Print

发布时间: 2019-04-16
摘要点击次数:
全文下载次数:
DOI: 10.11834/jig.180469
2019 | Volume 24 | Number 4




    计算机图形学    




  <<上一篇 




  下一篇>> 





拥有指数参数的新三角基
expand article info 汪凯, 张贵仓, 拓明秀
西北师范大学, 兰州 730070

摘要

目的 为了使构造的曲线拥有传统Bézier曲线的良好性质,同时还具备形状可调性、逼近性、保形性以及实用性。方法 首先在拟扩展切比雪夫空间的框架下,构造了一类具有全正性的拟三次三角Bernstein基函数,并给出了该基函数的性质;基于此基函数,构造了相应的拟三次三角Bézier曲线,分析了其曲线的性质,得到了生成曲线的割角算法以及C1,C2光滑拼接条件,同时还提出了一种估计曲线逼近控制多边形程度的三角Bernstein算子;接着在拟三次三角Bernstein基函数的基础上提出一种三角域上带3个指数参数的拟三次三角Bernstein-Bézier基,基于此基生成了一种三角域上的拟三次三角Bernstein-Bézier曲面,该曲面可以构建边界为椭圆弧、抛物线弧以及圆弧的曲面,此外,还提出一种实用的de-Casteljau-type算法,同时还给出了连接两个曲面的G1连续条件。结果 实验表明,本文在拟扩展切比雪夫空间中构造的具有全正性的曲线曲面,能够灵活地进行形状调整,而且具有良好的逼近性以及适用性。结论 本文在拟扩展切比雪夫空间的框架下构造了一类具有全正性的基函数,并以此基函数进行曲线曲面构造。实验表明本文构造的曲线具备传统三次Bézier曲线的所有优良性质,而且具有灵活的形状可调性。随着参数的增大,所生成的曲线能够更加逼近控制多边形,模拟控制多边形的行为。此外,本文在三角域上构造的曲面能够生成边界为椭圆弧的曲面。综上,本文提出的基函数满足几何工业的需要,是一种实用的方法。

关键词

拟扩展切比雪夫空间; 全正性; 割角算法; 三角域曲面; de-Casteljau-tpye算法

New trigonometric basis with exponential parameters
expand article info Wang Kai, Zhang Guicang, Tuo Mingxiu
Northwest Normal University, Lanzhou 730070, China
Supported by: National Natural Science Foundation of China (61861040)

Abstract

Objective The construction of basis functions has consistently been a difficult point of computer-aided geometric design (CAGD). The construction of a class of practical basis functions often plays a decisive role in the development of the geometric industry. The traditional Bézier curves and B-splines have been widely used in CAGD. However, when the control points are determined, the generated curve is relatively fixed with respect to the control points and has a certain rigidity. Although the proposed rational B-spline curve can adjust the curve by adjusting the weight factor, the rational methods have difficulty in predicting the influence of the weight factor on the curve due to its own shortcomings. Researchers have exerted efforts in the past two decades to solve this problem. However, most of the improved methods have the basic properties of the traditional Bézier method and the B-spline method, such as affine invariance, convex hull, non-negativity, geometric invariance and flexible shape adjustability. Moreover, the proposed curve can accurately represent special curves used in engineering, such as conic and hyperbolic curves. However, most of the literature does not discuss the variation diminishing the generated curve. The curve with vanishing variation must have convexity. The curve with the total positivity must have diminishing variation. Therefore, the total positivity of the basis functions indicates that these functions are suitable for geometric design. We can easily obtain rectangular patches with shape parameters through these new curves. However, the Bernstein-Bézier patch over the triangular domain is not a tensor product patch exactly. Therefore, we cannot obtain triangular surfaces with an adjustable shape using the method of tensor product. Surface modeling over triangular domain is important for many applications. Thus, the practical methods for generating surfaces over a triangular domain must be explored. The blossom property in quasi extended Chebyshev space is used to construct a group of optimal normalized entirely positive basis for curve and surface construction. This method enables the extended curve and surface, thereby maintaining the good nature of the traditional Bézier and B-spline methods while preserving shape, shape adjustability, and practicability. Method A class of cubic trigonometric quasi Bernstein basis functions with total positivity is constructed under the framework of the extended Chebyshev space, and the properties of the basis functions are provided. The corresponding curve is presented based on this basis function. The properties of the curve are analyzed. The cutting algorithm of the curve and the smooth connecting conditions are obtained. A trigonometric quasi Bernstein operator for estimating the degree of the control polygon is also proposed. Then, based on the cubic trigonometric quasi Bernstein basis function, a class of trigonometric polynomial basis functions with three shape parameters over the triangular domain is proposed. A type of triangular polynomial patch over the triangular domain is proposed based on this basis functions. This patch can be used to construct patches whose boundaries are elliptical arcs, parabolic arcs, and arcs. A practical de-Casteljau-type algorithm is proposed to calculate the proposed triangular polynomial surface efficiently and stably. In addition, G1 continuous conditions for joining two triangular polynomial patches are provided. Result Experimental results show that the proposed total positivity patch in the frame of Chebyshev space not only can adjust the shape flexibly but also has shape preservation and good approximation. Conclusion We construct a class of basis functions with total positivity under the framework of the extended Chebysh ev space, and construct the curve and surface with this basis function. Experimental results show that the curve constructed in this study has all the excellent properties of a traditional cubic Bézier curve and has flexible shape adjustability. As the parameters increase, the generated curve can be closer to the control polygon, thereby simulating its behavior. In addition, the surface constructed on the triangular domain can generate the surface whose boundaries are elliptical arcs. A de Casteljau-type algorithm for calculating the surface is also provided. In summary, the proposed basis function satisfies the requirements of the geometric industry and is a practical method.

Key words

quasi extended Chebyshev space; totally positivity; corner cutting algorithm; triangular patch; de-Casteljau-type algorithm

0 引言

基函数的构造一直是CAGD(computer aided geometric design)的一个难点,一类实用基函数的构造往往对几何工业的发展起着决定性的作用。众所周知,传统Bézier曲线和B样条已经被广泛运用于CAGD中,然而当控制顶点确定的情况下,生成的曲线相对于控制顶点较为固定,具有一定的刚性。尽管提出的有理B样条曲线可以通过调整权因子的方式对曲线进行调整,但是有理化方法因其本身的缺点,使得权因子对曲线的影响很难预测[1-3]。为了解决该类问题,二十多年来,研究人员做出了大量的努力,主要集中在两个方面:1)在经典的三次Bézier曲线以及B样条曲线中加入参数来增强曲线的调整能力[4-7];2)在非多项式空间,如三角空间、双曲空间构造类似经典Bézier以及B样条性质的曲线曲面[8-14]

上述改进的方法大都具备传统Bézier方法和B样条方法的基本性质,如凸包性、非负性、几何不变性以及灵活的形状可调性,并且提出的曲线能够精确地表示工程上常用的特殊曲线,如圆锥曲线、双曲线等。但是大多数文献并没有讨论生成曲线的变差缩减性,具有变差缩减性的曲线一定具有保凸性,具有全正性的曲线一定具有变差缩减性,因此是否具有全正性是衡量一组基函数是否适合曲线设计的标准之一[15-17]

众所周知,传统的矩形域曲面已经被广泛运用到CAGD中,具有很高的研究与应用价值。由于传统矩形域曲面是传统Bézier曲线通过张量积方法的直接扩展,因此很容易得到带形状参数的矩形域曲面,然而三角域上的曲面并不是一个张量积曲面,因此无法直接通过张量积的方法获得三角域上的带形状参数的曲面。但是在很多应用中,基于三角域的曲面模型非常重要,因此三角域曲面同样具有很高的研究价值。近年来,大量学者也为此做出了很多努力,具体可参考文献[18-23]以及其他相关文献。在文献[18]中,Cao等人构造了一种基于三角域的带一个形状参数的Bernstein-Bézier曲面,在控制网格固定的情况下,通过改变形状参数的值可以得到不同的曲面。在文献[19]中,Shen等人提出了一种基于三角域的带一个形状参数的线性类Bernstein三角多项式基函数。在文献[24]中,Zhang将文献[20]中的C-Bézier基扩展成了基于三角域的带一个形状参数的类Bézier基,该基可以生成边界为椭圆弧的曲面。在文献[21]中,Yang等人给出了一种带$3n(n+1)/2$个形状参数的三角Bézier曲面。在文献[22]中,Yan等人利用经典的递推算法将三角域上的2次Bernstein-Bézier基函数定义成$n$次的基函数,并且基于此基函数提出一种带一个形状参数的三角类Bernstein-Bézier曲面。Zhu等人[23]提出了一种基于三角域的带3个指数参数的$αβγ$-Bernstein-Bézier基函数,当参数取特定值时,该基可以退化成经典的三次Bernstein基函数和三次Said-Ball基函数。

拟扩展切比雪夫(QEC)空间具有适合构造B基的开花性质,成为适合几何设计的最大一类空间[25-30]。另外,基于三角函数空间的类Bézier基和类B样条基在保形设计中具有很大的潜力[31-34]。为此,本文在函数空间${\mathit{\boldsymbol{T}}}_{α, β}=\{1, \sin^{2}t, (1-\sin t)^{2}{\rm e}^{-α\sin t}, (1-\cos t)^{2}{\rm e}^{-β\cos t} \}$构造了一类具有指数函数的新三角基函数,该基不仅具有传统Bernstein基的所有良好性质,如:全正性、非负性、对称性等,而且该基函数中的参数具有张力的作用效果,进而构造的曲线具有变差缩减性等重要的保形性质。此外,在该基的基础上,提出一种在三角域上生成三角多项式曲面的实用方法。首先给出了三角域上带3个形状参数的三角多项式基函数,接着在该基的基础上提出了相应的三角多项式曲面,曲面中的3个形状参数对曲面具有可预测的调整作用。并且该曲面可以用来生成边界为抛物线弧、椭圆弧甚至圆弧的曲面。最后给出了三角域上三角多项式曲面的$\rm G^{1}$连续条件以及一种实用的计算三角域上三角多项式曲面的de-Casteljau算法。

1 拟三次三角Bézier曲线

1.1 最优规范全正基的构造

对任意的$t∈[0, {\rm{ \mathsf{ π} }} /2],α, β∈[0, +∞]$,考虑在如下的拟三次三角函数空间${\mathit{\boldsymbol{T}}}_{α, β}=\{1, \sin^{2}t, (1-\sin t)^{2}{\rm e}^{-α\sin t}, (1-\cos t)^{2}{\rm e}^{-β\cos t} \}$中构造带指数参数的拟三次三角Bernstein(QCT-Bernstein)基,则相应的母函数为

$ \begin{array}{*{20}{c}} {\mathit{\boldsymbol{ \boldsymbol{\varPhi} }}\left( t \right) = }\\ {\left( {{{\sin }^2}t,{{\left( {1 - \sin t} \right)}^2}{{\rm{e}}^{ - \alpha \sin t}},{{\left( {1 - \cos t} \right)}^2}{{\rm{e}}^{ - \beta \cos t}}} \right)} \end{array} $ (1)

由文献[25],只要证明拟三次三角函数空间${\mathit{\boldsymbol{T}}}_{α, β} $的微分空间

$ \begin{array}{*{20}{c}} {{\rm{D}}{\mathit{\boldsymbol{T}}_{\alpha ,\beta }} = \left\{ {2\sin t\cos t,{{\rm{e}}^{ - \alpha \sin t}}\left( {1 - \sin t} \right) \times } \right.}\\ {\cos t\left( { - 2 - \alpha + \alpha \sin t} \right),}\\ {\left. { - {{\rm{e}}^{ - \beta \cos t}}\left( {1 - \cos t} \right)\sin t\left( { - 2 - \beta + \beta \cos t} \right)} \right\}} \end{array} $

