Vision-Based Autonomous UAV Tracking and Control

Abstract

This paper presents a vision-based control approach for unmanned aerial vehicles (UA Vs), focusing on the detection and tracking of airborne objects. The proposed system integrates deep-learning-based object detection using YOLO models with computationally efficient tracking algorithms to ensure realtime performance. The control methodology involves extracting positional information from visual data and generating attitude commands to regulate UA V movement via MA VLink communication. The implementation is optimized for deployment on a Raspberry Pi 5, leveraging OpenCV and NCNN frameworks. Experimental results demonstrate the system’s capability to detect and track small UA Vs while maintaining high frame rates, enabling reliable feedback-based flight adjustments.