1
$\begingroup$

In the x-y plane, a parametric curve defined by $\mathcal{C}$,

$$\pmb{\mathcal{C}= (x,y) \equiv \left( u_t ,v_t \right)},$$

touches a line $\pmb{L_\theta(x,y)}$ where $\theta$ is the orientation of the line with respect to the x-axis.

I am interested in distance $\pmb{d_\theta}$ which is defined as the distance between the points of intersection between $\pmb{\mathcal{C}}$ and $\pmb{L_\theta}$. This is shown in the figure below.

Assume that the parametric equations of the curve are known. Is there's a nice mathematical formulation for solving this problem. May be related to Radon Transform?

I am able to solve this problem numerically using Python. But I have problems formalizing this as an analytical solution.

Any hints or connections is welcome.

Figure

  • 0
    Why new labels u,v for x,y ?2017-01-14
  • 0
    In the diagram, the image of $\mathcal{C}$ is a polar graph, i.e., it crosses each ray through the origin exactly once. Are you assuming this condition generally? Separately, you write "[the curve] touches _a_ line $L_{\theta}$", but the diagram shows the curve crossing _the_ line _through the origin_ making angle $\theta$ with the positive $x$-axis. Is that your intent?2017-01-14
  • 1
    @ Andrew D. Hwang I am assuming the condition generally. This is right. Also, the line always passes through the origin. As a result, for each $\theta$, there's a new pair of points of intersection between the curve and the line.2017-01-15

1 Answers 1

1

$\newcommand{\Cv}{\mathcal{C}}\DeclareMathOperator{\Atan}{atan2}$In the diagram, and as clarified in the comments, the curve $\Cv$ is a polar graph $r = f(\theta)$, for some positive function $f$ to be determined from the component functions $(x, y)$ of the parametrization. In terms of $f$, the desired distance function is $$ d_{\theta} = f(\theta) + f(\theta + \pi). $$

To find $f$, write $\theta = \Theta(t)$ and \begin{alignat*}{2} x(t) &= r\cos\theta &&= f(\theta) \cos\theta, \\ y(t) &= r\sin\theta &&= f(\theta) \sin\theta, \end{alignat*} so that $$ f(\theta) = \sqrt{x(t)^{2} + y(t)^{2}} \tag{1} $$ and $$ \Theta'(t) = \frac{d\theta}{dt} = \frac{y(t) x'(t) - x(t) y'(t)}{x(t)^{2} + y(t)^{2}}. \tag{2} $$ Under the geometric assumptions on $\Cv$, the preceding expression is non-vanishing, and without loss of generality (i.e., assuming $\Cv$ is traced counterclockwise) may be taken to be positive. If $[a, b]$ is the domain of the given parametrization of $\Cv$, then $$ \Theta(t) = \int_{a}^{t} \frac{y(\tau) x'(\tau) - x(\tau) y'(\tau)}{x(\tau)^{2} + y(\tau)^{2}}\, d\tau \tag{3} $$ defines an invertible function $\Theta:[a, b] \to [0, 2\pi]$. The function $f$ is given by (1): $$ f(\theta) = \sqrt{x(\Theta^{-1}(\theta))^{2} + y(\Theta^{-1}(\theta))^{2}}. $$

While this analytic expression is exact, it entails inverting the definite integral (3), and may therefore be inconvenient for practical use. If instead the goal is to calculate $d_{\theta}$ for finitely many specified values of $\theta$, it suffices, for a given $\theta$, to find (using Newton's method, say) the values $t_{1}$ and $t_{2}$ such that $$ \Atan(y(t_{1}), x(t_{1})) = \theta,\qquad \Atan(y(t_{2}), x(t_{2})) = \theta + \pi, $$ so that $$ d_{\theta} = \sqrt{x(t_{1})^{2} + y(t_{1})^{2}} + \sqrt{x(t_{2})^{2} + y(t_{2})^{2}}. $$ As usual, $\Atan(y, x)$ denotes the branch of polar angle of the ray from the origin through $(x, y)$ and taking values in $(-\pi, \pi)$.