Saliency Superpixel Matching for Underwater Terrain-Aided Localization

Underwater terrain image matching provides a new approach for underwater vehicle terrain matching localization. Underwater terrain images are usually generated by the projection of elevation values measured by acoustic bathymetry sensors. Due to the special characteristics of the sensor working principle, underwater terrain structure and underwater working environment, the underwater terrain images have flat local variations and there are often nonlinear intensity differences and plane rotations between image pairs, thus posing a great challenge to the current image matching algorithm. To address this problem, we propose a new terrain image matching method, which refers to the main steps of feature point-based matching methods and replaces the process of extracting feature points by selecting saliency regions in the image. Specifically, we propose an image matching scheme based on saliency superpixel. The saliency values of the superpixels are quantified by the richness of local topographic information and are therefore more robust to intensity differences. Then, we describe the superpixels in terms of their domains using the maximum index map and topographic information, and we also improve the robustness of the maximum index map to rotation by rotating the domains. Experimental results show that our method is more robust to nonlinear intensity differences and rotations in the underwater terrain image matching task and outperforms the compared methods.