Title: Real-time human search and monitoring system using unmanned aerial vehicle
Authors: Cheok Jun Hong; Vimal Rau Aparow; Hishamuddin Jamaluddin
Addresses: Automated Vehicle Engineering System (AVES) Research Group, Department of Electrical and Electronics Engineering, Faculty of Science and Engineering, University of Nottingham Malaysia, Jalan Broga, Semenyih, Selangor, Malaysia ' Automated Vehicle Engineering System (AVES) Research Group, Department of Electrical and Electronics Engineering, Faculty of Science and Engineering, University of Nottingham Malaysia, Jalan Broga, Semenyih, Selangor, Malaysia ' Department of Electrical and Electronics Engineering, Faculty of Engineering and Information Technology, Southern University College, Skudai, Johor, Malaysia
Abstract: Search and rescue operations after a major disaster are often hindered by inaccessibility to the affected area due to damaged infrastructure. However, this approach can lead to oversight due to the altitude and limited field of view. Therefore, a real-time human detection system using a fixed-wing quadcopter UAV equipped with Internet of Things devices for Mobile-Edge Computing is proposed. An onboard Raspberry Pi companion computer controls the UAV and streams aerial imagery from the camera attached to offload human detection computation tasks to a computing server at the edge using 4G cellular network technology. The computing server utilises YOLOv3 deep neural network trained on the VisDrone and SARD data sets to detect humans. The detection results are transmitted to the Ground Control Station to locate the victim's position. Offloading heavy computation from the UAV to the cloud improves power efficiency and precision while maintaining real-time capabilities with accuracy of 78.72%.
Keywords: unmanned aerial vehicle; 4G; deep learning; search and rescue; mobile-edge computing; fog computing.
DOI: 10.1504/IJVAS.2023.136180
International Journal of Vehicle Autonomous Systems, 2023 Vol.17 No.1/2, pp.106 - 132
Received: 29 Nov 2022
Accepted: 29 Oct 2023
Published online: 19 Jan 2024 *