Optical three-dimensional coordinate meter with vision guidance for submillimeter pointing to tags

A laser rangefinder-based experimental 3D coordinate meter with vision guidance for pointing at visible features is presented, and its measurement performance on circular tags is tested. This is conceived of as part of an automated 3D measurement system concept that comprises CAD-based graphic measurement planning producing a measurement model file (MMF), actual measurements executed according to the information in the MMF and with vision guidance, analysis of the results, and decisions regarding acceptance of the part measured, etc. The automated 3D measurement will be used for dimensional control of large manufactured objects and in other industrial applications. The experimental system is based on the Acman 200 Automatic 3D coordinate meter, in which the target aiming sensor based on a position-sensitive detector is replaced with the vision sensor and a control system. The control system guides the laser rangefinder to point at the nominal position of the target point according to the information in the MMF, then the vision sensor is used to guide the measurement exactly to the target point, and finally the range value and the two corresponding angle values to the target point are measured and the related x, y, z coordinates are computed. The vision sensor employs template matching and centroid methods for coarse and accurate tag finding, respectively, and it is situated on the same optical axis with the laser rangefinder and red pointing beam in the measuring head of the coordinate meter. Experimental results on the reliability, repeatability, accuracy, stability, and execution time for pointing operation and 3D coordinate measurement are given i a measurement volume of 13 × 11 × 4 m. The repeatability of the vision-guided pointing was better than 0.02 mrad, and the worst-case pointing accuracy was 0.04 mrad. The repeatability of the 3D coordinate measurements was 0.3 mm (standard deviation of x, y, and z) on the average, and the accuracy was better than ±1 mm (for x, y, and z) for 81% of the measured coordinate values.