View-based underwater SLAM using a stereo camera

This paper proposes an underwater visual simultaneous localization and mapping system using a stereo camera. In front-end, the image data is used to execute feature detection and matching. And the vehicle motion is estimated by the feature matching result. In back-end, a global optimization method represented by a graph, is used to eliminate the errors. The mapping result is in the from 3-D point cloud, which is established by the information from color image and depth image.