超二次関数を用いたモデルと測定点の一致方法に関する研究

抄録

This paper describes how to fit a model and 3D range points. The model consists of superquadrics. M-estimation is used for the function fitting problem. And genetic algorithm is used for the determination of the initial parameter values. Our algorithm proceed followed steps.
First, a distance of point from the each part of superquadrics is calculated and a point of attribute is determined. And all of the points are calculated as same. Next, M-estimators of each range point is calculated. And the estimation function is defined by the M-estimators. Third, the polytope method is used for optimization of the estimation function. And the optimization is iterated using initial parameter values given by genetic algorithm. The experiment was conducted by the object of simple structure.