Skip to content

Affine transformations

An affine transformation is a geometrical transformation that preserves lines (they remain straight). It may alter distances, angles, scales, orientations. Affine transforms are linear transformations and commonly implemented using matrix algebra.

The Transform2d class in djutils-draw performs affine transforms on 2D objects; the Transform3d class operates on 3D objects. Unlike all other objects in the djutils-draw package, the transform objects are mutable. The djutils-draw transformations are constructed as the identity transformation, then modified by applying some sequence of translations, rotations, shear-operations and reflections. As each of these modifications of a transform returns the resulting transform; these operations can be chained. Finally the transformation can operate on a Point or Bounds object or on an Iterator<Point>. (This will probably be extended to all Drawable objects.)

Setting up a Transform2d or Transform3d involves some math that is relatively expensive when rotations are involved. Subsequently applying such a transformation to coordinates is very fast as that only involves a few multiplications and additions.

Translating and Rotating

In this example, a unity Transform3d object is constructed and then a translation is added to it, and then a rotation. The net effect on the Point3d is that is first rotated and then translated.

    Point3d point = new Point3d(1, 0, 0);
    System.out.println(point);
    Transform3d transform1 = new Transform3d();
    transform1.translate(2, 5, 8); // Translate
    System.out.println(transform1.transform(point));
    transform1.rotY(Math.PI / 2);
    System.out.println(transform1.transform(point));

Outputs

Point3d [x=1.000000, y=0.000000, z=0.000000]
Point3d [x=3.000000, y=5.000000, z=8.000000]
Point3d [x=2.000000, y=5.000000, z=7.000000]

The resulting transformation could have been created in a single line of Java code using chaining:

    Transform3d transform2 = new Transform3d().translate(2, 5, 8).rotY(Math.PI / 2);
    System.out.println(transform2.transform(point));

Outputs:

Point3d [x=2.000000, y=5.000000, z=7.000000]

Of course, the Transform3d class also implements rotations around the X and Z axes. As the Transform2d class only implements rotation perpendicular to the X-Y plane this method is simply named rotation.

Shearing

The shear operator transforms a rectangle into a parallelogram. The Transform2d class implements one shear(double, double) method. The Transform3d class implements three shear methods: shearXY(double, double), shearYZ(double, double) and shearZX(double, double).

    Transform2d transform = new Transform2d().shear(2, 3);
    for (int x = 0; x < 2; x++)
    {
        for (int y = 0; y < 2; y++)
        {
            Point2d p = new Point2d(x, y);
            System.out.println(p + " -> " + transform.transform(p));
        }
    }

Outputs:

Point2d [x=0.000000, y=0.000000] -> Point2d [x=0.000000, y=0.000000]
Point2d [x=0.000000, y=1.000000] -> Point2d [x=2.000000, y=1.000000]
Point2d [x=1.000000, y=0.000000] -> Point2d [x=1.000000, y=3.000000]
Point2d [x=1.000000, y=1.000000] -> Point2d [x=3.000000, y=4.000000]

The X-coordinates are incremented by the first parameter of the shear times the Y-coordinate. The Y-coordinates are incremented by the second parameter of the shear times the X-coordinate.

    Transform3d transform = new Transform3d().shearXY(2, 3);
    for (int x = 0; x < 2; x++)
    {
        for (int y = 0; y < 2; y++)
        {
            for (int z = 0; z < 2; z++)
            {
                Point3d p = new Point3d(x, y, z);
                System.out.println(p + " -> " + transform.transform(p));
            }
        }
    }

Outputs:

Point3d [x=0.000000, y=0.000000, z=0.000000] -> Point3d [x=0.000000, y=0.000000, z=0.000000]
Point3d [x=0.000000, y=0.000000, z=1.000000] -> Point3d [x=2.000000, y=3.000000, z=1.000000]
Point3d [x=0.000000, y=1.000000, z=0.000000] -> Point3d [x=0.000000, y=1.000000, z=0.000000]
Point3d [x=0.000000, y=1.000000, z=1.000000] -> Point3d [x=2.000000, y=4.000000, z=1.000000]
Point3d [x=1.000000, y=0.000000, z=0.000000] -> Point3d [x=1.000000, y=0.000000, z=0.000000]
Point3d [x=1.000000, y=0.000000, z=1.000000] -> Point3d [x=3.000000, y=3.000000, z=1.000000]
Point3d [x=1.000000, y=1.000000, z=0.000000] -> Point3d [x=1.000000, y=1.000000, z=0.000000]
Point3d [x=1.000000, y=1.000000, z=1.000000] -> Point3d [x=3.000000, y=4.000000, z=1.000000]

The shearXY operation does not modify the Z coordinates. The X-coordinates are incremented by the first parameter of the shear times the Z value. The Y-coordinates are incremented by the second parameter of the shear times the Z value. Similarly, there shearYZ operation preserves the X-coordinate values and the shearZX operation preserves the Y-coordinate values.

Scaling

A transform can also perform scaling by independent factors along each axis.

    Transform2d transform = new Transform2d().scale(4, 6);
    for (int x = 0; x < 2; x++)
    {
        for (int y = 0; y < 2; y++)
        {
            Point2d p = new Point2d(x, y);
            System.out.println(p + " -> " + transform.transform(p));
        }
    }

Outputs:

Point2d [x=0.000000, y=0.000000] -> Point2d [x=0.000000, y=0.000000]
Point2d [x=0.000000, y=1.000000] -> Point2d [x=0.000000, y=6.000000]
Point2d [x=1.000000, y=0.000000] -> Point2d [x=4.000000, y=0.000000]
Point2d [x=1.000000, y=1.000000] -> Point2d [x=4.000000, y=6.000000]

Reflecting

The Transform2d object implements reflectX() and reflectY(). the Transform3d object additionally implements reflectZ(). These methods invert that particular coordinate.

Transforming multi-point Drawable objects

All preceding examples of using transforms showed how a transform can be applied to a Point. Each Drawable object implements getPoints() which constructs an Iterator<Point> that will yields all points of the Drawable. Also, Drawable objects that are constructed from a varying number of points have a constructor that takes such an iterator. This can be used to construct a new transformed Drawable with a single line of code:

    PolyLine2d line = new PolyLine2d(new Point2d(1, 2), new Point2d(2, 3), new Point2d(5, 0));
    System.out.println(line);
    Transform2d transform = new Transform2d().scale(2, 3);
    PolyLine2d transformedLine = new PolyLine2d(transform.transform(line.getPoints()));
    System.out.println(transformedLine);

Outputs:

PolyLine2d [x=1.000000, y=2.000000, x=2.000000, y=3.000000, x=5.000000, y=0.000000]
PolyLine2d [x=2.000000, y=6.000000, x=4.000000, y=9.000000, x=10.000000, y=0.000000]