ICode9

精准搜索请尝试: 精确搜索
首页 > 其他分享> 文章详细

散点拟合圆

2022-05-29 19:04:06  阅读:190  来源: 互联网

标签:begin end align 散点 tag 拟合 aX sum


最近工作中遇到一个拟合圆的问题,通过找到的轮廓点(存在缺失的情况),需要找出圆心及半径。这里采用最小二乘法进行拟合,并记录一下具体的推导过程。最小二乘法是解决曲线拟合问题最常用的方法,其基本思路是:令

\[f(x) = \alpha_1 \phi_1(x) + \alpha_2 \phi_2(x) + ... + \alpha_m \phi_m(x) \]

其中\(\phi_k(x)\)是事先选择的一组线性无关函数,\(\alpha_k\)是待定系数 ,拟合准则是使\(y_i(i = 1, 2, ..., n)\)与\(f(x_i)\)的距离\(\delta_i\)的平方和最小。注意:最小二乘拟合对于离群点很敏感,如果存在离群点可以采用RANSAC(Random Sample Consensus)算法进行拟合

最小二乘法拟合圆推导流程

1、确定待定系数

设圆的圆心为\((A, B)\),半径为\(R\),则圆的方程可以写成

\[ \begin{align} R^2 &= (x-A)^2 + (y - B)^2 \tag{1}\\ &= x^2 - 2Ax + A^2 + y^2 - 2By + B^2 \tag{2} \end{align} \]

令\(a = -2A, b = -2B, c = A^2 + B^2 - R^2\),则圆的曲线方程可写成(3)式:

\[ \begin{align} x^2 + y^2 +ax + by + c = 0 \tag{3} \end{align} \]

只需要求出参数\(a, b, c\)即可得到圆心及半径:

\[ \begin{align} A &= -\frac{a}{2} \tag{4}\\ B &= -\frac{b}{2} \tag{5}\\ R &= \frac{1}{2}\sqrt{a^2 + b^2 - 4c} \tag{6} \end{align} \]

所以这里待定系数为\(a, b, c\)。

2、确定距离和

样本集\((X_i, Y_i), i = {1, 2, 3, ..., N}\)中第\(i\)个样本到圆心的距离记为\(d_i\),

\[{d_i}^2 = (X_i - A)^2 + (Y_i - B)^2 \tag{7} \]

则其与半径之间的误差为:

\[\delta_i = {d_i}^2 - R^2 = {X_i}^2 + {Y_i}^2 + aX_i + bY_i + c \tag{8} \]

所以所有样本的误差和\(Q(a, b, c)\)为

\[ \begin{align} Q(a, b, c) &= \sum_{i = 1}^{N}{\delta_i}^2 \tag{9} \\ &= \sum_{i = 1}^{N}{({X_i}^2 + {Y_i}^2 + aX_i + bY_i + c)}^2 \tag{10} \end{align} \]

现在问题变成求参数\(a,b,c\)使得误差和\(Q(a, b, c)\)最小。

3、求解

\(Q(a, b, c)\)的每一项都大于等于0,所以函数存在大于或等于零的极小值,极大值为正无穷大。
\(Q(a, b, c)\)分别对参数\(a,b,c\)求偏导,然后令偏导数为0,即可求出极值点。

\[ \begin{align} \frac{{\rm d}Q(a, b, c)}{{\rm d}a} &= \sum_{i = 1}^{N} 2({X_i}^2 + {Y_i}^2 + aX_i + bY_i + c)X_i = 0 \tag{11} \\ \frac{{\rm d}Q(a, b, c)}{{\rm d}b} &= \sum_{i = 1}^{N}2({X_i}^2 + {Y_i}^2 + aX_i + bY_i + c)Y_i = 0 \tag{12} \\ \frac{{\rm d}Q(a, b, c)}{{\rm d}c} &= \sum_{i = 1}^{N}2({X_i}^2 + {Y_i}^2 + aX_i + bY_i + c) = 0 \tag{13} \end{align} \]

$式(11) * N - 式(13) * \sum_{i = 1}^{N} X_i $消去c

\[ \begin{align} 0 &= N\sum_{i = 1}^{N} 2({X_i}^2 + {Y_i}^2 + aX_i + bY_i + c)X_i - \sum_{i = 1}^{N}2({X_i}^2 + {Y_i}^2 + aX_i + bY_i + c) * \sum_{i = 1}^{N}X_i \tag{14} \\ &= N\sum_{i = 1}^{N} 2({X_i}^2 + {Y_i}^2 + aX_i + bY_i)X_i - \sum_{i = 1}^{N}2({X_i}^2 + {Y_i}^2 + aX_i + bY_i) * \sum_{i = 1}^{N}X_i \tag{15}\\ &= (N\sum_{i = 1}^{N}{X_i}^2 - \sum_{i = 1}^{N}X_i \sum_{i = 1}^{N}X_i)a + (N\sum_{i = 1}^{N}X_i Y_i - \sum_{i = 1}^{N}X_i \sum_{i = 1}^{N}Y_i)b + N\sum_{i = 1}^{N}{X_i}^3 + N\sum_{i = 1}^{N}X_i{Y_i}^2 - N\sum_{i = 1}^{N}({X_i}^2 + {Y_i}^2)\sum_{i = 1}^{N}X_i \tag{16} \end{align} \]

