A 3-LEVEL CHECKERBOARD PATTERN (TCP) PROJECTION METHOD FOR CURVED SURFACE MEASUREMENT

This paper describes a new pattern projection method by which the 3D range and shape of a curved surface can be measured simply, rapidly and precisely. A three-level checkerboard pattern (TCP) is projected onto a curved surface, and the projection pattern on the surface is observed by a well-defined camera. The spatial feature codes of the TCP are used to match an observed pattern image to the slide mask pattern. This method provides high spatial resolution and makes it possible to take highly accurate measurements of stable and moving object surfaces. A new approach for rapid and precise calibration of the slide projector is also presented. In our experiments using real objects about 0.8 m from the camera, the surface measurement error was 0.5-1 mm in S.D. and 2 mm maximum, which corresponds to a relative accuracy of about 0.2%. This result coincides well with the theoretical prediction from image digitizing errors. The semi-automatic procedure for surface measurement together with camera and projector calibrations can be completed within 40 s, which includes CPU time of about 5 s in VAX11/780 (1 MIPS)