Wenglor B50 - Basic Character Geometry; Examples; Size of the ROI

151
Smart Camera / Vision-Sensor / 1D-/2D-Code-Scanner / OCR Reader
11.15.1. Basic Character Geometry
• The OCR Reader functions ideally as of a character height of 25 pixels. In this case, as a rule, the gap
between the characters are large enough for the characters to be separated.
• The OCR Reader functions ideally when the gap between the characters is half as large as the character
width.
• If “non-linear calculation of the binarization threshold” is used, the gap between the characters should not
be any larger than one character. Otherwise the gap itself might be recognized as a character under certain
circumstances. In this case, two objects should be used.
• Processing time has a quadratic relationship to character size. If a character string with a character
height of 25 pixels requires 20 ms for the reading algorithm, time is increased to 80 ms for a character
height of 50 pixels.
11.15.1.1. Examples
Font: Arial Standard
Height: 30 pixels
 “02” cannot be segmented.
Font: OCR B
Height: 30 pixels
 All characters can be readily segmented.
Font: OCR B
Height: 30 pixels
Binarization: “non-linear calculation”
 Excessively large spaces are seen as separate
segments.
11.15.2. Size of the ROI
If the region of interest is too large, the algorithm for determining the binarization threshold does not function
reliably.
The following rule of thumb applies:
Left/right edge spacing: 1 x character width
Top/bottom edge spacing: 0.5 x character width
This “quiet zone” should not be interfered with by other characters or objects.
Furthermore, processing time also increases when the ROI is too large.
If the position of the character string to be read is not consistent in actual practice, it’s usually better to use lo-
calization instead of a very large ROI. Processing time for localization plus reading is usually less than required
for a large ROI.