recognized, an email notification containing a face image is sent to the house owner. The Google Assistant could handle user requests successfully with a success rate of 92.8% from 147 trials.