Package org.geotools.referencing.operation.builder

A package of convenience classes which use control points common to two data sets to derive empirically the transformation parameters needed to convert positions between the coordinate systems of the two data sets.

Note the implementation is currently (January 2008) limited to two dimensions. The methods, however, are generic and could be expanded to three dimensions someday. At that time, we will probably make minor changes to the API.

The package consists of two types of convenience classes: the various builder classes which use a set of individual control points to obtain the conversion and the GridToEnvelopeMapper class which derives the conversion from a grid range to a georeferenced Envelope.

The builder classes should be used by users who have two data sets that are known to share certain common points but who currently do not line up. This could be the case, for example, if a user has two data sets describing the same region but one of these has an unknown coordinate referencing system. In this situation, there is no way to convert coordinate positions between the two data sets. However, if the user can identify a series of positions coupled in each data set, a Builder can calculate an empirical conversion between the two data sets. The different Builder classes use different mathematical approaches to obtain the empirical estimate.

The GridToEnvelopeMapper should be used by users who have a grid, such as an image, which is not georeferenced but the user knows the grid is aligned in one of the four cardinal directions and the user can identify the outer georeferenced envelope of the grid. The Mapper class can then calculate an empirical conversion object to map positions in the image coordinate system to georeferenced positions.

The builder classes require a matched set of known positions, one from a "source" data set and another from a "target" data set; the builder will then provide a structure which contains a conversion object to transform positions from the "source" coordinate system to the "target" coordinate system. The builders require a list of MappedPosition objects which are associations of a DirectPosition in the "source" data set with another DirectPosition in the "target" data set. The getTransformation() method in the builder can then be used to provide a Transformation object from which the user can obtain the MathTransform to use for conversion operations.

Different builders use different mathematical approaches for obtaining the empirical estimate of the conversion parameters. The builders are:

with the mathematical details of each estimation procedure explained in the documentation of the builder class itself. The first four of these use a least squares estimation method in which, if the system is over-determined by having more than the minimum number of control points necessary to derive the estimate, the best matching parameter estimate will be obtained by minimising the sum of the squared distances to the points. The RubberSheet algorithm uses a linear interpolation between the various control points.