是闭区间$[0, {\rm{ \mathsf{ π} }} /2]$上的一个3维QEC空间。

定理1   对任意的$α, β∈[0, +∞] $${\rm D} {\mathit{\boldsymbol{T}}}_{α, β}$为闭区间$[0, {\rm{ \mathsf{ π} }} /2]$上的一个3维QEC空间。

证明  对任意的$ξ_{i}∈ {\bf R}, i=0, 1, 2 $,考虑如下的线性组合

$ \begin{array}{*{20}{c}} {{\xi _0}\left[ {2\sin t\cos t} \right] + }\\ {{\xi _1}\left[ {{{\rm{e}}^{ - \alpha \sin t}}\left( {1 - \sin t} \right)\cos t\left( { - 2 - \alpha + \alpha \sin t} \right)} \right] + }\\ {{\xi _2}\left[ { - {{\rm{e}}^{ - \beta \cos t}}\left( {1 - \cos t} \right)\sin t\left( { - 2 - \beta + \beta \cos t} \right)} \right] = 0} \end{array} $ (2)

$t=0$时,由式(2)可得$ξ_{1}=0$

$t={\rm{ \mathsf{ π} }} /2$时,由式(2)可得$ξ_{2}=0 $;

最后可得$ξ_{0}=0$。由此可见${\rm D} {\mathit{\boldsymbol{T}}}_{α, β}$是一个3维空间。

下面先证明$ {\rm D} {\mathit{\boldsymbol{T}}}_{α, β}$是开区间$(0, {\rm{ \mathsf{ π} }} /2)$上的一个3维完备扩展切比雪夫(ECC)空间。对任意$t∈[a, b] \subset (0, {\rm{ \mathsf{ π} }} /2)$,令

$ \begin{array}{l} u\left( t \right) = \frac{{{{\rm{e}}^{ - \alpha \sin t}}\cos t}}{{{{\sin }^2}t}}\left[ {2 + {\alpha ^2}{{\left( {1 - \sin t} \right)}^2}\sin t + } \right.\\ \left. {\alpha \left( {1 - {{\sin }^2}t} \right) + 2\alpha \sin t\left( {1 - \sin t} \right)} \right] > 0 \end{array} $

$ \begin{array}{l} u'\left( t \right) = - \frac{{{{\rm{e}}^{ - \beta \sin t}}}}{{\sin t}}\left[ {2 + {\alpha ^2}{{\left( {1 - \sin t} \right)}^2}\sin t + } \right.\\ \left. {\alpha \left( {1 - {{\sin }^2}t} \right) + 2\alpha \sin t\left( {1 - \sin t} \right)} \right] - \\ \frac{{{{\rm{e}}^{ - \alpha \sin t}}{{\cos }^2}t}}{{{{\sin }^3}t}}\left\{ {4 + {\alpha ^3}{{\left( {1 - \sin t} \right)}^2}{{\sin }^2}t + \alpha \left( {2 + 4\sin t} \right) + } \right.\\ \left. {2{\alpha ^2}\sin t\left[ {\left( {1 - {{\sin }^2}t} \right) + \sin t\left( {1 - \sin t} \right)} \right]} \right\} < 0 \end{array} $

$ \begin{array}{l} v\left( t \right) = \frac{{{{\rm{e}}^{ - \beta \cos 0t}}\sin t}}{{{{\cos }^2}t}}\left[ {2 + {\beta ^2}{{\left( {1 - \cos t} \right)}^2}\cos t + } \right.\\ \left. {\beta \left( {1 - {{\cos }^2}t} \right) + 2\beta \cos t\left( {1 - \cos t} \right)} \right] > 0 \end{array} $

$ \begin{array}{l} v'\left( t \right) = \frac{{{{\rm{e}}^{ - \beta \cos t}}}}{{\cos t}}\left[ {2 + {\beta ^2}{{\left( {1 - \cos t} \right)}^2}\cos t + } \right.\\ \left. {\beta \left( {1 - {{\cos }^2}t} \right) + 2\beta \cos t\left( {1 - \cos t} \right)} \right] + \\ \frac{{{{\rm{e}}^{ - \beta \cos t}}{{\sin }^2}t}}{{{{\cos }^3}t}}\left\{ {4 + {\beta ^3}{{\left( {1 - \cos t} \right)}^2}{{\cos }^2}t + \beta \left( {2 + 4\cos t} \right) + } \right.\\ \left. {2{\beta ^2}\cos t\left[ {\left( {1 - {{\cos }^2}t} \right) + \cos t\left( {1 - \cos t} \right)} \right]} \right\} > 0 \end{array} $

因此,关于函数$u(t)$$v(t)$的朗斯基行列式[35]

$ \begin{array}{*{20}{c}} {W\left( {u,v} \right)\left( t \right) = u\left( t \right)v'\left( t \right) - u'\left( t \right)v\left( t \right) > 0}\\ {\forall t \in \left[ {0,{\rm{ \mathsf{ π} }}/2} \right]} \end{array} $

$t∈[a, b]$,定义如下3个权函数

$ \left\{ \begin{array}{l} {w_0}\left( t \right) = 2\sin t\cos t\\ {w_1}\left( t \right) = Au\left( t \right) + Bv\left( t \right)\\ {w_2}\left( t \right) = C\frac{{W\left( {u,v} \right)\left( t \right)}}{{{{\left[ {Au\left( t \right) + Bv\left( t \right)} \right]}^2}}} \end{array} \right. $

式中,$A, B, C$是3个任意的正实数。显然,权函数$w_{i}(t)(i=0, 1, 2)$均为$[a, b]$${\bf C} ^{∞}$正的有界函数。考虑如下由$w_{i}(t)(i=0, 1, 2)$定义的ECC空间

$ \left\{ \begin{array}{l} {u_0}\left( t \right) = {w_0}\left( t \right)\\ {u_1}\left( t \right) = {w_0}\left( t \right)\int_a^t {{w_1}\left( {{t_1}} \right){\rm{d}}{t_1}} \\ {u_2}\left( t \right) = {w_0}\left( t \right)\int_a^t {{w_1}\left( {{t_1}} \right)} \int_a^{{t_1}} {{w_2}\left( {{t_2}} \right){\rm{d}}{t_2}{\rm{d}}{t_1}} \end{array} \right. $

可以验证,$u_{0}(t), u_{1}(t), u_{2}(t)$均为函数$\left(\begin{align} & 2\sin \ t\cos \ t, {{\rm e}^{-\alpha \sin \ t}}(1-\sin \ t)\cos \ t(-2\text{-}\alpha +\alpha \sin \ t), \\ & -{{\text{e}}^{-\beta \text{cos }\!\!~\!\!\text{ }\ t}}(1-\text{cos}\ ~t)\text{sin}\ ~t(-2-\beta +\beta \text{cos}~\ t) \\ \end{align} \right)$的线性组合,由文献[25-30, 36]可知函数空间$ {\rm D} {\mathit{\boldsymbol{T}}}_{α, β}$为闭区间$[a, b]$上的ECC空间,由于闭区间$[a, b]$为开区间$(0, {\rm{ \mathsf{ π} }} /2)$上的任意子空间,因此函数空间$ {\rm D} {\mathit{\boldsymbol{T}}}_{α, β}$为开区间$(0, {\rm{ \mathsf{ π} }} /2)$上的ECC空间。

下面证明$ {\rm D} {\mathit{\boldsymbol{T}}}_{α, β}$$[0, {\rm{ \mathsf{ π} }} /2]$上的一个QEC空间。由文献[25-30, 36]知,要证明$ {\rm D} {\mathit{\boldsymbol{T}}}_{α, β}$为QEC空间,只需证明$ {\rm D} {\mathit{\boldsymbol{T}}}_{α, β}$空间中的任意非零函数在区间$[0, {\rm{ \mathsf{ π} }} /2] $上只有2个零点即可。考虑$ {\rm D} {\mathit{\boldsymbol{T}}}_{α, β}$中非零函数

$ \begin{array}{*{20}{c}} {F\left( t \right) = {C_0}\left[ {2\sin t\cos t} \right] + }\\ {{C_1}\left[ {{{\rm{e}}^{ - \alpha \sin t}}\left( {1 - \sin t} \right)\cos t\left( { - 2 - \alpha + \alpha \sin t} \right)} \right]}\\ {{C_2}\left[ { - {{\rm{e}}^{ - \beta \cos t}}\left( {1 - \cos t} \right)\sin t\left( { - 2 - \beta + \beta \cos t} \right)} \right]} \end{array} $

式中,$t∈[0, {\rm{ \mathsf{ π} }} /2]$。由于$ {\rm D} {\mathit{\boldsymbol{T}}}_{α, β}$$(0, {\rm{ \mathsf{ π} }} /2)$上的ECC空间,函数$F(t)$$(0, {\rm{ \mathsf{ π} }} /2) $上至多只有2个零点。

假设$t=0$为函数$F(t)$的零点,则有$C_{1}=0$

在此情况下,如果$C_{2}=0$,则此时$t=0$$t={\rm{ \mathsf{ π} }} /2$分别为$F(t)$的单根;

如果$C_{0}=0$,则$t=0$$F(t)$的唯一根;

如果$C_{0}C_{2}>0$,即$C_{0}>0, C_{2}>0$$C_{0} < 0$$C_{2} < 0$

$ \begin{array}{*{20}{c}} {F\left( t \right) = {C_0}\left[ {2\sin t\cos t} \right] + }\\ {{C_2}\left[ { - {{\rm{e}}^{ - \beta \cos t}}\left( {1 - \cos t} \right)\sin t\left( { - 2 - \beta + \beta \cos t} \right)} \right] = }\\ {\sin t\left[ {2{C_0}\cos t + {C_2}{e^{ - \beta \cos t}}\left( {1 - \cos t} \right)\left( {2 + \beta - \beta \cos t} \right)} \right]} \end{array} $

很明显$t=0$$F(t)$的根,而$t={\rm{ \mathsf{ π} }} /2$不为$F(t)$的根。当$t∈(0, {\rm{ \mathsf{ π} }} /2)$时,

$ {{\rm{e}}^{ - \beta \cos t}}\left( {1 - \cos t} \right)\left( {2 + \beta - \beta \cos t} \right) > 0 $

所以$F(t)$$t∈(0, {\rm{ \mathsf{ π} }} /2)$时恒正或者恒负。因此$F(t)$$t∈[0, {\rm{ \mathsf{ π} }} /2]$上只有1个零点。

如果$C_{0}C_{2} < 0$,即$C_{0}>0, C_{2} < 0$$C_{0} < 0, $ $C_{2}>0$,再令

$ g\left( t \right) = 2{C_0}\cos t + {C_2}{{\rm{e}}^{ - \beta \cos t}}\left( {1 - \cos t} \right)\left( {2 + \beta - \beta \cos t} \right) $

$ \begin{array}{l} g'\left( t \right) = \sin t\left\{ { - 2{C_0} + } \right.\\ \left. {{{\rm{e}}^{ - \beta \cos t}}{C_2}\left[ {2 + 4b + {b^2} - 2b\left( {2 + b} \right)\cos t + {b^2}{{\cos }^2}t} \right]} \right\} \end{array} $

$h(t)=2+4b+b^{2}-2b(2+b)\cos t+b^{2}\cos^{2}t$,容易证明$h(t)≥4$

可知当$C_{0}>0, C_{2} < 0$$g′(t) < 0$时,$g(t)$$t∈(0, {\rm{ \mathsf{ π} }} /2)$单调递减;当$C_{0} < 0, C_{2}>0$$g′(t)>0$时,$g(t)$$t∈(0, {\rm{ \mathsf{ π} }} /2)$单调递增。

