Character recognition applications are frequently used in Industry. Industrial character recognition has significant differences when compared to document type character recognition applications.

Frequent industrial character recognition applications are:

• Reading characters in the form of scraping or embossing on a metal surface
• Reading from surfaces such as plastic or glass
• Reading the label or print on the product
• Reading various characters on products moving on a conveyor
• Reading of information such as Serial Number, expiry date
• Reading information such as Chassis number, VIN, product code

As can be imagined, it is not possible to implement such applications by using a general purpose OCR application. Using various lighting and advanced image processing techniques, it is possible to develop high performance OCR (Character Reading) applications.

OCR applications require application-specific lighting and the use of appropriate lenses. If reading from cylindrical side surfaces is the case, then multiple camera systems or pericentric lenses can be used to see the whole surface.

As lighting, proper illumination required by the application such as coaxial, dome, led, ring needs to be chosen correctly.

In some applications, with photometric stereo techniques, it is possible to read hard-to-see texts.

Some sample application pictures we are working on:

Reading the laser print on cylindrical surface

Recognition of the characters arranged at a circular angle

Reading handwritten numbers. (Only numeric characters can be recognised.)

Reading characters written in dot matrix and decomposing different products

the OCR process,by using deep learning and classification methods, teaching to the machine in a sense, it is possible to achieve a very high success rate on character recognition.

Industrial OCR processing usually is not limited to recognition only. Often there are steps such as finding the product first, then searching for pre-determined coordinates, and then recognizing the text in the area of interest. In some cases, some part of the characters can be seen dark and some part can be seen light. In such cases advanced image processing techniques are used.

Derivative gauss filtered picture.

As a result, industrial character recognition is a field with its own unique challenges. Correct camera, lens, lighting choice and advanced image processing functions need to be addressed.