Module name: |
MAT |
Module identifier: |
IG_REC_RM_MAT |
Filling methods supported: |
IG_REC_FM_OCRA, IG_REC_FM_OCRB, IG_REC_FM_MICR, IG_REC_FM_DOTDIGIT, IG_REC_FM_DASHDIGIT |
Filters supported: |
all elements except IG_REC_FILTER_LOWERCASE |
Trade-off supported: |
none |
Knowledge base file: |
SFOCR.TUD |
Training supported: |
no |
This section provides information about the following:
This module is designed to read certain groups of fixed-font characters specially designed for OCR or imaging applications, in which no two characters have similar shapes. Each character group has its own filling method. Application areas are in banking, check or waybill handling, product distribution and document validation, where high accuracy can be vital. It also handles some non-fixed print styles.
The above-mentioned supported filling methods cover the ranges of characters that the MAT engine knows. The description of these ranges can be found in OCR special filling methods.
When this module is handling fixed fonts, well printed on a firm surface, with a limited number of possible characters, it should provide high accuracy recognition. This can be further enhanced by use of regular expressions (UD-checking) or user-written checking available through the checking module. This module doesn't interpret the page-level recognition trade-off settings. This recognition is fairly tolerant of differing resolution settings. The dot- and dash-digit symbols are typically printed onto product containers, which may be curved or molded; furthermore many different standards are used. For high accuracy here, the texts to be read should be found on flat surfaces (on packing documents rather than product containers).
If the application does not need spell checking, it can be switched off; this may increase the module's performance.
This module supports the checking module fully, and also all filter elements (however, IG_REC_FILTER_LOWERCASE will have no effect). Characters should be at least 4 x 8 pixels large and should not be touching. IG_REC_FM_OCRA and IG_REC_FM_OCRB support both single and multi-line zones, the other filling methods (IG_REC_FM_MICR, IG_REC_FM_DOTDIGIT, and IG_REC_FM_DASHDIGIT) do not; each line must be zoned separately. MICR strings found on checks are usually single-line only.