Application of Computer Vision in Helicopter Obstacle Detection and Avoidance

Abstract

This paper explores the application of computer vision technology in helicopter obstacle detection and avoidance. A multi-sensor fusion and deep learning-based obstacle detection system was developed, employing an improved YOLOv5s model for real-time recognition. The system demonstrated excellent performance in complex environments, achieving an accuracy rate of 98% and a positioning precision of 0.1 meters. By combining the improved RRT* algorithm and model predictive control, intelligent path planning and dynamic obstacle avoidance were implemented, achieving a 99.5% avoidance success rate in typical scene tests. The results indicate that computer vision-based methods significantly outperform traditional technologies in key metrics such as detection distance, accuracy, and response time, providing a new technical avenue for enhancing helicopter flight safety.