Yun-Sik Kim, Myung-Il Roh, Ha-Yun Kim, In-Chang Yeo, Nam-Sun Son, "A Method for Robust Tracking and Fusion of Maritime Obstacles Using Multiple Sensor Data", Proceedings of ISOPE (International Society of Offshore and Polar Engineers) 2025, Goyang, Korea, pp. 1849-1854, 2025.06.01-06
International Conference
2025.02.10 14:14
Yun-Sik Kim, Myung-Il Roh, Ha-Yun Kim, In-Chang Yeo, Nam-Sun Son, "A Method for Robust Tracking and Fusion of Maritime Obstacles Using Multiple Sensor Data", Proceedings of ISOPE 2025, Goyang, Korea, 2025.06.01-06
조회 수 321
| 첨부 '1' |
|---|
| Abstract | For the safe navigation of USVs (Unmanned Surface Vehicles), it is crucial that they autonomously recognize maritime obstacles and accurately detect surroundings. To achieve this, USVs typically utilize various sensors such as RADAR (RAdio Detection And Ranging) and cameras. While RADAR has certain limitations, including lower resolution for object classification and reduced accuracy in lateral distance estimation compared to cameras, it offers important advantages such as long-range detection, relative speed measurement, and reliable functionality regardless of lighting conditions. In contrast, cameras provide high-resolution images but face challenges in low-light environments and provide substantial amounts of data that require extensive processing. Since each sensor has its characteristics, such as detection range, frequency, and error distribution, achieving consistent and accurate obstacle tracking with a single sensor is challenging. Therefore, it is necessary to integrate multiple sensors, combining RADAR’s robustness under various weather conditions with the high-resolution capabilities of cameras. In this study, a method for robust tracking and fusion was proposed to track the trajectory, COG (Course Over Ground), and SOG (Speed Over Ground) of maritime obstacles around the USV, utilizing data collected from multiple sensors. Specifically, three different sensor fusion algorithms—low-level fusion, middle-level fusion, and high-level fusion—were implemented within a tracking system. Each algorithm integrated sensor data at various levels. Low-level fusion combined raw data to minimize information loss, while middle-level fusion processes extracted features to enhance object recognition. High-level fusion integrated the outputs from various sensors to improve decision-making. Each fusion approach presented unique advantages and limitations depending on the specific situations, making it essential to choose the appropriate approach based on the application context for optimal performance. The effectiveness of these methods was evaluated using detection data obtained from RADAR and cameras in real-world experiments, assessing their performance in diverse conditions. The fusion results were subsequently compared with ground truth data. The analysis confirmed that integrating data from multiple sensors significantly improved the accuracy of obstacle tracking around the USV, thereby enhancing the overall safety and reliability of autonomous navigation. |
|---|---|
| Publication Date | 2025-06-03 |
