We have scans (PDF) of a number of reports documenting early geometric models in the COMGEOM format (a now obsolete format, but the models are interesting nonetheless). These reports contain the actual geometry defining the model as pages and pages of numbers and letters. Unfortunately, the quality is sufficiently poor that optical character recognition (OCR) has a very high rate of error.

This task is to attempt the manual transcription of the MEP-021A Generator Set model described in the report ''A Combinatorial Geometry Computer Description of the MEP-021A Generator Set'' (see the References list below for the link that will let you download the PDF). One possible approach is to use Acrobat Reader or some other PDF reader select and copy the OCR text, paste that to a text file as a starting point, and then manually correct it. There may also be some patterns that will allow for semi-automated processing (for example, if 5 zeros in a row are commonly replaced with the character ''O'' instead of 0, a search and replace is in order.) However you wish to approach it is fine, but remember that the goal is not just the extraction of the OCR text but the production of an accurate transcription of the file. The OCR text can be used as a starting point but it will NOT be accurate.

The goal is to have a file that can be fed to BRL-CAD's comgeom-g importer to generate an accurate .g file. If the generator conversion goes well, there are a significant number of other models (some of them considerably more complex) that we are interested in that can be used to create more tasks of this nature.

References:

• http://www.dtic.mil/docs/citations/ADA073408

Code:

• src/conv/comgeom

Please discuss your progress with the developers. This task has the potential for being broken up into multiple tasks depending on the time and accuracy of your conversion.