| Abstract | The unmanned surface vehicle (USV) is recently getting more attention with the development of autonomous navigation technology. The USV operates for various purposes, such as marine reconnaissance and survey, and it can operate as a group according to specific scenarios. In this study, a method of obstacle detection and tracking of USVs using multi-view images in marine environment was proposed. We assumed each USV detected marine obstacles by the camera mounted on itself and tracked the position and speed of the obstacles. For this, the YOLO algorithm, which is one of the real-time object detection algorithms based on CNN (Convolutional Neural Network), was adopted after fine tuning. Subsequently, the relative position of the detected obstacle was estimated through the position estimation algorithm of a monocular camera. Then, the position, speed, and course of each obstacle were tracked using the extended Kalman filter. Finally, the accuracy of the tracking results was improved through the data fusion of the tracked data from multiple USVs. The proposed method was validated through the actual field test using multiple USVs. As a result, it was confirmed that the proposed method could be useful in detecting and traking the obstacles in maritime environment. |
|---|---|
| Publication Date | 2022-09-13 |
Jeong-Ho Park, Myung-Il Roh, Hye-Won Lee, Jisang Ha, Yeong-Min Jo, Nam-Sun Son, "Obstacle Detection and Tracking of Unmanned Surface Vehicles Using Multi-view Images in Marine Environment", Proceedings of ICCAS(International Conference on Computer Applications in Shipbuilding) 2022, Yokohama, Japan, 2022.09.13-15