Yun-Sik Kim, Myung-Il Roh, Ha-Yun Kim, In-Chang Yeo, Nam-Sun Son, "An Improved Method for Detection and Tracking of Maritime Obstacles Using Multiple-Sensor Fusion in Real-World Maritime Environments", Proceedings of G-NAOE 2026, Houston, USA, 2026.10.20-24
| Abstract | For the safe and reliable navigation of USVs (Unmanned Surface Vehicles), it is essential to detect and track maritime obstacles autonomously. To achieve this, USVs typically employ multiple sensors, such as a camera and RADAR (RAdio Detection And Ranging), as a single sensor cannot provide consistent performance under differences in sensing range, frequency, and error characteristics. In particular, a camera and RADAR are complementary, since their strengths compensate for each other’s limitations. A camera provides high-resolution visual data for detailed recognition but is sensitive to lighting conditions and requires significant processing. In contrast, RADAR provides long-range detection and stable performance under various lighting and weather conditions, but its low resolution limits object classification and lateral distance estimation. Accordingly, integrating a camera and RADAR is important to leverage their complementary strengths and achieve reliable detection and tracking across diverse maritime environments. To address this challenge, this study proposed an improved method for detecting and tracking maritime obstacles using multiple-sensor fusion in real-world maritime environments. The proposed method consists of three main stages: (1) detection and localization, (2) tracking, and (3) sensor fusion. In the detection and localization stage, we proposed a 3D detection-based localization method for the camera to jointly estimate obstacle detection and localization, reducing error accumulation from a sequential 2D detection-based localization method. For RADAR, obstacle detection and localization were performed using CA-CFAR (Cell-Averaging Constant False Alarm Rate) to ensure robust detection in cluttered maritime environments. In the tracking stage, a hybrid tracking method combined EKF (Extended Kalman Filter)-based models with a learning-based model to improve stability and adaptability under varying conditions. In the sensor fusion stage, tracking results from the camera and RADAR were integrated using a sensor-level fusion method based on the estimated error covariance. The proposed method was validated through field experiments using camera and RADAR measurements in real-world maritime environments, and the results demonstrated improved detection accuracy and tracking stability compared to single-sensor methods. |
|---|---|
| Publication Date | 2026-10-20 |
