Product Identification

In the industry, several product type identification codes are developed.
The most well-known is the bar code, which can be found on food packages. The barcode itself is a robust but low efficient code.
Several types and codes (called, as symbologies, e.g. EAN=-13, UPC-A, Code 93 and code 25) are developed through the years and even the libraries (bookland) and blood banks (Codabar USD-4, 2of7) developed their own barcodes.
Even the mail service developed their own Postal Barcodes (e.g. UK Royal Mail RM4SCC, Postnet, Australia Post and Deutschen Post).
code93a issn code1 Philips dotcode-a datamatrix maxicode
To increase the efficiency of these codes, a 2D codes was developed, e.g. Philips Dot Code A, Code 1, Data Matrix, Maxicode and Axtec Code.
These 2D codes use a 2D reader to read the code, These readers are more expensive than 1D code readers. Even colour codes are created as ultra code which also needs a color 2D scanner.
Almost every code uses a check digit to realize a robust detection.
postnet The Postnet code (picture left) is actually a binary code with 5 digits. The long bar is comparable with a binary 1. The code itself contains a check digit based on Mod 10 algorithm.
(The value of the check digit is that when added to the sum of other digits in the barcode, results in a total that is multiple of 10. For example: The check digit for 91801 is 1 since 9+1+8+0+1+ 1 =20, which is two times the multiple of 10).

But all of these codes are not readable or hardly readable for the human eye. Therefore, alphanumeric codes, that are readable, are used. The advantage is, of course, that every human can read the code so no code reader is necessary. In the industry this alphanumeric identification, e.g. production factory, date and machine number, are often used.
Product Identification/images/Complete image numeric code For the production of CRT (Cathode Ray Tube) screens at Philips Display Components at Aachen an alphanumeric code, containing production date and factory number, is used. This code is applied by a laser just after the screen has left the glass mould. The screen is still at a high temperature when the laser scribes this code at the outside of the screen.
For Statistical Process Control and traceability it is important to read this code on several places in the factory. Therefore a special reader is developed to read this numerical code. The quality of the code on the screen depends on the positioning of the screen in front of the laser and the focusing of the laser. In the picture left, a complete image of the code is shown. On the code area the glass does have a polished surface. Outside this area the glass shows a rough structure.

Vision inspection setup

A qualitative good image with a high contrast can only achieved when a large diffuse illumination is used under a certain angle with the screen.
In fact, the polished surface of the screen is used as a mirror.
The camera is positioned at a different angle, which is shown in the drawings of the final version.
The results show the contour of the figures. Only the contour radius of every figure gives the contrast.

Product Identification Number Prod Identification first test setup Camera inline setup Monitor inline result
The first set-up in the production environment is shown in these three pictures below.
First picture is an overview where the screen transport is clearly visible; and at the other side the vision system is positioned.
In the center picture, a close view from the side of the camera is given. The code will be read at the area where a blue guidance wheel is missing.
On the picture to the right, the test results are visible on the monitor screen.

Inspection inline setup_2 small.jpg Inspection inline setup_1 small.jpg The final version (see drawings) is built and tested in production.
For the final version, a digital camera, two red high power led arrays (Luxeon) and a double very thin diffuser at the front are used.


Software

In case a good contrast is available a normal template matching detects the individual characters. The match percentage in this case is very high. In the software, an selection mechanism is also available, so matched characters outside a certain area are rejected.
When the contrast is very low, an extra check is done using another set of templates.
Still it is possible that a character is not readable.
Number 5.jpg Number 3.jpg Worse number 1.jpg Good number-1.jpg When characters look alike, e.g. 5 and 3, a sub-image (template) of the different parts of the characters is necessary. In fact, after recognizing these types of characters, an extra template match on that particular area is executed.
The numbers 1 in the left picture do have a lower contrast then the numbers 1 in the right picture.
Remark: this code is not provided with a check digit like the Postnet code. Adding this kind of check possibilities results in a more robust detection of the code.