Optical Inspection Systems

Fraunhofer Institute for Integrated Circuits

Symbol Based Identification

Solutions for difficult identification tasks

container

Symbol based identification offers systematic solutions for difficult identification tasks. This allows the identification of objects that can not be easily labeled by machine readable characters or bar codes.
To quote as an example: engraved symbols or characters on die cast parts or tyres. A further application is the reading of characters under "real world" conditions, e.g. on containers. The core technologies cover imaging methods like 3-D-sensors or color cameras, as well as the latest methods in image processing for extraction and interpretation of symbols and labels.

The approach processing the example picture on the left is as follows:

  • Identification of regions carrying labels or characters (segmentation)
  • This includes the correction of perspective deformations.
  • Optical Character Recognition, (OCR) The OCR used here fulfills higher demands than standard OCR, as it allows:
             1. arbitrary character and background colors
             2. different fonts and font sizes
             3. unknown text orientation
             4. heavily deformated texts (uneven surfaces)
             5. noisy images (rust, stains, shadows)
             6. defect characters (scratches, rust, masking) 
  • Enhancing reliability by syntactical verification (e.g. check digits)