Flight Obstacle Detection Using A Fisheye Camera for Unmanned Aircraft Risk Degradation-
Tatsuki YAMANAKA, Yuichi Yaguchi · The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec) · 2025
Uncrewed Aerial Systems (UAS) must be capable of autonomously detecting and avoiding obstacles for safe operation. Object recognition with a wide 360-degree field of view is essential to achieve this objective. However, conventional object detection algorithms often cannot handle the distortion of fisheye camera images. This research focuses on developing a deep learning-based object recognition system using yolov5, which is specialized for fisheye camera images. We aim to improve detection accuracy by building a dataset of labeled fisheye images and fine-tuning the model. This research will contribute to developing autonomous navigation for unmanned aerial vehicles.