An Autonomous Scheme Leveraging Deep Learning Techniques to Adjust the Gimbal Camera Angle in Real-Time for Victim Positioning by Drone

Search and rescue operations are expensive and resource-intensive. It's critical to locate missing persons promptly to increase their chances of survival and reduce search and rescue expenses. Unmanned aerial vehicles (UAVs), commonly known as drones, are increasingly utilized to monitor and support rescue missions. When used with unmanned aerial vehicles (UAVs), search algorithms allow for real-time analysis of image sequences and timely reporting of results. This has made it feasible for rescue teams to respond to the scene as soon as possible, textcolorbluewhich has benefited the search efforts. Artificial Intelligence (AI) algorithms use Convolutional Neural Networks (CNN) to identify small objects in real-time. This paper uses the Yolov5 model to detect victims on the ground and can identify and locate victims at up to 20 frames per second, ensuring real-time broadcasting with error in victim positioning reduced to around 10 meters. This study also examines the impact of camera rotation angle on the accuracy of coordinates and position of victims in different scenarios with the proposed gimbal camera control method in conjunction with deep learning techniques to improve GPS errors.