可见$g(t)$$t∈(0, {\rm{ \mathsf{ π} }} /2)$上至多只有1个零点,而$t=0$$F(t)$的根,所以$F(t)$$t∈[0, {\rm{ \mathsf{ π} }} /2]$上至多只有2个零点。类似地,若$t={\rm{ \mathsf{ π} }} /2$$F(t)$的零点,可以证明函数$F(t)$$t∈[0, {\rm{ \mathsf{ π} }} /2]$上至多只有2个零点。证毕。

由于$ {\rm D} {\mathit{\boldsymbol{T}}}_{α, β}$$[0, {\rm{ \mathsf{ π} }} /2]$上的3维QEC空间,由文献[25]知$ {\mathit{\boldsymbol{T}}}_{α, β}$中存在开花,这意味着对于$α, β∈[0, +∞]$$ {\mathit{\boldsymbol{T}}}_{α, β}$适合曲线设计。由文献[25]还知$ {\mathit{\boldsymbol{T}}}_{α, β}$$[0, {\rm{ \mathsf{ π} }} /2]$上具有一个B基(即最优规范全正基)。

定理2   对于任意的$α, β∈[0, +∞]$$t∈[0, {\rm{ \mathsf{ π} }} /2]$,函数空间$ {\mathit{\boldsymbol{T}}}_{α, β}$中的B基为

$ \left\{ \begin{array}{l} {T_0}\left( t \right) = {\left( {1 - \sin t} \right)^2}{{\rm{e}}^{ - \alpha \sin t}}\\ {T_1}\left( t \right) = 1 - {\sin ^2}t - {\left( {1 - \sin t} \right)^2}{{\rm{e}}^{ - \alpha \sin t}}\\ {T_2}\left( t \right) = 1 - {\cos ^2}t - {\left( {1 - \cos t} \right)^2}{{\rm{e}}^{ - \beta \cos t}}\\ {T_3}\left( t \right) = {\left( {1 - \cos t} \right)^2}{{\rm{e}}^{ - \beta \cos t}} \end{array} \right. $ (3)

称之为QCT-Bernstein基函数。

证明   对于任意的$α, β∈[0, +∞]$,由式(1)可得

$ \mathit{\boldsymbol{ \boldsymbol{\varPhi} }}\left( 0 \right) = \left( {0,1,0} \right) $

$ \mathit{\boldsymbol{ \boldsymbol{\varPhi} }}\left( {{\rm{ \mathsf{ π} }}/2} \right) = \left( {1,0,1} \right) $

$ \mathit{\boldsymbol{ \boldsymbol{\varPhi} '}}\left( 0 \right) = \left( {2, - \left( {\alpha + 2} \right),0} \right) $

$ \mathit{\boldsymbol{ \boldsymbol{\varPhi} '}}\left( {{\rm{ \mathsf{ π} }}/2} \right) = \left( {0,0,2 + \beta } \right) $

$ \mathit{\boldsymbol{ \boldsymbol{\varPhi} ''}}\left( 0 \right) = \left( {2,{\alpha ^2} + 4\alpha + 2,0} \right) $

$ \mathit{\boldsymbol{ \boldsymbol{\varPhi} ''}}\left( {{\rm{ \mathsf{ π} }}/2} \right) = \left( { - 2,0,{\beta ^2}4\beta + 2} \right) $

由此可得

$ {\mathit{\Pi }_0} = \mathit{\boldsymbol{ \boldsymbol{\varPhi} }}\left( 0 \right) = \left( {0,1,0} \right) $

$ {\mathit{\Pi }_3} = \mathit{\boldsymbol{ \boldsymbol{\varPhi} }}\left( {{\rm{ \mathsf{ π} }}/2} \right) = \left( {0,0,2 + \beta } \right) $

$ \left\{ {{H_1}} \right\} = {O_{S{C_1}}}\mathit{\boldsymbol{ \boldsymbol{\varPhi} }}\left( 0 \right) \cap {O_{S{C_2}}}\mathit{\boldsymbol{ \boldsymbol{\varPhi} }}\left( {{\rm{ \mathsf{ π} }}/2} \right) = \left( {0,0,0} \right) $

$ \left\{ {{H_2}} \right\} = {O_{S{C_1}}}\mathit{\boldsymbol{ \boldsymbol{\varPhi} }}\left( 0 \right) \cap {O_{S{C_1}}}\mathit{\boldsymbol{ \boldsymbol{\varPhi} }}\left( {{\rm{ \mathsf{ π} }}/2} \right) = \left( {1,0,0} \right) $

$t∈[0, {\rm{ \mathsf{ π} }} /2]$,由$\mathit{\boldsymbol{ \boldsymbol{\varPhi} }}{\rm{ }}\left( t \right) = \sum\limits_{i = 0}^3 {{A_i}\left( t \right)} {\mathit{\Pi }_i}{\rm{ }}$可得

$ \left\{ \begin{array}{l} {T_2}\left( t \right) + {T_3}\left( t \right) = so{m^2}t\\ {T_0}\left( t \right) = {\left( {1 - \sin t} \right)^2}{{\rm{e}}^{ - \alpha \sin y}}\\ {T_3}\left( t \right) = {\left( {1 - \cos t} \right)^2}{{\rm{e}}^{ - \beta \cos t}} \end{array} \right. $

由上式连同$\sum\limits_{i = 0}^3 {{T_i}\left( t \right)} =1$,容易推导出式(3)。下面验证$T_{i}(t), i=0, 1, 2, 3$恰为$ {\mathit{\boldsymbol{T}}}_{α, β}$中的B基。

首先,证明$T_{i}(t), i=0, 1, 2, 3$线性无关。对于任意的$ξ_{i}∈ {\mathbf R} (i=0, 1, 2, 3)$,考虑如下的线性组合

$ \sum\limits_{i = 0}^3 {{\xi _i}{T_i}\left( t \right)} = 0 $ (4)

两边对$t$求导得

$ \sum\limits_{i = 0}^3 {{\xi _i}{{T'}_i}\left( t \right)} = 0 $ (5)

$t=0$分别代入式(4)(5),得

$ \left\{ \begin{array}{l} \xi = 0\\ \left( {\alpha + 1} \right)\left( {{\xi _0} - {\xi _1}} \right) = 0 \end{array} \right. $

由此可得$ξ_{0}=ξ_{1}=0$。类似地,由上面两式可得$ξ_{2}=ξ_{3}=0$

容易验证$T_{i}(t), i=0, 1, 2, 3$$[0, {\rm{ \mathsf{ π} }} /2]$上具有非负性,且在$(0, {\rm{ \mathsf{ π} }} /2)$上具有严格的正性。

因为$T_{i}(t), i=0, 1, 2, 3$具有如下的端点性质:

$T_{0}(0)=1$$T_{0}(t)$$t={\rm{ \mathsf{ π} }} /2$处有二重根;

$T_{3}({\rm{ \mathsf{ π} }} /2)=1$$T_{3}(t)$$t=0$处有二重根;

对于$i=1, 2$$T_{1}(t)$$T_{2}(t)$$t=0$$t={\rm{ \mathsf{ π} }} /2$均有2个根。

因此由文献[25]可知QCT-Bernstein基$T_{i}(t), i=0, 1, 2, 3$为函数空间$ {\mathit{\boldsymbol{T}}}_{α, β}$中的B基。证毕。

为了方便讨论,将定义相应的基函数为$T_{i}(t; α, β), i=0, 1, 2, 3$,或者$T_{i}(t; α), i=0, 1, T_{i}(t; β), $ $i=2, 3$图 1给出不同参数下的基函数图像。

图 1 不同参数下的QCT-Bernstein基函数图像
Fig. 1 QCT-Bernstein basis functions with different shape parameters

1.2 拟三次三角Bézier曲线

1.2.1 拟三次三角Bézier曲线的定义和性质

定义1   对于给定的控制点$P_{i}∈ {\bf R} ^{2}/ {\bf R} ^{3}(i=0, 1, 2, 3)$,称

$ T\left( {t;\alpha ,\beta } \right) = \sum\limits_{i = 0}^3 {{P_i}{T_i}\left( {t;\alpha ,\beta } \right)} $ (6)

为带2个指数参数$α$$β$的QCT-Bézier曲线,式中,$t∈[0, {\rm{ \mathsf{ π} }} /2], α, β∈[0, +∞)$

由于QCT-Bernstein基函数具有单位性、非负性和全正性,因此相应的QCT-Bézier曲线具有仿射不变性、凸包性和变差缩减性。此外当$α, β∈[0, +∞)$,有如下端点性质

$ \left\{ \begin{array}{l} P\left( {0;\alpha ,\beta } \right) = {P_0}\\ P\left( {{\rm{ \mathsf{ π} }}/2;\alpha ,\beta } \right) = {P_3}\\ P'\left( {0;\alpha ,\beta } \right) = \left( {2 + \alpha } \right)\left( {{P_1} - {P_0}} \right)\\ P'\left( {0;{\rm{ = }}\alpha ,\beta } \right) = \left( {2 + \beta } \right)\left( {{P_3} - {P_3}} \right)6\\ 2\left( {{P_2} - {P_1}} \right)\\ P''\left( {{\rm{ \mathsf{ π} /2;}}\alpha {\rm{,}}\beta } \right) = \left( {\beta + 4\beta + 2} \right)\left( {{P_3} - {P_2}} \right) + \\ 2\left( {{P_1} - {P_2}} \right) \end{array} \right. $

1.2.2 QCT-Bézier曲线的形状控制

对于$t∈[0, {\rm{ \mathsf{ π} }} /2]$,改写式(6)为

$ \begin{array}{*{20}{c}} {T\left( {t;\alpha ,\beta } \right) = {P_1}{{\cos }^2}t + {P_2}{{\sin }^2}t + }\\ {{T_0}\left( {t;\alpha } \right)\left( {{P_0} - {P_1}} \right) + {T_3}\left( {t;\beta } \right)\left( {{P_3} - {P_2}} \right)} \end{array} $ (7)

显然,对于$t∈(0, {\rm{ \mathsf{ π} }} /2)$$T_{0}(t; α)$关于$α$单调递减,这表明QCT-Bézier曲线随着$α$的增大沿着与边向量$ {\mathit{\boldsymbol{P}}} _{0} {\mathit{\boldsymbol{P}}} _{1}$相同的方向移动。若$α$减小,则情况相反。$β$关于边向量$ {\mathit{\boldsymbol{P}}} _{3} {\mathit{\boldsymbol{P}}} _{2}$具有类似的效果。当$α$$β$增大时,QCT-Bézier曲线将分别趋向控制点$P_{2}$$P_{3}$;当$α$$β$相等且同时增大或者减小时,QCT-Bézier曲线将沿着与边向量$ {\mathit{\boldsymbol{P}}} _{2} {\mathit{\boldsymbol{P}}} _{1}$相同或相反的方向移动。这表明$α$$β$具有张力作用效果。图 2显示的是不同参数对QCT-Bézier曲线的影响。

图 2 不同参数对QCT-Bézier曲线的影响
Fig. 2 QCT-Bézier curves with different shape parameter values

1.2.3 QCT-Bézier曲线的割角算法

下面开发一种稳定高效计算QCT-Bézier曲线的割角算法。为此,将QCT-Bézier曲线写成式(8)的形式。图 3展示了利用割角算法计算QCT-Bézier曲线的全过程以及案例。

