An accuracy evaluation system for ship components was developed at the University of Tokyo in order to explore the application of laser scanners to shipbuilding. In this system, the shapes of the curved shell plates are obtained as point cloud data. The point cloud data and the design data are registered together and the displacement errors are evaluated and visualized in the form of color maps and histograms. k-d tree construction is performed on the scanned point cloud data to fasten the neighbor searching of each point. To extract the continuous domain of the component from the point cloud data, a region growing method is performed on every neighbor point of the seed point. Then the neighbor domain, which is separated by obstacles' shadows from the extracted domain, is recognized by conducting curved surface fitting and B-spline curved line fitting at each edge point of the extracted domain. In order to make the whole system easier for workers to use in shipyards, the developed automation engine of measurement/analysis flow sends every system operation command instead of manually operation. By adopting the developed system, the rate of the finished plates that fell out of the fixed accuracy's tolerance rage decreased from 75% to 45% within just 5 months.