Resumen de Vision-Based Human Posture Detection from a Virtual Home-Care Unmanned Aerial Vehicle

Andrés Bustamante Londoño, Lidia María Belmonte, Antonio Pereira González, Pascual González Galindo, Antonio Fernández Caballero , Rafael Morales Ruiz

• Monitoring is essential to provide assistance to people who require home care due to their age or health condition. This paper presents the vision-based detection of three postures of a person (standing, sitting and laying down) from an unmanned aerial vehicle. The proposal uses the MediaPipe Pose Python module, considering only seven skeleton points and a set of trigonometric calculations. The work is evaluated in a Unity virtual reality (VR) environment that simulates the monitoring process of an assistant UAV. The images acquired by the UAV’s on-board camera are sent from the VR visualiser to the Python module via the Message Queue Telemetry Transport (MQTT) protocol. The simulation shows very promising results for the detection of a person’s postures.