Human Detection for Search and Rescue Applications with UAVs and Mixed Reality Interfaces

Human Detection (HD)in natural disaster areas with the use of Unmanned Aerial Vehicles (UAV)or drones is an important approach to preserve human lives in Search and Rescue (SAR)operations, allowing to rescue injured people through previous and fast exploration from safer zones. Nevertheless, constraints on artificial vision and interactive interfaces remain a challenge, namely the ones related to Computer Vision (CV)techniques, since HD cameras drones need to transmit video streams in real-time to Ground Stations (GS). Additionally, GS interfaces of this type of applications are usually not easy to use for unexperienced pilots. In this work, we propose an architecture for SAR operations, integrating an interactive interface that can be used with commercial UAV and Head-Mounted Display (HMD)devices, which provide Mixed Reality (MR)for GS. Results on an application prototype show that a fine tunning process of a non-traditional object detection model with SAR images can improve human detection precision with real-time video streaming.