图 3 割角算法
Fig. 3 Corner cutting algorithm

$ \begin{array}{*{20}{c}} {A\left( {t;\alpha ,\beta } \right) = \left( {\begin{array}{*{20}{c}} {1 - {{\sin }^2}t}&{1 - {{\cos }^2}t} \end{array}} \right) \times }\\ {\left( {\begin{array}{*{20}{c}} {1 - \sin t}&{\sin t}&0\\ 0&{\cos t}&{1 - \cos t} \end{array}} \right) \times }\\ {\left( {\begin{array}{*{20}{c}} {\frac{{{{\rm{e}}^{ - \alpha \sin t}}}}{{1 + \sin t}}}&{\frac{{1 + \sin t - {{\rm{e}}^{ - \alpha \sin t}}}}{{1 + \sin t}}}&0&0\\ 0&{\frac{{\cos t}}{{\sin t + \cos t}}}&{\frac{{\sin t}}{{\sin t + \cos t}}}&0\\ 0&0&{\frac{{1 + \cos t - {{\rm{e}}^{ - \beta \cos t}}}}{{1 + \cos t}}}&{\frac{{{{\rm{e}}^{ - \beta \cos t}}}}{{1 + \cos t}}} \end{array}} \right) \times }\\ {\left( {\begin{array}{*{20}{c}} {{P_0}}\\ {{P_1}}\\ {{P_2}}\\ {{P_3}} \end{array}} \right)} \end{array} $ (8)

1.2.4 椭圆和抛物线的精确表示

本文提出的QCT-Bézier曲线可精确表示椭圆和抛物线。对$α=β=2$,取控制点

$ \left\{ \begin{array}{l} {P_0} = \left( {{x_0} + a,{y_0}} \right)\\ {P_1} = \left( {{x_0} + a,{y_0} + b/2} \right)\\ {P_2} = \left( {{x_0} + a/2,{y_0} + b} \right)\\ {P_3} = \left( {{x_0},{y_0} + b} \right) \end{array} \right. $

则得式(6)中$T(t; 2, 2)$的参数方程为

$ \left\{ \begin{array}{l} x\left( t \right) = {x_0} + a\cos t\\ y\left( t \right) = {y_0} + b\sin t \end{array} \right.\;\;\;t \in \left[ {0,{\rm{ \mathsf{ π} }}/2} \right] $

这表明$T(t; 2, 2)$为一段四分之一椭圆弧,当$a=b$时,$T(t; 2, 2)$为一段四分之一圆弧。在实际应用中,通过限制参数$t∈[θ_{1}, θ_{2}]$,即可生成所需的任意一段椭圆弧。

此外,对$α=β=2,b-a>0$,取控制点为$P_{0}=(b, c_{2}b^{2}+c_{1}b+c_{0}), P_{1}=(b, c_{2}b^{2}+c_{1}b+c_{0}), P_{2}=((a+b)/2, c_{2}ab+c_{1}(a+b)/2+c_{0})$$P_{3}=(a, c_{2}a^{2}+$ $c_{1}a+c_{0})$,则$T(t; 2, 2)$可写成如下参数方程形式

$ \left\{ \begin{array}{l} x\left( t \right) = \left( {b - a} \right)\cos t + a\\ y\left( t \right) = {c_2}{\left[ {\left( {b - a} \right)\cos t + a} \right]^2} + \\ {c_1}\left[ {\left( {b - a} \right)\cos t + a} \right] + {c_0} \end{array} \right. $

式中,$t∈[0, {\rm{ \mathsf{ π} }} /2]$。这表明QCT-Bézier曲线$T(t; 2, 2)$为一段椭圆弧$y=c_{2}x^{2}+c_{1}x+c_{0}, x∈[a, b]$图 4给出QCT-Bézier曲线对圆和抛物线弧的精确表示(实线部分)。

图 4 圆弧和抛物线弧的精确表示
Fig. 4 The representation of elliptic and parabolic arcs

1.2.5 QCT-Bézier曲线的拼接

在实际应用过程中,需要通过拼接QCT-Bézier曲线来生成几何造型复杂的曲线。设两段QCT-Bézier曲线分别为

$ {F_1}\left( {t;{\alpha _1},{\beta _1}} \right) = \sum\limits_{i = 0}^3 {{P_i}{A_i}\left( {t;{\alpha _1},{\beta _1}} \right)} $ (9)

$ {F_1}\left( {t;{\alpha _2},{\beta _2}} \right) = \sum\limits_{i = 0}^3 {{P_i}{A_i}\left( {t;{\alpha _2},{\beta _2}} \right)} $ (10)

显然,若控制顶点满足$P_{3}=Q_{0}$,则这两段曲线段构成$ {\rm C}^{0}$连续的曲线。

为了方便讨论,对于节点$u_{1} < u_{2} < u_{3}$,设由式(9)和式(10)拼接的曲线$F(u)$

$ F\left( u \right) = \left\{ \begin{array}{l} {F_1}\left( {\frac{{\rm{ \mathsf{ π} }}}{2} \times \frac{{u - {u_1}}}{{{h_1}}};{\alpha _1},{\beta _1}} \right)\;\;\;\;\;\;u \in \left[ {{u_1},{u_2}} \right]\\ {F_2}\left( {\frac{{\rm{ \mathsf{ π} }}}{2} \times \frac{{u - {u_2}}}{{{h_2}}};{\alpha _2},{\beta _2}} \right)\;\;\;\;\;u \in \left[ {{u_2},{u_3}} \right] \end{array} \right. $ (11)

式中,$h_{i}=u_{i+1}-u_{i}, i=1, 2$

定理3   对任意$α_{i}, β_{i}∈[0, +∞),i=1, 2$,曲线$F(u)$在节点$u_{2}$$\rm C^{1}$连续,如果条件

$ {P_3} = {Q_0} = \frac{{\left( {{\alpha _2} + 2} \right){h_1}{Q_1} + \left( {{\beta _1} + 2} \right){h_2}{P_2}}}{{\left( {{\alpha _2} + 2} \right){h_1} + \left( {{\beta _1} + 2} \right){h_2}}} $ (12)

成立。进一步,对任意$α_{i}, β_{i}∈[0, +∞),i=1, 2$,曲线$F(u)$在节点$u_{2}$$\rm C^{2}$连续, 如果式(12)和如下条件同时成立

$ \begin{array}{*{20}{c}} {{Q_2} = \frac{1}{{2{\alpha _2}h_1^2}}\left\{ {2\beta {h_1}{h_2} + {h_2}\left[ {{\alpha _2}{h_2}\left( {\beta _1^2 + 4{\beta _1} + 2} \right) + } \right.} \right.}\\ {\left. {\left. {{\beta _1}{h_1}\left( {\alpha _2^2 + 4{\alpha _2} + 2} \right)} \right]} \right\}\left( {{P_3} - {P_2}} \right) + \frac{{h_2^2}}{{h_1^2}}\left( {{P_1} - {P_2}} \right) + {P_3}} \end{array} $

证明  对任意$α_{i}, β_{i}∈[0, +∞),i=1, 2,$容易验证

$ \left\{ \begin{array}{l} F\left( {u_2^ - } \right) = {P_3}\\ F\left( {u_2^ + } \right) = {Q_0}\\ F'\left( {u_2^ - } \right) = \frac{{\rm{ \mathsf{ π} }}}{2}\frac{{\left( {{\beta _1} + 3} \right)}}{{{h_1}}}\left( {{P_3} - {P_2}} \right)\\ F'\left( {u_2^ + } \right) = \frac{{\rm{ \mathsf{ π} }}}{2}\frac{{\left( {{\alpha _2} + 3} \right)}}{{{h_2}}}\left( {{Q_1} - {Q_0}} \right)\\ F''\left( {u_2^ - } \right) = {\left( {\frac{{\rm{ \mathsf{ π} }}}{{2{h_1}}}} \right)^2}\left[ {\left( {\beta _1^2 + 4{\beta _1} + 2} \right) \times } \right.\\ \left. {\left( {{P_3} - {P_2}} \right) + 2\left( {{P_1} - {P_2}} \right)} \right]\\ F''\left( {u_2^ + } \right) = {\left( {\frac{{\rm{ \mathsf{ π} }}}{{2{h_2}}}} \right)^2}\left[ {\left( {\alpha _2^2 + 4{\alpha _2} + 2} \right) \times } \right.\\ \left. {\left( {{Q_0} - {Q_1}} \right) + 2\left( {{Q_2} - {Q_1}} \right)} \right] \end{array} \right. $

由此可得$F(u^{-}_{2})=F(u^{+}_{2})$, $F′(u^{-}_{2})=F′(u^{+}_{2}),F″(u^{-}_{2})=F″(u^{+}_{2})$。证毕。

图 5显示的是QCT-Bézier的拼接。对于$\rm C^{1}$拼接条件,形状参数为$α_{1}=β_{1}=α_{2}=β_{2}=0$,对于$\rm C^{2}$拼接条件,形状参数为$α_{1}=1, β_{1}=0$$ α_{2}=1, β_{2}=1$

图 5 QCT-Bézier曲线的拼接
((a)$\rm C^{1}$ continue curve; (b)$\rm C^{2}$ continue curve)
Fig. 5 connecting of the QCT-Bézier curves

1.2.6 拟三角Bernstein算子

拟三角Bernstein算子对测量向量函数空间的近似性质是非常有用的,算子的第3大特征值和1的差值可以粗略地估计曲线与控制多边形的近似程度[37],差值越小说明曲线和控制多边形越接近。本小节将构造一种拟三角Bernstein算子来分析给出的曲线式(6)和相应控制多边形的逼近程度。

对于任意的$α∈[1, +∞), t∈[0, {\rm{ \mathsf{ π} }} /2]$,假设

$ \begin{array}{*{20}{c}} {{f_1}\left( {t;\alpha } \right) = \frac{1}{\alpha }{T_1}\left( {t;\alpha } \right) + \left( {1 - \frac{1}{\alpha }} \right){T_2}\left( {t;\alpha } \right) + }\\ {{T_3}\left( {t;\alpha } \right)} \end{array} $ (13)

式中,$T_{j}(t; α), j=0, 1, 2, 3$是式(3)给出的QCT-Bernstein基函数。

直接计算有

$ \begin{array}{l} \frac{{{\rm{d}}{f_1}\left( {t;\alpha } \right)}}{{{\rm{d}}t}} = 2\left( {1 - \frac{2}{\alpha }} \right)\sin t\cos t + \\ {{\rm{e}}^{ - \alpha \sin t}}\cos t\left( {1 - \sin t} \right)\left( {2 + \alpha - \alpha \sin t} \right) + \\ {{\rm{e}}^{ - \alpha \cos t}}\sin t\left( {1 - \cos t} \right)\left( {2 + \alpha - \alpha \cos t} \right) > 0 \end{array} $

不难发现,对任意的$α∈[1, +∞), t∈[0, {\rm{ \mathsf{ π} }} /2]$,函数$f_{1}(t)$是关于变量$t$的一个单调递增函数。此外,$f_{1}(t)+f_{1}({\rm{ \mathsf{ π} }} /2-t)=1$。由此可见,函数$f_{1}(t)$和函数$I(t)=t$具有很多类似的性质。

众所周知,经典的Bernstein算子具有$f_{0}(t)=1$。类似地,本文模仿经典Bernstein算子来构造拟三角Bernstein算子,使得构造的算子具有$f_{0}(t)=1$和式(13)的形式。由于$f_{1}(t; α)$关于变量$t$严格单调递增,且$f_{1}(0;α)=0$$f_{1}({\rm{ \mathsf{ π} }} /2;α)=1$,对于任意固定的$(1/α)∈(0, 1]$,存在唯一的$t^{*}_{α}∈(0, {\rm{ \mathsf{ π} }} /2]$, 有$f_{1}(t^{*}_{α}; α)=1/α$。在以上的条件下,可以定义拟三角Bernstein算子$B→T_{α, α}$,具体如下

