AUTONOMOUS SEARCH AND RESCUE DRONE WITH REAL-TIME YOLO OBJECT DETECTION AND GPS NAVIGATION USING RASPBERRY PI 4

Abstract

The development of intelligent unmanned aerial vehicles with onboard computer vision capabilities is increasingly relevant to search and rescue operations. However, real-time human detection on compact UAV platforms remains challenging due to limited computational resources, varying environmental conditions, and the small size of distant targets. This study presents a UAV-based human detection and localization system integrating a Raspberry Pi 4 Model B, YOLOv8, and u-blox NEO-M8N GPS. The system performs onboard processing of the video stream and associates detected targets with the UAV's geographical position. The YOLOv8 model was trained using the VisDrone2019-DET dataset and evaluated through real-world flight experiments under different distances, altitudes, and lighting conditions. The system achieved an average detection confidence of 0.67, with a maximum confidence of 0.81, and a real-time processing speed of 4–8 FPS. Human targets were detected at distances of up to 30 m, while GPS positioning errors ranged from 1.6 to 2.8 m, with a mean error of approximately 2.12 m. The results demonstrate the capability of the proposed UAV platform to perform onboard human detection and GPS-based localization in real-time, providing a basis for UAV-assisted search and rescue applications.