Transforms

This page is up-to-date as of Parallax version 0.28.2

A transform in Parallax is a data object that represents a coordinate transformation from one coordinate system (CS) to another (plus metadata). A transform is generated from a set of correspondence points - pairs of 3D coordinates that define a mapping from one coordinate system (the "from" CS) to another (the "to" CS). Once a transform has been generated, it can be used to convert 3D object points from one system to another (bidirectionally).

Transforms can be saved and loaded as Python Python pickle files.

The data contained within a transform object/file includes:

• Name
• "From" coordinate system
• "To" coordinate system
• The internal parameters (rotation, translation, scaling, etc. - the specifics will depend on the type of transform used)
• RMSE - the root-mean-square re-projection error of the correspondence points
• Reprojection convergence - the variation (RMS) in the reprojected correspondence points as we sample n-1 of the points, for all n permutations.
• Parameter convergence - the variation (RMS) in the internal parameters as we sample n-1 of the points, for all n permutations.

Transforms in Parallax can be applied forward or backward, and composed into composite transforms.