$ \begin{array}{l} B\left( f \right) = f\left( 0 \right){T_0}\left( {t;\alpha } \right) + f\left( {t_\alpha ^ * } \right){T_1}\left( {t;\alpha } \right) + \\ f\left( {{\rm{ \mathsf{ π} /2}} - t_\alpha ^ * } \right){T_2}\left( {t;\alpha } \right) + f\left( {{\rm{ \mathsf{ π} /2}}} \right){T_3}\left( {t;\alpha } \right) \end{array} $ (14)

式中,$f∈ {\mathit{\boldsymbol{C}}} [0, {\rm{ \mathsf{ π} }} /2]$

明显有

$ B\left( {{f_0}} \right) = {f_0}\left( t \right),B\left( {{f_1}} \right) = {f_1}\left( {t;\alpha } \right) $

这意味着特征方程$f_{0}(t)$$f_{1}(t; α)$对应的特征值为$λ_{0}=λ_{1}=1$

对于任意的$α∈[1, +∞), t∈[0, {\rm{ \mathsf{ π} }} /2]$,定义函数

$ {f_2}\left( {t;\alpha } \right) = 1 - {T_0}\left( {t;\alpha } \right) - {T_3}\left( {t;\alpha } \right) $ (15)

该函数满足$f_{2}(0;α)=f_{2}({\rm{ \mathsf{ π} }} /2;α)=0$$f_{2}(t; α)= f_{2}({\rm{ \mathsf{ π} }} /2-t; α)$,因此有

$ \begin{array}{*{20}{c}} {B\left( {{f_2}} \right) = {f_2}\left( {t_\alpha ^ * ;\alpha } \right){T_1}\left( {t;\alpha } \right) + }\\ {{f_2}\left( {\left( {{\rm{ \mathsf{ π} /2}} - t_\alpha ^ * } \right);\alpha } \right){T_2}\left( {t;\alpha } \right) = }\\ {{f_2}\left( {t_\alpha ^ * ;\alpha } \right)\left[ {{T_1}\left( {t;\alpha } \right) + {T_2}\left( {t;\alpha } \right)} \right] = }\\ {{f_2}\left( {t_\alpha ^ * ;\alpha } \right){f_2}\left( {t;\alpha } \right)} \end{array} $

这表明$f_{2}(t; α)$是一个特征函数,则相应的特征值为$λ_{2, α}=f_{2}(t^{*}_{α}; α)$

此外,对于任意的$α∈[1, +∞), t∈[0, {\rm{ \mathsf{ π} }} /2]$,定义函数

$ {f_3}\left( {t;\alpha } \right) = {T_1}\left( {t;\alpha } \right) - {T_2}\left( {t;\alpha } \right) $ (16)

该函数满足$f_{3}(0;α)=f_{3}({\rm{ \mathsf{ π} }} /2;α)=0$以及$f_{3}(0;α)=-f_{3}({\rm{ \mathsf{ π} }} /2-t; α)$,因此有

$ \begin{array}{*{20}{c}} {B\left( {{f_3}} \right) = {f_3}\left( {t_\alpha ^ * ;\alpha } \right){T_1}\left( {t;\alpha } \right) + }\\ {{f_3}\left( {\left( {{\rm{ \mathsf{ π} /2}} - t_\alpha ^ * } \right);\alpha } \right){T_2}\left( {t;\alpha } \right) = }\\ {{f_3}\left( {t_\alpha ^ * ;\alpha } \right)\left[ {{T_1}\left( {t;\alpha } \right) - {T_2}\left( {t;\alpha } \right)} \right] = }\\ {{f_3}\left( {t_\alpha ^ * ;\alpha } \right){f_3}\left( {t;\alpha } \right)} \end{array} $

这表明$f_{3}(t; α)$也是一个特征函数,则相应的特征值为$λ_{3, α}=f_{3}(t^{*}_{α}; α)$

下面进一步说明$0≤λ_{3, α}≤λ_{2, α} <λ_{1}=λ_{0}=1$。首先,很明显$λ_{2, α}=f_{2}(t^{*}_{α}; α) <1$。其次由于$t^{*}_{α}∈[0, {\rm{ \mathsf{ π} }} /2]$,直接计算有

$ \begin{array}{l} {\lambda _{2,\alpha }} - {\lambda _{3,\lambda }} = {f_2}\left( {t_\alpha ^ * ;\alpha } \right) - {f_3}\left( {t_\alpha ^ * ;\alpha } \right) = \\ 1 - {T_0}\left( {t_\alpha ^ * ;\alpha } \right) - {T_3}\left( {t_\alpha ^ * ;\alpha } \right) - {T_1}\left( {t_\alpha ^ * ;\alpha } \right) + \\ {T_2}\left( {t_\alpha ^ * ;\alpha } \right) = 2{T_2}\left( {t_\alpha ^ * ;\alpha } \right) > 0 \end{array} $

最后,对于任意$α∈[1, +∞)$,从定理2可知,$(T_{0}(t; α), T_{1}(t; α), T_{2}(t; α), T_{3}(t; α))$形成函数空间$ {\mathit{\boldsymbol{T}}} _{α, α}$的最优规范全正基。由文献[17]可知,对于任意的$0≤t_{0} <t_{1} <t_{2}<t_{3}≤{\rm{ \mathsf{ π} }} /2$,配置矩阵$(T_{j}(t_{i}; α))_{0≤i, j≤3}$的子行列式均为非负的。因此,对于$t^{*}_{α}∈(0, {\rm{ \mathsf{ π} }} /2]$,有

$ \left| {\begin{array}{*{20}{c}} {{T_1}\left( {t_\alpha ^ * ;\alpha } \right)}&{{T_2}\left( {t_\alpha ^ * ;\alpha } \right)}\\ {{T_1}\left( {{\rm{ \mathsf{ π} }}/2;\alpha } \right)}&{{T_2}\left( {{\rm{ \mathsf{ π} }}/2;\alpha } \right)} \end{array}} \right| = {\lambda _{3,\lambda }} \ge 0 $

从上式可以得出$λ_{3, λ}≥0$。综上,给出了拟三角Bernstein算子在$α∈[1, +∞)$上的情形,$α∈(0, 1]$的情形可以类似讨论,只需做变换$α′=1/α$即可。

从上面的讨论中,可以发现$λ_{2, α}$是拟三角Bernstein算子$B$的第3大特征值,因此$λ_{2, α}$反映了生成曲线$T(t; α, α)$和它对应控制多边形的逼近程度。当$α=2$时,可得出$t^{*}_{2}={\rm{ \mathsf{ π} }} /2$, 因此$1-λ_{2, 2}=1-(1-2×(1- {\sqrt 2} /2)^{2}{\rm e}^{-2×({\sqrt 2} /2)})≈0.041 \;7$。从文献[38]中,可以发现经典的三次Bernstein算子的第3大特征值为$λ_{2, 2}≈0.666 \;7$,则第3大特征值和1的差值为$1-λ_{2, 2}≈0.333 \;3$,这也意味着当$α=2$时,生成的QCT-Bezier曲线比经典的三次Bézier曲线更逼近控制多边形。此外,有

$ \mathop {\lim }\limits_{\alpha \to + \infty } 1 - {\lambda _{2,\alpha }} = 0 $

因此,随着参数$α$的增大,曲线越来越逼近控制多边形,这也意味着随着参数的增大,生成的曲线也就越能够反映控制多边形的行为。图 6给出了不同参数下QCT-Bézier曲线和经典的三次Bézier曲线与控制多边形的逼近比较。

图 6 逼近比较
Fig. 6 Approximation comparison

2 三角域上拟三次三角Bernstein-Bézier基

本节将扩展QCT-Bernstein基(式(3))形成三角域上的拟三次三角Bernstein-Bézier(QTC-Bernstein-Bézier)基。

2.1 三角域上QCT-Bernstein-Bézier基的构造

定义2   对任意的$α, β, γ∈[0, +∞]$$D=\{(u, v, w) |u+v+w ={\rm{ \mathsf{ π} }} /2, u≥0, v≥0, w≥0\}$,称如下10个三角多项式为三角域$ \mathit{\boldsymbol{D}} $上带3个指数参数$α$, $β$$γ$的QCT-Bernstein-Bézier基函数

$ \left\{ \begin{array}{l} T_{3,0,0}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right) = {\left( {1 - \cos u} \right)^2}{{\rm{e}}^{ - \alpha \cos u}}\\ T_{0,3,0}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right) = {\left( {1 - \cos v} \right)^2}{{\rm{e}}^{ - \beta \cos v}}\\ T_{0,0,3}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right) = {\left( {1 - \cos w} \right)^2}{{\rm{e}}^{ - \gamma \cos w}}\\ T_{2,1,0}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right) = \cos w\sin v\left( {1 - \cos u} \right) \times \\ \frac{{1 + \cos u - \left( {1 - \cos u} \right){{\rm{e}}^{ - \alpha \cos u}}}}{{\cos u}}\\ T_{2,0,1}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right) = \cos v\sin w\left( {1 - \cos u} \right) \times \\ \frac{{1 + \cos u - \left( {1 - \cos u} \right){{\rm{e}}^{ - \alpha \cos u}}}}{{\cos u}}\\ T_{1,2,0}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right) = \cos w\sin u\left( {1 - \cos v} \right) \times \\ \frac{{1 + \cos v - \left( {1 - \cos v} \right){{\rm{e}}^{ - \beta \cos v}}}}{{\cos v}}\\ T_{0,2,1}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right) = \cos u\sin w\left( {1 - \cos v} \right) \times \\ \frac{{1 + \cos v - \left( {1 - \cos v} \right){{\rm{e}}^{ - \beta \cos v}}}}{{\cos v}}\\ T_{1,0,2}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right) = \cos v\sin u\left( {1 - \cos w} \right) \times \\ \frac{{1 + \cos w - \left( {1 - \cos w} \right){{\rm{e}}^{ - \gamma \cos w}}}}{{\cos w}}\\ T_{0,1,2}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right) = \cos u\sin v\left( {1 - \cos w} \right) \times \\ \frac{{1 + \cos w - \left( {1 - \cos w} \right){{\rm{e}}^{ - \gamma \cos w}}}}{{\cos w}}\\ T_{1,1,1}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right) = 1 - \sum\limits_{\begin{array}{*{20}{c}} {i + j + k = 3,}\\ {i \cdot j \cdot k \ne 1} \end{array}} {T_{i,j,k}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right)} \end{array} \right. $ (17)

注意,对任意的$α, β, γ∈[0, +∞)$,有

$ \begin{array}{l} \mathop {\lim }\limits_{u \to {\rm{ \mathsf{ π} }}/2} \frac{{1 + \cos u - \left( {1 - \cos u} \right){{\rm{e}}^{ - \alpha \cos u}}}}{{\cos u}} = \\ \mathop {\lim }\limits_{u \to {\rm{ \mathsf{ π} }}/2} \frac{{ - \sin u - \sin u\left[ {1 + \alpha \left( {1 - \cos u} \right)} \right]{{\rm{e}}^{ - \alpha \cos u}}}}{{ - \sin u}} = \\ \mathop {\lim }\limits_{u \to {\rm{ \mathsf{ π} }}/2} 1 + \left[ {1 + \alpha \left( {1 - \cos u} \right)} \right]{{\rm{e}}^{ - \alpha \cos u}} = \alpha + 2 \end{array} $

类似地,

$ \mathop {\lim }\limits_{u \to {\rm{ \mathsf{ π} }}/2} \frac{{1 + \cos v - \left( {1 - \cos v} \right){{\rm{e}}^{ - \beta \cos v}}}}{{\cos v}} = \beta + 2 $

$ \mathop {\lim }\limits_{u \to {\rm{ \mathsf{ π} }}/2} \frac{{1 + \cos w - \left( {1 - \cos w} \right){{\rm{e}}^{ - \gamma \cos w}}}}{{\cos w}} = \gamma + 2 $

