You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
150 lines
5.6 KiB
150 lines
5.6 KiB
3 months ago
|
// David Eberly, Geometric Tools, Redmond WA 98052
|
||
|
// Copyright (c) 1998-2021
|
||
|
// Distributed under the Boost Software License, Version 1.0.
|
||
|
// https://www.boost.org/LICENSE_1_0.txt
|
||
|
// https://www.geometrictools.com/License/Boost/LICENSE_1_0.txt
|
||
|
// Version: 4.0.2019.08.13
|
||
|
|
||
|
#pragma once
|
||
|
|
||
|
#include <Mathematics/ApprGaussian2.h>
|
||
|
#include <Mathematics/Hyperellipsoid.h>
|
||
|
#include <Mathematics/Projection.h>
|
||
|
|
||
|
namespace gte
|
||
|
{
|
||
|
// The input points are fit with a Gaussian distribution. The center C of
|
||
|
// the ellipse is chosen to be the mean of the distribution. The axes of
|
||
|
// the ellipse are chosen to be the eigenvectors of the covariance matrix
|
||
|
// M. The shape of the ellipse is determined by the absolute values of
|
||
|
// the eigenvalues. NOTE: The construction is ill-conditioned if the
|
||
|
// points are (nearly) collinear. In this case M has a (nearly) zero
|
||
|
// eigenvalue, so inverting M can be a problem numerically.
|
||
|
template <typename Real>
|
||
|
bool GetContainer(int numPoints, Vector2<Real> const* points, Ellipse2<Real>& ellipse)
|
||
|
{
|
||
|
// Fit the points with a Gaussian distribution. The covariance matrix
|
||
|
// is M = sum_j D[j]*U[j]*U[j]^T, where D[j] are the eigenvalues and
|
||
|
// U[j] are corresponding unit-length eigenvectors.
|
||
|
ApprGaussian2<Real> fitter;
|
||
|
if (fitter.Fit(numPoints, points))
|
||
|
{
|
||
|
OrientedBox2<Real> box = fitter.GetParameters();
|
||
|
|
||
|
// If either eigenvalue is nonpositive, adjust the D[] values so
|
||
|
// that we actually build an ellipse.
|
||
|
for (int j = 0; j < 2; ++j)
|
||
|
{
|
||
|
if (box.extent[j] < (Real)0)
|
||
|
{
|
||
|
box.extent[j] = -box.extent[j];
|
||
|
}
|
||
|
}
|
||
|
|
||
|
// Grow the ellipse, while retaining its shape determined by the
|
||
|
// covariance matrix, to enclose all the input points. The
|
||
|
// quadratic form that is used for the ellipse construction is
|
||
|
// Q(X) = (X-C)^T*M*(X-C)
|
||
|
// = (X-C)^T*(sum_j D[j]*U[j]*U[j]^T)*(X-C)
|
||
|
// = sum_j D[j]*Dot(U[j],X-C)^2
|
||
|
// If the maximum value of Q(X[i]) for all input points is V^2,
|
||
|
// then a bounding ellipse is Q(X) = V^2, because Q(X[i]) <= V^2
|
||
|
// for all i.
|
||
|
|
||
|
Real maxValue = (Real)0;
|
||
|
for (int i = 0; i < numPoints; ++i)
|
||
|
{
|
||
|
Vector2<Real> diff = points[i] - box.center;
|
||
|
Real dot[2] =
|
||
|
{
|
||
|
Dot(box.axis[0], diff),
|
||
|
Dot(box.axis[1], diff)
|
||
|
};
|
||
|
|
||
|
Real value =
|
||
|
box.extent[0] * dot[0] * dot[0] +
|
||
|
box.extent[1] * dot[1] * dot[1];
|
||
|
|
||
|
if (value > maxValue)
|
||
|
{
|
||
|
maxValue = value;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
// Arrange for the quadratic to satisfy Q(X) <= 1.
|
||
|
ellipse.center = box.center;
|
||
|
for (int j = 0; j < 2; ++j)
|
||
|
{
|
||
|
ellipse.axis[j] = box.axis[j];
|
||
|
ellipse.extent[j] = std::sqrt(maxValue / box.extent[j]);
|
||
|
}
|
||
|
return true;
|
||
|
|
||
|
}
|
||
|
|
||
|
return false;
|
||
|
}
|
||
|
|
||
|
// Test for containment of a point inside an ellipse.
|
||
|
template <typename Real>
|
||
|
bool InContainer(Vector2<Real> const& point, Ellipse2<Real> const& ellipse)
|
||
|
{
|
||
|
Vector2<Real> diff = point - ellipse.center;
|
||
|
Vector2<Real> standardized{
|
||
|
Dot(diff, ellipse.axis[0]) / ellipse.extent[0],
|
||
|
Dot(diff, ellipse.axis[1]) / ellipse.extent[1] };
|
||
|
return Length(standardized) <= (Real)1;
|
||
|
}
|
||
|
|
||
|
// Construct a bounding ellipse for the two input ellipses. The result is
|
||
|
// not necessarily the minimum-area ellipse containing the two ellipses.
|
||
|
template <typename Real>
|
||
|
bool MergeContainers(Ellipse2<Real> const& ellipse0,
|
||
|
Ellipse2<Real> const& ellipse1, Ellipse2<Real>& merge)
|
||
|
{
|
||
|
// Compute the average of the input centers.
|
||
|
merge.center = (Real)0.5 * (ellipse0.center + ellipse1.center);
|
||
|
|
||
|
// The bounding ellipse orientation is the average of the input
|
||
|
// orientations.
|
||
|
if (Dot(ellipse0.axis[0], ellipse1.axis[0]) >= (Real)0)
|
||
|
{
|
||
|
merge.axis[0] = (Real)0.5 * (ellipse0.axis[0] + ellipse1.axis[0]);
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
merge.axis[0] = (Real)0.5 * (ellipse0.axis[0] - ellipse1.axis[0]);
|
||
|
}
|
||
|
Normalize(merge.axis[0]);
|
||
|
merge.axis[1] = -Perp(merge.axis[0]);
|
||
|
|
||
|
// Project the input ellipses onto the axes obtained by the average
|
||
|
// of the orientations and that go through the center obtained by the
|
||
|
// average of the centers.
|
||
|
for (int j = 0; j < 2; ++j)
|
||
|
{
|
||
|
// Projection axis.
|
||
|
Line2<Real> line(merge.center, merge.axis[j]);
|
||
|
|
||
|
// Project ellipsoids onto the axis.
|
||
|
Real min0, max0, min1, max1;
|
||
|
Project(ellipse0, line, min0, max0);
|
||
|
Project(ellipse1, line, min1, max1);
|
||
|
|
||
|
// Determine the smallest interval containing the projected
|
||
|
// intervals.
|
||
|
Real maxIntr = (max0 >= max1 ? max0 : max1);
|
||
|
Real minIntr = (min0 <= min1 ? min0 : min1);
|
||
|
|
||
|
// Update the average center to be the center of the bounding box
|
||
|
// defined by the projected intervals.
|
||
|
merge.center += line.direction * ((Real)0.5 * (minIntr + maxIntr));
|
||
|
|
||
|
// Compute the extents of the box based on the new center.
|
||
|
merge.extent[j] = (Real)0.5 * (maxIntr - minIntr);
|
||
|
}
|
||
|
|
||
|
return true;
|
||
|
}
|
||
|
}
|