摘要：The evaluation process of the curved shell plates' accuracy carries great risks for no quantitative criteria, and great depending on implicit knowledge, the skill and experience of the workers. An accuracy evaluation method for the curved shell plates using laser scanners was proposed. The shape of the objects is obtained as point cloud data. In the system, the point cloud data and the design data are registered and the displacement errors are evaluated and visualized by a color map and histogram. Before the two input data's registration, the points of curved shell plates are extracted using region growing method. To extract the curved shell plates from the point cloud with noises, a new extraction algorithm which is more efficient and faster is proposed in this paper. And the ICP algorithm for registration of the design data and the point cloud data is also developed to automatize the registration operation. In practice stage, automation engine which can operate the scanner directly has been arranged. Also, the measurement and analysis flow has been totally automated for ease of use. By deploying the improved system to factory for about 5 months, the evaluating process was automatized, which lead directly to the immediate correction processing and obvious decrease of the error dispersion.
