Accuracy Optimization Research On NURBS-based Human Calf Modeling

High-precision parameterized models of specific human body parts are essential for achieving refined and precise design and manufacturing of medical products and instruments in the field of healthcare and rehabilitation.In this study,we focused on the precise customization of medical compression stockings and carried out related research on highprecision reconstruction of the human calf surface based on small number of key points.Non-uniform rational b-spline(NURBS)is currently the standard for geometric modeling and data exchange in the industry.In NURBS surface fitting,the configuration of the knot vector directly determines the quality of the surface reconstruction,which is very challenging due to non-linearity and non-convex properties of this problem.On the other hand,T-splines have recently becomes frequently-used modeling approach in related fields.As a generalization of NURBS,T-splines break the restriction of the NURBS tensor product topology and demonstrate great potential in reverse engineering,isogeometric analysis,numerical control machining,and other fields.On these basis,this work conducted technologies improvements and realize optimized knot placement and enhanced T-spline modeling to achieve highprecision reconstruction of leg surfaces.The specific work and achievements are presented below.1.An improved meshing method for point cloud was proposed.The point cloud is reconstructed layer by layer,and each section contour was fitted with high-precision NURBS curves.Then,the obtained section curves are discretized,sampled with a uniform arc length using a unified and uniform radian sequence in polar coordinates.As a result,high-precision uniformly distributed mesh points are obtained,which provided a solid foundation for subsequent surface reconstruction work.2.A particle swarm optimization-based surface reconstruction method was proposed.Firstly,closed curve fitting method for leg cross-sectional contours was improved using periodic knot vectors and control points,which can achieve second-order continuous NURBS closed curve fitting and further a smooth closed surface for the leg.Then,a particle swarm optimization-based knot replacement method is proposed to adjust the positions of knots based on a the fitness function in order to determine the optimal node vector.Experimental results showed that the surface reconstruction accuracy of the calf is improved by an average of 10%compared to traditional knot placement method,with an error as low as 0.29 mm,which is nearly70% lower than previous works.3.T-spline skinning algorithm was implemented for leg surface reconstruction.Theoretical aspects of T-splines and T-spline surface skinning methods were studied.For the task of reconstructing the calf surface,the approximation method for the control curve in the T-spline skinning algorithm was improved.Specifically,the control curve was discretized and directly approximated using NURBS curve fitting method to obtain an approximate control curve.This approach improved computational efficiency without compromising precision.The experimental results demonstrated that compared to the B-spline method,T-spline skinning reduced the number of control points by up to 38%,and achieved reconstruction accuracy with an error as low as 0.2mm.These results validated the superiority of T-splines.Compared to previous works,the proposed methods in this thesis fully utilized the shape information conveyed by the leg point cloud,and achieved significant improvement in modeling accuracy.The related methods and techniques can not only be applied to the precise and fast customization of compression stockings,but also provide important references for the personalized design and manufacturing of other compression garments.