An Arctic underwater terrain matching method integrating template matching and DEM super-resolution

Terrain matching serves as a critical technique in underwater terrain-aided navigation systems, playing a vital role in ensuring high-precision positioning and navigation stability. However, underwater terrain matching in the complex Arctic waters faces two core challenges: the low resolution of reference Digital Elevation Models (DEMs) and the poor quality of measured multibeam sonar imagery. These two factors severely constrain matching accuracy, making it difficult to meet the demands of high-precision underwater positioning and navigation. To address these challenges, an underwater terrain matching method integrating template matching with DEM super-resolution is proposed. Specifically, a self-supervised learning-based super-resolution approach is first designed to enhance the spatial resolution of the reference DEM, achieving resolution-level alignment with the measured multibeam sonar imagery. Subsequently, a multi-channel oriented gradient template matching method incorporating terrain slope features is introduced to extract more discriminative high-level terrain information from the aligned images, significantly improving matching accuracy and noise resistance between sonar imagery and DEM. This method balances image detail restoration with feature matching accuracy, substantially enhancing the overall performance of the navigation system. Extensive comparative experiments conducted on multibeam measurement data from four Arctic routes demonstrate that the proposed method outperforms existing mainstream methods in both matching accuracy and robustness, while maintaining stable performance under various noise conditions. Simulation tests further validate that the method achieves excellent positioning accuracy in both rugged and flat terrain regions, demonstrating broad application prospects. • A terrain matching method tailored for Arctic underwater environments. • Self-supervised learning aligns DEM resolution with sonar imagery. • Terrain-aware template matching improves multimodal data registration. • Method achieves robust localization across diverse Arctic terrains.