A Clustering-Aided Template Matching Algorithm Towards Underwater SLAM Using Imaging Sonar

In man-made marine infrastructures, elements such as pillars, cables or ducts are common, which provide distinctive landmarks for Simultaneous Localization and Mapping purposes. In this work, we concentrate on the application of template matching to acoustic imagery for landmark detection and tracking, building on the modeling of common elements in marine environments. The proposed algorithm extends on the original method by employing a density-based clustering technique for match candidate selection and leveraging vehicle inertial information to identify regions of interest in the acquired images, tackling performance deterioration resulting from motion-induced image deformation and overall acoustic feature ambiguity. Experimental results are provided based on datasets collected in a testing pool environment.