ABSTRACT

This chapter looks at the main approaches for obtaining 3D information using camera systems and outlines a step-by-step workflow for recovery 3D shape using stereo imaging. It focuses on stereo systems as it is felt that the effort and expense associated with underwater inspections warrant the use of more dependable solutions. The chapter provides key steps such as stereo camera calibration, finding matching points in two images, and generating a 3D mesh and demonstrates the process on stereo images of a ship hull that is colonized by barnacles. It focuses on checkerboard-based calibration and rectification and discusses strategies for performing self-calibration. The chapter outlines the process of upgrading and cleaning a 3D point cloud to a watertight mesh and describes an advanced stereo correspondence algorithm. It shows how the problem of disparity map estimation can be formulated on a Markov random field and solved using Belief Propagation.