Dynamic Maritime Environment Object Tracking Method for MASS

Accurate visual tracking of maritime objects is vital for safe navigation in Maritime Autonomous Surface Ships (MASS). However, occlusion, non-linear object motion, and camera blur pose significant challenges in real-world maritime environments. To address these issues, this work proposes the Dynamic Maritime Object Tracking model, which enhances detection accuracy and enables robust tracking under highly dynamic conditions. The tracking stage employs an Observation-centric Kalman filter (OCKF) to correct large prediction errors caused by non-linear motion. Additionally, the introduction of the Robust Horizontal Observation-Centric Momentum Estimation (RHOCME) strategy better addresses the issues of frequent ID loss caused by shaking of MASS-induced instability. Experiments on the Singapore Maritime Dataset show that our model outperforms state-of-the-art methods under highly dynamic conditions, enhancing MASS situational awareness and improving navigational safety.