因此对于不同的指数参数,在三角域$ \mathit{\boldsymbol{ D}} $上基函数式(17)均有意义。

引理1   对$u+v+w={\rm{ \mathsf{ π} }} /2$,下式成立

$ 1 - \left( {{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w} \right) = 2\sin u\sin v\sin w $

证明  对$u+v+w={\rm{ \mathsf{ π} }} /2$,直接计算得

$ \begin{array}{l} 1 - \left( {{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w} \right) = \\ \frac{1}{2}\left( {\cos 2u + \cos 2v + \cos 2w - 1} \right) = \\ \cos \left( {u + v} \right)\cos \left( {u - v} \right) - {\sin ^2}w = \\ \cos \left( {u - v} \right)\sin w - \cos \left( {u + v} \right)\sin w = \\ \left[ {\cos \left( {u - v} \right) - \cos \left( {u + v} \right)} \right]\sin w = \\ 2\sin u\sin v\sin w \end{array} $

证毕。

2.2 三角域上QCT-Bernstein-Bézier基的性质

定理4   三角域上QCT-Bernstein-Bézier基具有如下性质

1) 单位性:$\sum\limits_{i + j + k = 3} { T ^{3}_{i, j, l}(u, v, w; α, β, γ)} =1$

2) 非负性:对于任意的$i, j, k∈ {\mathbf N}, i+j+k=3$,有$T ^{3}_{i, j, k}(u, v, w; α, β, γ)≥0$

3) 对称性:

$ \begin{array}{l} T_{i,j,k}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right) = T_{j,i,k}^3\left( {v,u,w;\beta ,\alpha ,\gamma } \right) = \\ T_{j,k,i}^3\left( {v,w,u;\beta ,\gamma ,\alpha } \right) = T_{i,k,j}^3\left( {u,w,v;\alpha ,\gamma ,\beta } \right) = \\ T_{k,i,j}^3\left( {w,u,v;\gamma ,\alpha ,\beta } \right) = T_{i,j,k}^3\left( {w,v,u;\gamma ,\beta ,\alpha } \right) \end{array} $

4) 边界性:当3个变量$α, β$$γ$中的其中之一取值为0时,基函数式(17)退化为相应的带2个指数参数的QCT-Bernstein基函数,如式(3)。

5) 线性无关性:$\{T ^{3}_{i, j, k}(u, v, w; α, β, γ), i+j+k=3\}$线性无关。

证明  下面证明性质2)和5)。其余性质容易证明。

性质2),对任意$α, β, γ≥0, i, j, k∈{\mathbf N}, i+j+k=3$$i×j×k≠1$,显然有

$ T_{i,j,k}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right) \ge 0 $

此外,对于$T ^{3}_{1, 1, 1}(u, v, w; α, β, γ)≥0$,由引理1可得性质5),对任意$α, β, γ≥0, ξ_{i, j, k}∈{\bf R}, (i, j, k∈ {\mathbf N}, i+j+k=3)$。考虑线性组合

$ \sum\limits_{i + j + k = 3} {{\lambda _{i,j,k}}T_{i,j,k}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right)} = 0 $ (18)

$w=0$,有

$ \sum\limits_{i = 0}^3 {{\xi _{i,\left( {3 - i} \right),0}}{T_i}\left( {u;\alpha ,\beta } \right)} = 0 $ (19)

因此,由QCT-Bézier基的线性无关性,可得$ξ_{i, (3-i), 0}= 0(i=0, 1, 2, 3)$。类似可得$ξ_{i, 0, (3-i)}=$ $ξ_{0, i, (3-i)}=0(i=0, 1, 2, 3)$。最后可得$ξ_{1, 1, 1}=0$。证毕。

图 7给出了部分QCT-Bernstein-Bézier基的图形,指数参数值为$α=β=γ=2$

图 7 三角域上部分QCT-Bernstein-Bézier基的图形
Fig. 7 Some plots of QCT-Bernstein-Bézier basis functions

2.3 三角域上QCT-Bernstein-Bézier曲面

定义3   对任意的实数$α, β, γ∈[0,+∞)$$ { \mathit{\boldsymbol{D}}} =\{(u, v, w)|u+v+w={\rm{ \mathsf{ π} }} /2, u≥0, v≥0, w≥0\}$,给定控制顶点$P_{i, j, k}∈ {\bf R} ^{3}(i, j, k∈ {\bf N}, i+j+k=3)$,称

$ \begin{array}{*{20}{c}} {R\left( {u,v,w} \right) = \sum\limits_{i + j + k = 3} {T_{i,j,k}^3\left( {u,v,w;\alpha ,\beta ,\gamma } \right)} }\\ {{P_{i,j,k}},\left( {u,v,w} \right) \in \mathit{\boldsymbol{D}}} \end{array} $ (20)

为三角域上的QCT-Bernstein-Bézier曲面。

由三角域上QCT-Bernstein-Bézier基的性质,易推出相应曲面的性质:

1) 仿射不变性和凸包性。由于基函数具有单位性和非负性,故相应的曲面具有仿射不变性和凸包性。

2) 插值角点性质。直接计算可得

$ \begin{array}{*{20}{c}} {R\left( {{\rm{ \mathsf{ π} }}/2,0,0} \right) = {P_{3,0,0}},R\left( {0,{\rm{ \mathsf{ π} }}/2,0} \right) = {P_{0,3,0}},}\\ {R\left( {0,0,{\rm{ \mathsf{ π} }}/2} \right) = {P_{0,0,3}}} \end{array} $

3) 角点切平面性质。令$w={\rm{ \mathsf{ π} }} /2-u-v$,可得

$ \frac{{{\rm{d}}R\left( {u,v,w} \right)}}{{{\rm{d}}u}}\left| {_{\left( {{\rm{ \mathsf{ π} }}/2,0,0} \right)}} \right. = \left( {2 + \alpha } \right)\left( {{P_{3,0,0}} - {P_{2,0,1}}} \right) $

$ \frac{{{\rm{d}}R\left( {u,v,w} \right)}}{{{\rm{d}}v}}\left| {_{\left( {{\rm{ \mathsf{ π} }}/2,0,0} \right)}} \right. = \left( {2 + \alpha } \right)\left( {{P_{2,1,0}} - {P_{2,0,1}}} \right) $

$ \frac{{{\rm{d}}R\left( {u,v,w} \right)}}{{{\rm{d}}u}}\left| {_{\left( {0,{\rm{ \mathsf{ π} }}/2,0} \right)}} \right. = \left( {2 + \beta } \right)\left( {{P_{1,2,0}} - {P_{0,2,1}}} \right) $

$ \frac{{{\rm{d}}R\left( {u,v,w} \right)}}{{{\rm{d}}v}}\left| {_{\left( {0,{\rm{ \mathsf{ π} }}/2,0} \right)}} \right. = \left( {2 + \beta } \right)\left( {{P_{0,3,0}} - {P_{0,2,1}}} \right) $

$ \frac{{{\rm{d}}R\left( {u,v,w} \right)}}{{{\rm{d}}u}}\left| {_{\left( {0,0,{\rm{ \mathsf{ π} }}/2} \right)}} \right. = \left( {2 + \gamma } \right)\left( {{P_{1,0,2}} - {P_{0,0,3}}} \right) $

$ \frac{{{\rm{d}}R\left( {u,v,w} \right)}}{{{\rm{d}}v}}\left| {_{\left( {0,0,{\rm{ \mathsf{ π} }}/2} \right)}} \right. = \left( {2 + \gamma } \right)\left( {{P_{0,1,2}} - {P_{0,0,3}}} \right) $

这表明QCT-Bernstein-Bézier曲面在3个角点$({\rm{ \mathsf{ π} }} /2, 0, 0), (0, {\rm{ \mathsf{ π} }} /2, 0), (0, 0, {\rm{ \mathsf{ π} }} /2)$的切平面分别由控制点$P_{3, 0, 0}, P_{2, 1, 0}, P_{2, 1, 0};P_{0, 3, 0}, P_{1, 2, 0}, P_{0, 2, 1};P_{0, 0, 3}, P_{1, 0, 2}, P_{0, 1, 2}$生成。

4) 边界性质。当$w=0$$R(u, v, w)$退化为式(21)定义的带指数参数$α, β$的QCT-Bézier曲线

$ R\left( {u,v,0} \right) = \sum\limits_{i = 0}^3 {{P_{i,3 - i,0}}{A_i}\left( {u;\alpha ,\beta } \right)} $ (21)

类似地,$R(0, v, w)$为带指数参数$β, γ$的QCT-Bézier曲线,而$R(u, 0, w)$为带指数参数$α, γ$的QCT-Bézier曲线。因此,QCT-Bernstein-Bézier曲面的边界曲线也可精确表示椭圆弧和抛物线弧。图 8展示了边界分别为抛物线弧、椭圆弧和圆弧的QCT-Bernstein-Bézier曲面。其中参数为$α=β=γ=0$,控制点为$\{P_{3, 0, 0}=(0, -4, 0), P_{0, 3, 0}=(2, 0, 0)$, $P_{0, 0, 3}=(0, 0, 2), P_{2, 1, 0}=(1,-4, 0)$, $P_{2, 0, 1}=(0, -4, 1), P_{1, 2, 0}=(2, 0, 0), P_{0, 2, 1}=(2, 0, 1)$, $P_{1, 0, 2}=(1, 0, 2), P_{0, 1, 2}=(0,-2, 2), P_{1, 1, 1}=(1, -2, 1)\}$。直接计算可得相应边界曲线的参数方程,分别为$u=0, v=-4\sin x, w=2\cos x; u=2\sin x, v=0, w=2\cos x; $ $u=2\cos x, v=-4+4\cos^{2}x, w=0$, 其中$x∈[0, {\rm{ \mathsf{ π} }} /2]$

图 8 边界为圆弧、椭圆弧以及抛物线弧的QCT-Bernstein-Bézier曲面
Fig. 8 QCT-Bernstein-Bézier patches whose boundaries are arcs of elliptic, circle or parabola

5) 形状调整性质。由于QCT-Bernstein-Bézier曲面式(20)含有3个指数参数$α, β, γ$,从而在控制网格固定时,通过改变$α, β, γ$的值可对曲面进行形状调整。当增大$α, β, γ$的值时,曲面将向控制网格逼近,因此$α, β, γ$具有张力效果。此外,由曲面的边界性质易得,每条边界曲线$R(0, v, w)$, $R(u, 0, w)$$R(u, v, 0)$只和其中2个参数有关,和另外一个参数无关。这表明某一个参数的改变,只能够影响其中的2条边界曲线的形状。图 9给出了不同指数参数对QCT-Bernstein-Bézier曲面的影响。

图 9 三角域上QCT-Bernstein-Bézier曲面
Fig. 9 QCT-Bernstein-Bézier patches over triangular domain

2.4 de Casteljau-type算法

下面给出一种生成QCT-Bernstein-Bézier曲面高效稳定的de-Casteljau-type算法。对任意$(u, v, w)∈ \mathit{\boldsymbol{ D }}$,记

$ \left\{ \begin{array}{l} {f_1}\left( {u,v,w} \right) = \frac{{\sin u\cos w\left( {{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w} \right)}}{{\cos w\left( {\sin u + \sin v} \right)\left( {{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w} \right) + \sin w\left( {{{\sin }^2}u + {{\sin }^2}v} \right)}}\\ {f_2}\left( {u,v,w} \right) = \frac{{\sin u\cos w\left( {{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w} \right)}}{{\cos w\left( {\sin u + \sin v} \right)\left( {{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w} \right) + \sin w\left( {{{\sin }^2}u + {{\sin }^2}v} \right)}}\\ {f_3}\left( {u,v,w} \right) = \frac{{\sin w\left( {{{\sin }^2}u + {{\sin }^2}v} \right)}}{{\cos w\left( {\sin u + \sin v} \right)\left( {{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w} \right) + \sin w\left( {{{\sin }^2}u + {{\sin }^2}v} \right)}} \end{array} \right. $

