Coastal Ship Tracking with Memory-Guided Perceptual Network

Coastal ship tracking is used in many applications, such as autonomous navigation, maritime rescue, and environmental monitoring. Many general object-tracking methods based on deep learning have been explored for tracking, but they often fail to accurately track ships challenging scenarios, occlusion, scale variation, motion blur. We propose a memory-guided perception network (MGPN) address these issues. MGPN has two main innovative improvements. The dynamic memory mechanism (DMM) the proposed method stores past features of tracked target enhance model’s feature fusion capability temporal dimension. Meanwhile, hierarchical context-aware module (HCAM) enables interaction different scales, global local information, discrepancy targets improve spatial These innovations robustness reduce inaccuracies bounding boxes. conducted an in-depth ablation study demonstrate effectiveness DMM HCAM. Finally, influenced by above points, achieved state-of-the-art performance large offshore dataset, which contains scenarios complex backgrounds, varying scales.