This study proposes a method to estimate vessel positions in a harbor by relating real-world coordinates obtained from RTK-GNSS to image coordinates extracted from video footage through a projective transformation (homography). Unlike conventional approaches that rely on fixed landmarks, the proposed method utilizes the vessel's own trajectory to establish calibration correspondences based on time-synchronized RTK-GNSS measurements and image-based vessel tracking. To ensure robustness against measurement noise, the Random Sample Consensus (RANSAC) algorithm is applied to remove outliers prior to estimating the transformation matrix. Field experiments were conducted by filming vessels navigating within a harbor, and the estimated coordinates were compared with ground truth data from RTK-GNSS. The results demonstrated that the proposed method achieved a practically acceptable level of accuracy for trajectory analysis, with average positional errors ranging from several meters to approximately 20 meters. These findings indicate that the proposed method is effectively applicable to harbor monitoring and maritime traffic surveys without the need for ground control points. Future work will focus on improving robustness under varying environmental conditions and developing real-time processing capabilities.
