Building a model requires that a set of mathematical functions be built to establish correspondence between any point in the base geometry and the corresponding point in the target geometry.

The way that Register establishes this set of functions is by placing a set of corresponding points, called control points, and interpolating for the remainder of space around the control points. The control points are represented by color marks in both base and uncorrected spaces. The interpolation is done with smooth two dimensional functions like Helmert, Affine, Similitude, Projective, and up to a third degree Polynomial.

To test the accuracy of the interpolation, Register warps a representation of the base vector data to fit the uncorrect data. This visualization illustrates the fit that occurs when you resample the uncorrect data to fit the base data.

This visual feedback assists you as you gather more points and improve the fit. After enough points are gathered to satisfy the requirements of the selected transformation model, the uncorrect image or vectors can be resampled using this model. Since the fit has already been determined visually, the result is “what you see is what you get” resampling.

Warping the base vector data is done dynamically and is normally fast. It is recommended that you warp a small but significant set of base vector data. The set of base vector data can be acquired by extraction from one or several base documents, in digital or analog formats: design files, digital images, analog maps, drawings, plans, lists of known points, and so forth.

With a design file, extract a few representative elements or parts of elements. With a digital image, extract a few representative elements by interpretation. With a hard copy document, digitize a few representative elements. With a list of known points, no base vector data set is needed to use the dynamic warping capability since the control point marks play this role.

You can also use Register to build a model without a set of base vector data and without dynamic warping, which resembles the method that is commonly used with most geometric correction software. However, with this method there is no means to test the accuracy of the model prior to resampling the image or vector file.

Another unique feature of Register is the alleviation of the constraint of finding “true” or “exact” control points. With Register, you can use “fuzzy” control points, because of the visual feedback provided by dynamic warping.

A true or exact control point is a point which can be precisely identified on both the base data set and on the data to be registered. Examples include the crossing of two roads, the corner of a part, the corner of a building, and so forth. A fuzzy control point is a point which can be localized approximately on both documents. Examples include the center of a lake, window, or field, or the top or side of a car, and so forth.

When you use fuzzy control points, pick each point approximately in the base space and pin it in the uncorrected space by determining its general position. The dynamic warping shows if your placement is accurate enough. If not, just move the point.

The following design elements are supported for dynamic warping: