Multi‐Beam and Side‐Scan Sonar Image Registration of Submarine Pipeline Based on Phase Congruency

Submarine pipeline is the “main artery” of marine oil and gas transportation, which needs to be detected regularly. However, it is difficult for a single detection equipment to comprehensively analyze and identify the state of the pipeline. Multi‐beam echo sounder can obtain sonar images with accurate position information, and side‐scan sonar system can obtain sonar images with rich texture information, which are complementary to each other. Therefore, in order to accurately fuse these two different sources of seafloor images together, it is necessary to register them. To this end, a new image registration method of multi‐beam and side‐scan sonar based on phase congruency is proposed in this paper. First, the phase congruency model calculation is carried out in the feature extraction stage to obtain the maximum and minimum moments of phase congruency, and the maximum moment map is combined with Canny edge detection. Then the improved Harris corner detection operator and FAST detection operator are used to detect the edge features and corner features, respectively, to obtain stable edge and corner points as the feature points to be matched. Then in the feature description stage, the maximum index map (MIM) feature descriptor is constructed using the improved spatial structure based on the GLOH descriptor and combined with the SURF feature descriptor to form a hybrid descriptor. Finally, in the feature matching stage, the similarity metric of the descriptors is carried out by a two‐stage metric strategy to obtain homonymous point pairs, and the false matches are eliminated by fast sample consensus (FSC) to finally realize the image registration. Experimental results on multiple sets of sonar image data show that the algorithm proposed in this paper exhibits excellent performance in terms of both image matching accuracy and registration precision. © 2025 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.