A Classical Computer Vision Pipeline for Underwater Detection of Long, Flexible, and Highly Deformable Curvilinear Objects

Vision-based algorithms have been widely used for detection of underwater objects that are long, flexible, and highly deformable - like cables, pipelines, ropes, wires, and long strands of fishing net. Most algorithms use a deep learning or a combination of deep learning and classical computer vision approach in order to detect these specific objects. This paper presents a pure classical computer vision approach that aims to solve this detection problem by combining several state-of-the-art computer vision techniques. The implemented algorithm is tested using real world data collected using a AUV.