A weighting registration method based on LM nonlinear optimization

When industrial CT equipment scans the part, a series of CT slices are got. Because of some practical reasons, it isn't avoidable to have some position and orientation problems between the part's three dimensional CT slice series and its CAD model. So it is necessary to match the part's three dimensional CT slice series and its CAD model. A weighting model registration method is proposed based on Levenberg-Marquardt nonlinear optimization. Firstly initial point set is selected from three dimensional CT slice series' contours, then optimized object function is defined as distances' weighting mean square sum between initial point set and CAD model. The minimal object function is searched to get the position and orientation registration parameters of space transform by Levenberg-Marquardt nonlinear optimization. Finally the method can establish space transform according to registration parameters from the part's three dimensional CT slice series to its CAD model. In the optimization, the method changes six variables' optimization to three variables' optimization. It makes the optimization simple and reduces a large amount of computation. Simulation results of hollow turbine blade show the capabilities of the registration algorithm.