Implementación de un módulo domótico de seguridad residencial aplicando técnicas de visión artificial para la identificación de rostros.

Abstract

This graduate research aimed to implement a home automation module for residential security applying artificial vision techniques for face detection. With the use of a conventional video surveillance network obtains the images through the RTSP communication protocol from an IP camera for further processing using a microcomputer such as the Raspberry Pi 4, those images are analyzed with artificial vision tools allowing to select of facial features through OpenCV libraries and HAAR classifiers, this algorithm runs on Pi OS operating system based Linux own of developing board, which also allows the activation of outputs in case of violating the video surveillance network by executing a security protocol such as alarm and presence simulation. The module consists of a graphical interface developed in Node-red that allows its activation or deactivation and a panic button if required. It is also linked to the social network Telegram that, thanks to its policies of Privacy, receive the notification of the status of the module and a photograph of the face of the person who is not registered in the database, including commands to turn the module on or off directly from the application mentioned above. It is concluded that the camera should be located at the height of 240 cm and the maximum detection range at 160 cm to obtain satisfactory results. It is recommended to enter at least 20 images for each type of environment into the database, that is, for sunny, cloudy, night climates, etc; this will improve the robustness of the module.