$式(12) * N - 式(13) * \sum_{i = 1}^{N} Y_i $得

\[ \begin{align} 0 &= N\sum_{i = 1}^{N} 2({X_i}^2 + {Y_i}^2 + aX_i + bY_i + c)Y_i - \sum_{i = 1}^{N}2({X_i}^2 + {Y_i}^2 + aX_i + bY_i + c) * \sum_{i = 1}^{N}Y_i \tag{17} \\ &= N\sum_{i = 1}^{N} 2({X_i}^2 + {Y_i}^2 + aX_i + bY_i)Y_i - \sum_{i = 1}^{N}2({X_i}^2 + {Y_i}^2 + aX_i + bY_i) * \sum_{i = 1}^{N}Y_i \tag{18}\\ &= (N\sum_{i = 1}^{N}{X_i}{Y_i} - \sum_{i = 1}^{N}X_i \sum_{i = 1}^{N}Y_i)a + (N\sum_{i = 1}^{N}{Y_i}^2 - \sum_{i = 1}^{N}Y_i \sum_{i = 1}^{N}Y_i)b + N\sum_{i = 1}^{N}{Y_i}^3 + N\sum_{i = 1}^{N}{X_i}^2{Y_i} - N\sum_{i = 1}^{N}({X_i}^2 + {Y_i}^2)\sum_{i = 1}^{N}Y_i \tag{19} \end{align} \]

\[ \begin{align} C &= N\sum_{i = 1}^{N} {X_i}^2 - \sum_{i = 1}^{N}X_i \sum_{i = 1}^{N}X_i \tag{20} \\ D &= N\sum_{i = 1}^{N} {X_i}{Y_i} - \sum_{i = 1}^{N}X_i \sum_{i = 1}^{N}Y_i \tag{21} \\ E &= N\sum_{i = 1}^{N} {X_i}^3 + N\sum_{i = 1}^{N} {X_i}{Y_i}^2 - \sum_{i = 1}^{N}({X_i}^2 + {Y_i}^2) \sum_{i = 1}^{N}X_i \tag{22} \\ G &= N\sum_{i = 1}^{N} {Y_i}^2 - \sum_{i = 1}^{N}Y_i \sum_{i = 1}^{N}Y_i \tag{23} \\ H &= N\sum_{i = 1}^{N} {X_i}^2Y_i + N\sum_{i = 1}^{N}{Y_i}^3 - \sum_{i = 1}^{N}({X_i}^2 + {Y_i}^2) \sum_{i = 1}^{N}Y_i \tag{24} \end{align} \]

可得:

\[ \begin{align} 0 &= Ca + Db + E \tag{25} \\ 0 &= Da + Gb + H \tag{26} \\ a &= \frac{HD - EG} {CG - D^2} \tag{27} \\ b &= \frac{HC - ED} {D^2 - GC} \tag{28} \\ c &= -\frac{\sum_{i = 1}^{N}({X_i}^2 + {Y_i}^2) + a\sum_{i = 1}^{N}X_i + b\sum_{i = 1}^{N}Y_i}{N} \tag{29} \end{align} \]

再结合式(4)(5)(6)即可求得圆心和半径

代码

参考文档

圆拟合算法
最小二乘法拟合圆公式推导及其实现
最小二乘法拟合圆

标签:begin,end,align,散点,tag,拟合,aX,sum
来源: https://www.cnblogs.com/xiaxuexiaoab/p/16276402.html

本站声明: 1. iCode9 技术分享网(下文简称本站)提供的所有内容,仅供技术学习、探讨和分享;
2. 关于本站的所有留言、评论、转载及引用,纯属内容发起人的个人观点,与本站观点和立场无关;
3. 关于本站的所有言论和文字,纯属内容发起人的个人观点,与本站观点和立场无关;
4. 本站文章均是网友提供,不完全保证技术分享内容的完整性、准确性、时效性、风险性和版权归属;如您发现该文章侵犯了您的权益,可联系我们第一时间进行删除;
5. 本站为非盈利性的个人网站,所有内容不会用来进行牟利,也不会利用任何形式的广告来间接获益,纯粹是为了广大技术爱好者提供技术内容和技术思想的分享性交流网站。

专注分享技术,共同学习,共同进步。侵权联系[81616952@qq.com]

Copyright (C)ICode9.com, All Rights Reserved.

ICode9版权所有