Search and rescue operation using UAVs: A case study

Abstract

Many people go missing in the wild every year. In this paper, the Search and Rescue (SAR) mission is conducted using a novel system comprising an Unmanned Aerial Vehicle (UAV) coupled with real-time machine-learning-based object detection system embedded on a smartphone. Human detection from UAV in the wilderness is a challenging task, because of many constraints involved such as lack of computing and communication infrastructures. We proposed a novel combination of a robust architecture deployed on a smartphone and a novel Convolutional Neural Network (CNN) model to fulfil the goals of the project. Our approach achieved 94.73% of accuracy and 6.8 FPS on a smartphone. Our approach is highly portable, cost-effective, fast with high accuracy. This novel system is expected to contribute significantly to maximise chances of saving lives in the wild. This developed system has been recently launched by Police Scotland to facilitate the SAR teams to locate missing persons in Scotland wilderness.