Large 3D free surface measurement using a mobile coded light-based stereo vision system

A mobile coded light-based stereo vision system for measuring large 3D free surface is proposed. The system combines the technology of binocular stereo vision and multi-line structured light, which simplifies the process of finding correspondence. It can be freely moved around measured objects, and acquires local 3D data in various viewpoints. The methods of extracting light stripe are outlined. A novel approach, pyramid sub-grate projecting approach, is developed for encoding structured light. The approach can subdivide a high-density light grate into multiple sparse sub-grates. Each of the sub-grates can be effortlessly encoded by the time-multiplexing coding method. Given a point on left image, the correspondent point on right image is the intersection of coded light stripes with the same code and epipolar line. Then, the depth is estimated by triangulation. Moreover, a new registering 3D data based on a planar baseline target is detailed. The mathematic model of this registration method is given. An objective function about all parameters of rotation matrix and transmission vector being established, and those parameters can be found by Levenberg-Marquardt optimization Algorithm. However, the transmission vector is scaled by an unknown coefficient. Then the method of ascertaining the coefficient is also described. Compared with iterative closest point (ICP) algorithm, a popular registration method, the present registering method costs much less time, and is suitable for all kinds of objects. Some experimental results are presented to demonstrate the proposed technology. The results show that high-resolution 3D data of free surface has been acquired and 3D data have been perfectly registered. (c) 2006 Elsevier B.V. All rights reserved.