Multi-axis, structured-light, three-dimensional laser scanning system: Modeling, calibration, and measurement uncertainty assessment.

Abstract

This dissertation has focused on the development of a three-dimensional, structured-light, multiple-axis laser scanning system and its modeling, calibration, and measurement uncertainty assessment techniques. A non-contact light-striping (structured-light) optical coordinate measuring machine (CMM) is developed. The configuration of the machine is multi-axis, capable of multi-degree-of-freedom three dimensional measurement. A system modeling approach based on a unique skewed frame representation without the use of pin-hole camera model assumption is presented. It is demonstrated that the extrinsic calibration matrix can be decomposed into two classes of transformations, one homogeneous and the other nonhomogeneous. The nonhomogeneous transformation between a Cartesian world frame and the non-Cartesian skewed sensor frame is studied. It is found that there are only four constraints in this transformation instead of the six constraints for common homogeneous coordinate transformations. The sensitivity of the dimensional deformation on the two skew angles is simulated. A new target-based method for the extrinsic calibration of the system is developed. The calibration is based upon the scanning of a physical artifact in the working volume of the measurement system whenever a change of position and/or orientation of the sensor occurs or during initial system setup. Based on the results of the modeling, a constrained optimization calibration algorithm is developed. Two types of targets are proposed: ball and tetrahedron. For ball target, due to distortion of the image, conjugate pair identification becomes non-trivial. A unique slice-interpolation algorithm is proposed. Through this calibration, the usually non-orthogonal skewed sensor coordinate system can be related to a Cartesian common world coordinate system. This allows correct interpretation of the measurement results. Experimental studies using both targets show that a micron level calibration accuracy can be achieved. The proposed system calibration procedure makes automated on-line calibration possible. An uncertainty propagation analysis has been performed on target-based extrinsic calibration of the system using first-order Taylor expansion for computation of uncertainties. Uncertainty transmission through each step of the calibration process is derived analytically. This makes on-line measurement uncertainty assessment possible. We have found that uncertainties of measurements are not homogeneous within the measurement volume. The results can readily be applied to uncertainty evaluation of any coordinate measuring device. Extensive statistical model testing has been performed. It is concluded that the analytical solutions are justified under normal machine condition.