式中,

$ \left\{ \begin{array}{l} {g_1}\left( {u,v,w} \right) = \left( {1 - \cos u} \right)\left( {{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w} \right)\\ {g_2}\left( {u,v,w} \right) = \sin v\cos w\left( {{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w} \right) + \\ \sin u\sin v\sin w\\ {g_3}\left( {u,v,w} \right) = \cos v\sin w\left( {{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w} \right) + \\ \sin u\sin v\sin w \end{array} \right. $

$ \left\{ \begin{array}{l} P_{2,0,0}^1 = \\ \frac{{{{\rm{e}}^{ - \alpha \cos u}}}}{{1 + \cos u}}{P_{3,0,0}} + \frac{{\left( {1 + \cos u - {{\rm{e}}^{ - \alpha \cos u}}} \right)\sin v\cos w}}{{\left( {1 + \cos u} \right)\cos u}}{P_{2,1,0}} + \\ \frac{{\left( {1 + \cos u - {{\rm{e}}^{ - \alpha \cos u}}} \right)\sin w\cos v}}{{\left( {1 + \cos u} \right)\cos u}}{P_{2,0,1}}\\ P_{0,2,0}^1 = \\ \frac{{{{\rm{e}}^{ - \beta \cos v}}}}{{1 + \cos v}}{P_{0,3,0}} + \frac{{\left( {1 + \cos v - {{\rm{e}}^{ - \beta \cos v}}} \right)\sin u\cos w}}{{\left( {1 + \cos v} \right)\cos v}}{P_{1,2,0}} + \\ \frac{{\left( {1 + \cos v - {{\rm{e}}^{ - \beta \cos v}}} \right)\sin w\cos u}}{{\left( {1 + \cos v} \right)\cos v}}{P_{0,2,1}}\\ P_{2,0,0}^1 = \\ \frac{{{{\rm{e}}^{ - \gamma \cos w}}}}{{1 + \cos w}}{P_{0,0,3}} + \frac{{\left( {1 + \cos w - {{\rm{e}}^{ - \gamma \cos w}}} \right)\sin v\cos u}}{{\left( {1 + \cos w} \right)\cos w}}{P_{1,0,2}} + \\ \frac{{\left( {1 + \cos w - {{\rm{e}}^{ - \gamma \cos w}}} \right)\sin v\cos u}}{{\left( {1 + \cos w} \right)\cos w}}{P_{0,1,2}} \end{array} \right. $

则可以改写QCT-Bernstein-Bézier曲面的表达式为

$ \begin{array}{*{20}{c}} {R\left( {u,v,w} \right) = }\\ {\frac{{1 - {{\cos }^2}u}}{{{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w}}\left[ {{g_1}\left( {u,v,w} \right)P_{2,0,0}^1 + } \right.}\\ {\left. {{g_2}\left( {u,v,w} \right)P_{1,1,0}^1 + {g_3}\left( {u,v,w} \right)P_{1,0,1}^1} \right] + }\\ {\frac{{1 - {{\cos }^2}v}}{{{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w}}\left[ {{g_1}\left( {v,u,w} \right)P_{0,2,0}^1 + } \right.}\\ {\left. {{g_2}\left( {u,v,w} \right)P_{1,1,0}^1 + {g_3}\left( {u,v,w} \right)P_{0,1,1}^1} \right] + }\\ {\frac{{1 - {{\cos }^2}w}}{{{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w}}\left[ {{g_1}\left( {w,v,u} \right)P_{0,0,2}^1 + } \right.}\\ {\left. {{g_2}\left( {w,u,v} \right)P_{0,1,1}^1 + {g_3}\left( {w,v,u} \right)P_{1,0,1}^1} \right]} \end{array} $ (22)

再记

$ \begin{array}{l} P_{1,0,0}^2 = {g_1}\left( {u,v,w} \right)P_{2,0,0}^1 + \\ {g_2}\left( {u,v,w} \right)P_{1,1,0}^1 + {g_3}\left( {u,v,w} \right)P_{1,0,1}^1 \end{array} $

$ \begin{array}{l} P_{0,1,0}^2 = {g_2}\left( {v,u,w} \right)P_{1,1,0}^1 + \\ {g_1}\left( {v,u,w} \right)P_{0,2,0}^1 + {g_3}\left( {v,u,w} \right)P_{0,1,1}^1 \end{array} $

$ \begin{array}{l} P_{0,0,1}^2 = {g_3}\left( {w,v,u} \right)P_{1,0,1}^1 + \\ {g_2}\left( {w,v,u} \right)P_{0,1,1}^1 + {g_1}\left( {w,v,u} \right)P_{0,0,2}^1 \end{array} $

则可进一步改写QCT-Bernstein-Bézier曲面的表达式为

$ \begin{array}{l} R\left( {v,u,w} \right) = \frac{{1 - {{\cos }^2}u}}{{{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w}}P_{1,0,0}^2 + \\ \frac{{1 - {{\cos }^2}v}}{{{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w}}P_{0,1,0}^2 + \\ \frac{{1 - {{\cos }^2}w}}{{{{\sin }^2}u + {{\sin }^2}v + {{\sin }^2}w}}P_{0,0,1}^2 = P_{0,0,0}^2 \end{array} $ (23)

对任意$u+v+w={\rm{ \mathsf{ π} }} /2$,容易验证$f_{1}(u, v, w)+$ $f_{2}(u, v, w)+f_{3}(u, v, w)=1$$g_{1}(u, v, w)+g_{2}(u, v, w)$$+g_{3}(u, v, w)=1$(利用引理1)。

2.5 曲面的拼接

定义两张QCT-Bernstein-Bézier曲面分别为

$ \begin{array}{*{20}{c}} {{R_1}\left( {v,u,w} \right) = \sum\limits_{i + j + k = 3} {T_{i,j,k}^3\left( {u,v,w;{\alpha _1},\beta ,\gamma } \right)} }\\ {{P_{i,j,k}},\left( {u,v,w} \right) \in \mathit{\boldsymbol{D}}} \end{array} $ (24)

$ \begin{array}{*{20}{c}} {{R_2}\left( {u,v,w} \right) = \sum\limits_{i + j + k = 3} {T_{i,j,k}^3\left( {u,v,w;{\alpha _2},\beta ,\gamma } \right)} }\\ {{Q_{i,j,k}},\left( {u,v,w} \right) \in \mathit{\boldsymbol{D}}} \end{array} $ (25)

显然,若控制点满足

$ {P_{0,j,k}} = {Q_{0,j,k}}\;\;\;j,k \in {\bf{N}},j + k = 3 $ (26)

则曲面式(24)和式(25)存在公共的边界:$R_{1}(0, v, w)=R_{2}(0, v, w), v+w={\rm{ \mathsf{ π} }} /2$。这时两张曲面形成一张$\rm C^{0}$连续曲面。

对公共边界曲线$R_{1}(0, v, {\rm{ \mathsf{ π} }} /2-v)$关于变量$v$求导,可得

$ \begin{array}{l} \frac{{{\rm{d}}{R_1}\left( {0,v,{\rm{ \mathsf{ π} }}/2 - v} \right)}}{{{\rm{d}}v}} = \\ \left[ { - {{\rm{e}}^{ - \beta \cos t}}\left( {1 - \cos t} \right)\sin t\left( { - 2 - \beta + \beta \cos t} \right)} \right] \times \\ \left( {{P_{0,3,0}} - {P_{0,2,1}}} \right) + 2\sin v\cos v\left( {{P_{0,2,1}} - {P_{0,1,2}}} \right) + \\ \left[ {{{\rm{e}}^{ - \gamma \sin t}}\left( {1 - \sin t} \right)\cos t\left( { - 2 - \gamma + \gamma \sin t} \right)} \right] \times \\ \left( {{P_{0,1,2}} - {P_{0,0,3}}} \right) \end{array} $ (27)

对曲面$R_{1}(u, v, {\rm{ \mathsf{ π} }} /2-u-v)$$R_{2}(u, v, {\rm{ \mathsf{ π} }} /2-u-v)$分别关于变量$u$求导,可得

$ \begin{array}{l} \frac{{{\rm{d}}{R_1}\left( {0,v,{\rm{ \mathsf{ π} }}/2 - u - v} \right)}}{{{\rm{d}}v}}\left| {_{u = 0}} \right. = \\ \left[ { - {{\rm{e}}^{ - \beta \cos t}}\left( {1 - \cos t} \right)\sin t\left( { - 2 - \beta + \beta \cos t} \right)} \right] \times \\ \left( {{P_{1,2,0}} - {P_{0,2,1}}} \right) + 2\sin v\cos v\left( {{P_{1,1,1}} - {P_{0,1,2}}} \right) + \\ \left[ {{{\rm{e}}^{ - \gamma \sin t}}\left( {1 - \sin t} \right)\cos t\left( { - 2 - \gamma + \gamma \sin t} \right)} \right] \times \\ \left( {{P_{1,0,2}} - {P_{0,0,3}}} \right) \end{array} $ (28)

$ \begin{array}{l} \frac{{{\rm{d}}{R_2}\left( {0,v,{\rm{ \mathsf{ π} }}/2 - u - v} \right)}}{{{\rm{d}}u}}\left| {_{u = 0}} \right. = \\ \left[ { - {{\rm{e}}^{ - \beta \cos t}}\left( {1 - \cos t} \right)\sin t\left( { - 2 - \beta + \beta \cos t} \right)} \right] \times \\ \left( {{Q_{1,2,0}} - {Q_{0,2,1}}} \right) + 2\sin v\cos v\left( {{Q_{1,1,1}} - {Q_{0,1,2}}} \right) + \\ \left[ {{{\rm{e}}^{ - \gamma \sin t}}\left( {1 - \sin t} \right)\cos t\left( { - 2 - \gamma + \gamma \sin t} \right)} \right] \times \\ \left( {{Q_{1,0,2}} - {Q_{0,0,3}}} \right) \end{array} $ (29)

两张QCT-Bernstein-Bézier曲面$R_{1}(u, v, w)$$R_{2}(u, v, w)$$G_{1}$连续条件是式(27)、(28)和(29)定义的向量对任意变量$v$均共线[17],具体表示为

$ \begin{array}{*{20}{c}} {\frac{{{\rm{d}}{R_2}\left( {u,v,{\rm{ \mathsf{ π} }}/2 - u - v} \right)}}{{{\rm{d}}u}}\left| {_{u = 0}} \right. = \varphi \frac{{{\rm{d}}{R_1}\left( {0,v,{\rm{ \mathsf{ π} }}/2 - v} \right)}}{{{\rm{d}}v}} + }\\ {\phi \frac{{{\rm{d}}{R_1}\left( {u,v,{\rm{ \mathsf{ π} }}/2 - u - v} \right)}}{{{\rm{d}}u}}\left| {_{u = 0}} \right.} \end{array} $

式中,$φ$$\phi $是两个任意常数。因此得到如下条件

$ \left\{ \begin{array}{l} {Q_{1,2,0}} - {Q_{0,2,1}} = \varphi \left( {{P_{0,3,0}} - {P_{0,2,1}}} \right) + \phi \left( {{P_{1,2,0}} - {P_{0,2,1}}} \right)\\ {Q_{1,1,1}} - {Q_{0,1,2}} = \varphi \left( {{P_{0,2,1}} - {P_{0,1,2}}} \right) + \phi \left( {{P_{1,1,1}} - {P_{0,1,2}}} \right)\\ {Q_{1,0,2}} - {Q_{0,0,3}} = \varphi \left( {{P_{0,1,2}} - {P_{0,0,3}}} \right) + \phi \left( {{P_{1,0,2}} - {P_{0,0,3}}} \right) \end{array} \right. $ (30)

