Underwater pipelines, being critical infrastructure for the transportation of hydrocarbons and other resources, require regular inspection of their condition, taking into account the economic and environmental nature of the consequences of possible accidents. Therefore, one of the key technological challenges today is the development of reliable methods for recognizing underwater pipelines for the purpose of their inspection using video information received by an autonomous unmanned underwater vehicle. A method is proposed for recognizing and tracking an underwater pipeline using optical images using an autonomous underwater vehicle, based on a multi-stage computational data processing scheme, including: vectorization of initial images on a contour basis, selection of visible boundaries of the pipeline in images and calculation of its spatial centerline. The method is based on the use of the author's modification of the Hough Transform algorithm with adaptive limitation of the analysis area and a new version of the author's method for constructing contours using the Otsu's method. The contours obtained using the method have minimal redundancy and sufficient accuracy to identify visible pipeline boundaries using a modified Hough algorithm. The method is characterized by low computational costs in comparison with analogues. The easy calculation of the centerline is carried out on the basis of the application of the local recognition algorithm previously developed by the authors. Computational experiments were conducted to obtain comparative estimates of reliability and computational performance in relation to the contour algorithms of Canny, K-means, Otsu and the boundary detection method (modification of the Hough method). Including comparison assessments with some analogues. The obtained assessments of the effectiveness of the proposed solutions confirmed their effectiveness.
Key words
underwater pipeline, inspection, autonomous underwater vehicle, recognition, optical images, contour algorithms, vectorization