Nonmetric Camera Calibration for Underwater Laser Scanning System

Abstract
This paper describes the development of an underwater camera system with a laser line source to measure seafloor features at millimeter scales. The quality of underwater photography is limited by the visibility of the water column. In real underwater environments, there are always suspended particles in the water column and light is scattered by these particles. As a result, photographic images taken under such conditions will be blurred. The stronger the light source is, the more severe the blurring will be. Therefore, it is difficult to discern the features of the intended target directly from the underwater images. To overcome this problem, a laser stripe can be projected onto the target and the profile of the target can be inferred from the displacement of the laser scan lines relative to a reference baseline. With a calibrated camera, the displacement expressed in pixels can be converted into the dimensions of the target in engineering units. To obtain a broader view in a closer distance, a wide-angle lens is usually used. As a result, the image taken with the wide-angle lens is nonlinear and is strongly distorted at the edges of the image. Calibration of a camera involves finding the optical and geometrical parameters of the camera and the environment in which it works. In this paper, a modified coordinate mapping calibration procedure is used. We divide the scope of the camera into several regions and build linear mappings between the world coordinate system and the pixels in the regions. We lay vertical and horizontal grid lines separated by 50 mm on an acrylic board that is aligned with the laser scanning sheet. These grid lines serve as longitude and latitude lines of a map. On the captured image, we curve-fit the grid points in pixel coordinates. A pair of interpolated longitude and latitude lines which pass through the target point are used to estimate the location of the point in the world coordinate system. We assess the accuracy of this procedure with test pieces (grooved blocks and seabed ripples) fabricated by a computer numerical control milling machine. Our measurements show that the error is less than 1.5 mm when the target is scanned from a distance of 1 m.

This publication has 13 references indexed in Scilit: