Unmanned aerial vehicle (UAV) photogrammetry is indispensable for Arctic sea ice research, enabling high-resolution mapping and supporting critical operations. However, drifting sea ice, which is characterized by continuous motion that violates the fundamental ‘static-scene’ assumption, and no ground control points (GCPs), poses challenges to orthomosaic accuracy. This study presents a systematic framework for assessing and improving the relative accuracy of UAV photogrammetry over drifting Arctic sea ice, using 18 shipborne UAV missions conducted during the Following Arctic/Antarctic iCE 2024 expedition. A time-dependent correction method based on synchronized vessel GNSS data, referred to as drift correction, was employed to compensate for drift in the exterior orientation parameters of the UAV imagery. In the absence of GCPs, relative accuracy was evaluated using shipborne constrained and check scale bars. Results show that under raw conditions, with a ground sampling distance (GSD) of 2–5 cm, the orthomosaics achieved a mean root mean square error (RMSE) of 0.304 m. The RMSE showed a strong positive correlation with ice drift speed (r = 0.71) and drift distance (r = 0.79), while the flight–drift angle showed negligible influence (r = −0.13). Drift correction alone reduced the mean RMSE to 0.075 m, achieving error reductions of up to 0.6 m under high-drift conditions. The incorporation of scale bar constraints further enhanced accuracy to a mean RMSE of 0.043 m. These findings validate the effectiveness of drift correction and scale-constraint-based optimization, offering a practical framework for accuracy assessment and mission implementation in dynamic polar environments.