Efficient computation of the Gauss-Newton direction when fitting NURBS using ODR

摘要: We consider a subproblem in parameter estimation using the Gauss-Newton algorithm with regularization for NURBS curve fitting. The is fitted to set of data points least-squares sense, where sum squared orthogonal distances minimized. Control-points and weights are estimated. knot-vector degree kept constant. In algorithm, search direction obtained from linear overdetermined system Jacobian residual vector. Because properties our problem, has particular sparse structure which suitable performing splitting variables. handling computational problems report accuracy different methods, elapsed real time. variables two times faster method than plain normal equations.