总结上述讨论,可以得到定理5。

定理5   对任意$α_{l}, β, γ∈[0, +∞], l=1, 2$,若控制网格满足条件式(26)和式(30),则曲面$R_{1}(u, v, w)$$R_{2}(u, v, w)$满足$G_{1}$连续条件。

由定理5可知,曲面式(24)和式(25)的$ \rm G_{1}$连续拼接条件与两张Bernstein-Bézier曲面的$ \rm G_{1}$连续拼接条件相似[39]。不同的是,可以通过调整两张三角多项式曲面的形状参数值得到不同的$ \rm G_{1}$连续曲面。图 10给出了不同参数值下的$ \rm G_{1}$连续曲面。

图 10 不同参数下的曲面拼接
Fig. 10 The connecting of patches with different shape parameters

3 结论

本文提出了一种具有指数参数的三角基函数进行曲线曲面构造,该参数具有张力作用效果,随着参数的增大,生成的曲线曲面会越来越逼近控制网格,从而能够更加准确地描述控制网格的行为。在光滑连续方面,本文给出了QCT-Bézier曲线的$\rm C^{1}$$\rm C^{2}$连续条件以及拼接两张三角域QCT-Bernstein-Bézier曲面的$\rm G^{1}$条件;在特殊曲线的描述方面,提出的QCT-Bézier曲线能够精确地描述椭圆弧以及抛物线弧,进而相应的三角域QCT-Bernstein-Bézier曲面能够生成具有边界为椭圆弧以及抛物线弧的曲面;在曲线曲面的高效稳定计算方面,本文给出了生成QCT-Bézier曲线的割角算法以及三角域上计算QCT-Bernstein-Bézier曲面的de-Casteljau-type算法,同时还提出了一种估计QCT-Bézier曲线和控制多边形逼近程度的拟三角Bernstein算子。综上,本文提出的基于带指数参数的曲线曲面构造方法保留了传统方法所有的优良性质,而且具有优良的形状可调性和逼近性,是一种有效实用的方法,适合于曲线曲面设计。实际上,构造的曲线曲面具有良好的性质,但同样还有很多问题需要解决,例如,曲线、曲面的形状分析,曲线向B样条曲线、曲面的扩展方法研究等,限于篇幅,此类讨论将另文叙述。

参考文献

  • [1] Farin G. Curves and Surfaces for CAGD[M]. 5th ed. San Francisco: Morgan Kaufmann, 2001.
  • [2] Hoschek J, Lasser D. Fundamentals of Computer Aided Geometric Design[M]. Natick, MA: AK Peters, 1993.
  • [3] Piegl L, Tiller W. The NURBS Book[M]. New York: Springer, 1995.
  • [4] Wang W T, Wang G Z. Bézier curves with shape parameter[J]. Journal of Zhejiang University-Science A, 2005, 6(6): 497–501. [DOI:10.1631/jzus.2005.A0497]
  • [5] Chen J, Wang G J. A new type of the generalized Bézier curves[J]. Applied Mathematics-A Journal of Chinese Universities, 2011, 26(1): 47–56. [DOI:10.1007/s11766-011-2453-8]
  • [6] Qin X Q, Hu G, Zhang N J, et al. A novel extension to the polynomial basis functions describing Bézier curves and surfaces of degree n with multiple shape parameters[J]. Applied Mathematics and Computation, 2013, 223: 1–16. [DOI:10.1016/j.amc.2013.07.073]
  • [7] Xu G, Wang G Z. Extended cubic uniform B-spline and α-B-spline[J]. Acta Automatica Sinica, 2008, 34(8): 980–984. [DOI:10.1016/S1874-1029(08)60047-6]
  • [8] Chen Q Y, Wang G Z. A class of Bézier-like curves[J]. Computer Aided Geometric Design, 2003, 20(1): 29–39. [DOI:10.1016/S0167-8396(03)00003-7]
  • [9] Han X L, Zhu Y P. Curve construction based on five trigonometric blending functions[J]. BIT Numerical Mathematics, 2012, 52(4): 953–979. [DOI:10.1007/s10543-012-0386-0]
  • [10] Su B Y, Tan J Q. A family of quasi-cubic blended splines and applications[J]. Journal of Zhejiang University- Science A, 2006, 7(9): 1550–1560. [DOI:10.1631/jzus.2006.A1550]
  • [11] Zhang J X, Tan J Q. Extensions of hyperbolic Bézier curves[J]. Journal of Engineering Graphics, 2011, 32(1): 31–38. [张锦秀, 檀结庆. 代数双曲Bézier曲线的扩展[J]. 工程图学学报, 2011, 32(1): 31–38. ] [DOI:10.3969/j.issn.1003-0158.2011.01.007]
  • [12] Wu H Y, Zuo H. Quadratic non-uniform hyperbolic B-spline curves with multiple shape parameters[J]. Journal of Computer-Aided Design & Computer Graphics, 2007, 19(7): 876–883. [邬弘毅, 左华. 多形状参数的二次非均匀双曲B-样条曲线[J]. 计算机辅助设计与图形学学报, 2007, 19(7): 876–883. ] [DOI:10.3321/j.issn:1003-9775.2007.07.011]
  • [13] Han X L, Zhu Y P. Total positivity of the cubic trigonometric Bézier basis[J]. Journal of Applied Mathematics, 2014, 2014: #198745. [DOI:10.1155/2014/198745]
  • [14] Wang K, Zhang G C, Gong J H. Constructing trigonometric polynomial curves and surfaces with two parameters[J]. Journal of Image and Graphics, 2018, 23(12): 1910–1924. [汪凯, 张贵仓, 龚进慧. 带两个参数的三角多项式曲线曲面构造[J]. 中国图象图形学报, 2018, 23(12): 1910–1924. ] [DOI:10.11834/jig.180328]
  • [15] Kvasov B I. Methods of Shape-Preserving Spline Approximation[M]. Singapore: World Scientific Publishing, 2000.
  • [16] Carnicer J M, Peña J M. Shape preserving representations and optimality of the Bernstein basis[J]. Advances in Computational Mathematics, 1993, 1(2): 173–196. [DOI:10.1007/BF02071384]
  • [17] Carnicer J M, Peña J M. Totally positive bases for shape preserving curve design and optimality of B-splines[J]. Computer Aided Geometric Design, 1994, 11(6): 633–654. [DOI:10.1016/0167-8396(94)90056-6]
  • [18] Cao J, Wang G Z. An extension of Bernstein-Bézier surface over the triangular domain[J]. Progress in Natural Science, 2007, 17(3): 352–357. [DOI:10.1080/10020070612331343269]
  • [19] Shen W Q, Wang G Z. Triangular domain extension of linear Bernstein-like trigonometric polynomial basis[J]. Journal of Zhejiang University-Science C, 2010, 11(5): 356–364. [DOI:10.1631/jzus.C0910347]
  • [20] Wei Y W, Shen W Q, Wang G Z. Triangular domain extension of algebraic trigonometric Bézier-like basis[J]. Applied Mathematics-A Journal of Chinese Universities, 2011, 26(2): 151–160. [DOI:10.1007/s11766-011-2672-z]
  • [21] Yang L Q, Zeng X M. Bézier curves and surfaces with shape parameters[J]. International Journal of Computer Mathematics, 2009, 86(7): 1253–1263. [DOI:10.1080/0020716070182-1715]
  • [22] Yan L L, Liang J F. An extension of the Bézier model[J]. Applied Mathematics and Computation, 2011, 218(6): 2863–2879. [DOI:10.1016/j.amc.2011.08.030]
  • [23] Zhu Y P, Han X L. A class of αβγ Bernstein-Bézier basis functions over triangular domain[J]. Applied Mathematics and Computation, 2013, 220: 446–454. [DOI:10.1016/j.amc.2013.06.043]
  • [24] Zhang J W. C-curves:an extension of cubic curves[J]. Computer Aided Geometric Design, 1996, 13(3): 199–217. [DOI:10.1016/0167-8396(95)00022-4]
  • [25] Mazure M L. Which spaces for design?[J]. Numerische Mathematik, 2008, 110(3): 357–392. [DOI:10.1007/s00211-008-0164-8]
  • [26] Mazure M L. Quasi-Chebyshev splines with connection matrices:application to variable degree polynomial splines[J]. Computer Aided Geometric Design, 2001, 18(3): 287–298. [DOI:10.1016/S0167-8396(01)00031-0]
  • [27] Mazure M L. On dimension elevation in quasi extended Chebyshev spaces[J]. Numerische Mathematik, 2008, 109(3): 459–475. [DOI:10.1007/s00211-007-0133-7]
  • [28] Mazure M L. On a general new class of quasi Chebyshevian splines[J]. Numerical Algorithms, 2011, 58(3): 399–438. [DOI:10.1007/s11075-011-9461-x]
  • [29] Mazure M L. Quasi Extended Chebyshev spaces and weight functions[J]. Numerische Mathematik, 2011, 118(1): 79–108. [DOI:10.1007/s00211-010-0312-9]
  • [30] Bosner T, Rogina M. Variable degree polynomial splines are Chebyshev splines[J]. Advances in Computational Mathematics, 2013, 38(2): 383–400. [DOI:10.1007/s10444-011-9242-z]
  • [31] Ibraheem F, Hussain M, Hussain M Z. Positive data visualization using trigonometric function[J]. Journal of Applied Mathematics, 2012, 2012: #247120. [DOI:10.1155/2012/247120]
  • [32] Hussain M, Saleem S. C1 rational quadratic trigonometric spline[J]. Egyptian Informatics Journal, 2013, 14(3): 211–220. [DOI:10.1016/j.eij.2013.09.002]
  • [33] Hussain M Z, Hussain M, Waseem A. Shape-preserving trigonometric functions[J]. Computational and Applied Mathematics, 2014, 33(2): 411–431. [DOI:10.1007/s40314-013-0071-1]
  • [34] Ibraheem F, Hussain M, Hussain M Z. Monotone data visualization using rational trigonometric spline interpolation[J]. The Scientific World Journal, 2014, 2014: #602453. [DOI:10.1155/2014/602453]
  • [35] Zhao K. On the extension of the second main theorem of Nevanlinna[J]. Journal of Hunan Agricultural University, 1997(03): 275–279. [赵坤. Nevanlinna第二基本定理的推广(英文)[J]. 湖南农业大学学报, 1997(03): 275–279. ]
  • [36] Zhu Y P. Research on theory and methods for geometric modeling based on basis functions possessing shape parameters[D]. Changsha: Central South University, 2014. [朱远鹏.基函数中带形状参数的几何造型理论与方法研究[D].长沙: 中南大学, 2014.] http://cdmd.cnki.com.cn/Article/CDMD-10533-1014403350.htm
  • [37] Carnicer J M, Mainar E, Peña J M. A Bernstein-like operator for a mixed algebraic-trigonometric space[C]. XXI Congreso de Ecuacuines Difernciales y Aplicaciones, XI Congreso de Matematica Aplicada, Ciudad Real, 2009: 1-7. https://www.researchgate.net/publication/242271432_A_Bernstein-like_operator_for_a_mixed_algebraic-trigonometric_space
  • [38] Cooper S, Waldron S. The eigenstructure of the Bernstein operator[J]. Journal of Approximation Theory, 2000, 105(1): 133–165. [DOI:10.1006/jath.2000.3464]
  • [39] Salomon D. The Computer Graphics Manual[M]. New York: Springer, 2011: 719-721.