A Preliminary Study on Positioning of an Underwater Vehicle Based on Feature Matching of Seafloor Images

Abstract

This study presents an algorithm to estimate the displacement of an underwater vehicle from the seafloor videos collected by the vehicle itself. The algorithm is developed based on image feature detection and matching. In this algorithm, the first step is to extract consecutive partially overlapped images from the collected seafloor video. Next is to correct the radial distortion in the images. Considering that the vehicle's attitude is changing with time while conducting seafloor imaging, tangential distortion in the images is corrected for the tilts of the image plane onboard the vehicle camera. The Scale Invariant Feature Transformation (SIFT) method is then adopted to detect and match features in an image sequence. During the interval between two consecutive images, the vehicle displacement in pixel units is estimated from the matched feature points in these two images. The last step of the algorithm is converting the displacement from pixels into units of physical length. The performance of the proposed feature-based positioning algorithm was evaluated based on a 10-minute seafloor video collected in 2015 by using the Fiber-optical Instrumentation Towed System (FITS) at the Yuan-An Ridge off southwest Taiwan.