Specifications
Print2CAD OCR 2013 - 132
22.2.5 Text Recognition Quality
The quality of the text recognition is, amongst other things, inuenced by the following
factors:
- Quality of the specied text areas (depending on the user)
- Extent and quality of the prototype database (given in the software)
- Extent and quality of the dictionaries (given in the software)
- Quality of the manual correction of the words (depending on the user)
- Chromaticity, contrast, layout and font of the original le
- Resolution and quality of the image le (depending on RAM)
Whereas a pure pattern recognition with Print2CAD achieves a correctness in the range
of 90% (every tenth gure gets recognized incorrectly), a good manual review achieves
a correctness of up to 99%.
Important!
While text allows a higher error ratio, numbers, e.g. dimension numbers, require a
thorough and repeated proofreading.
Especially numbers containing “3” and “8” should be checked to ensure that “3”
and “8” have not been interchanged.
The same applies to “5” and “6,” as well as “1” and “7.”
Important!
A thorough selection of the text through text areas (text and numbers seperated)
vastly improves the quality of the text recognition.
Error! Results will be inaccurate