Font optimization – Rockwell Automation 5370-OCR2 PAK User Manual

Page 57

Advertising
background image

Chapter 5

Creating and Editing Fonts

5–22

The most important aspect of creating inspection fonts is font optimization,
where the patterns and settings of individual characters are adjusted for best
performance. The success of an OCR application depends on correctly
setting up the font to recognize characters despite image variations and print
degradation, and to eliminate confusion between different characters that are
similar in appearance. The following guidelines are intended to assist you in
setting up an OCR application.

Reducing Image Variations

Variations in the acquired image make it more difficult to accurately
recognize characters. The following types of image variation occur:

Lighting variations
Variations in the lighting across the field of view should be minimized
through careful lighting design. Variations in brightness from one image
to the next, due to light source intensity variation, can be reduced by
using the light probe. Refer to the CVIM2 Reference Manual for a
description of light probe operation.
Important: Variations in brightness can also compensated for by using
the CVIM2 Image Tool to preprocess the image for the OCR Tool. Refer
to Appendix B for more information.

Blurring due to motion
Blurring can cause the characters to become elongated in one direction.
This can be a problem particularly if the degree of blurring varies.
Blurring should be reduced or eliminated by using strobe lighting or
electronic shutter cameras.

Handling Print Degradation

Different printing processes introduce different forms of degradation or
variation into the printed characters. Some common printing problems are:

Broken characters
When part of a character is missing, this reduces its degree of correlation
with the taught character pattern. Usually, characters that are partially
printed should be rejected. However, the system can be configured to
accept some degree of broken characters. This is accomplished by
adjusting the correlation percentage.

Changes in character shape
Some printing processes introduce changes in the shape of the characters.
The system will recognize characters that vary in shape if there is a
sufficiently wide border of don’t-care pixels around the edge of the
character in the taught character pattern. If the variation in character
shape is extreme, the system will not be able to recognize the characters.

Font Optimization

Advertising