#19 Error: singular matrix while computing thin plate spline

[Edward Uebler]

I ran too large files of 15 to 20 million points and the above error occurs in SD 2 after 7 to 13 passes. First Domain ran 40 to 50 passes. attcahed is screen captuiured output.

[Edward Uebler]

• attachment _edstestout.txt_ added

Malheur Test file from Ed Uebler

[Edward Uebler]

• attachment _muckout.txt_ added

Mt Hood test file from Jim Muckenhoupt

[Jimm Domingo]

• owner set to jimm_domingo
• status changed from new to accepted
• milestone set to 1.0 (b4)

Added new unit tests that reproduce the singular matrix errors for the set of control points reported in both attachments (revision r126).

[Jimm Domingo]

Added more unit tests that show that the singular matrix errors are independent of the control points' X,Y orientation (revision r128). Each set of control points reported in the attachments lie in a north-south line (X coordinate = same value). If the points are oriented in an east-west line (Y coordinate = same value), the errors still occur. The errors also occur if the points are oriented in a diagonal line running from southwest to northeast (Y = X).

[Jimm Domingo]

Added more unit tests that illustrate how to modify the control points to prevent singular matrix errors (revision r129).

[Jimm Domingo]

Modified the unit tests to check if a set of 3 control points denotes a vertical plane (r130). If they do, then they will result in a singular matrix error.

[Jimm Domingo]

• status changed from accepted to closed
• resolution set to fixed

Fixed this ticket in revision r131. Modified the sorting of the unclassified points into interpolation regions. The test that determines if a region has enough points has been modified to include an additional check for a region with 3 associated points. If the 3 points represent a vertical plane, then additional points are collected for the region.