26 Sep 2011
High speed image processing techniques and verification are well-established and can be used to check text, 1D bar-codes, 2D datamatrix codes individually or in various combinations. In general, the data being checked on any individual batch does not change – each label, or product within the batch contains the same information. However there are many applications, such as bank notes, cheques, lottery tickets, software security labels, mail out documents and other 'content critical' items where the data changes from label to label. Whether this is a sequential or non-sequential change, it represents a significant challenge to the image processing system. As in more traditional applications, variable data verification systems must ensure product integrity and eliminate the production of expensive waste. In order to do that, the system is required to provide 100% inspection by reading and verifying label numbering at web speeds in real time and provide a guarantee of print quality by rejecting incorrectly printed characters.
Any industrial vision system consists essentially of the same components: an appropriate light source to illuminate the area of interest, a camera (or cameras) and frame grabber to capture an image, a PC to process data, suitable image processing software and appropriate interfaces into the printing machine for triggering of the imaging system and for a reject mechanism.
Capturing images at high speed requires a fast shutter speed to eliminate blurring, this means that there is very little time to sense the light reflected from the document, and ambient light will produce a signal from the camera, which is too small to be of any use. The camera signal can be boosted to a useable level by illuminating the document with a very intense light.
To maintain registration and obtain full image access, it is necessary to install the camera and lighting assembly in the delivery section of the press. Ideally the cameras should view the sheet on a flat surface to minimise distortion. High resolution low distortion lenses are chosen to give the correct working height for the cameras taking into account the area available in the press. The H.S.N.V.S. (High Speed Number Verification System) from Verification Systems Ltd is a high performance, multiple-stream inspection system that provides 100% checking and verification of sequences and print quality of numbered documents on high speed printing and collating machines. H.S.N.V.S. has multiple high speed triggered cameras (one per stream) that capture images from the moving document. Printed data is decoded and checked for correct quality and placement. All of the imaging components (cameras, lenses, framegrabbers etc) for the system are supplied by STEMMER IMAGING, Europe's largest independent supplier of imaging components to OEMs.
Key to the performance of the system is the custom-designed software. The software developed for this product and others in the range uses high speed pattern recognition modules designed specifically for alignment, optical character recognition (OCR) and verification from STEMMER IMAGING's Common Vision Blox imaging toolkit. The OCR software allows the training and recognition of objects of all shapes. Powerful neural net technology gives the tool high processing speeds and high recognition reliability even in poor quality images.
The Tool is trained with an interactive process, in which examples of both good (positive) and bad (negative) features or characters are trained, ensuring robust distinction between good and bad components. The result of the learning process is a "Classifier" which describes the characteristics of the learned objects.
During the training phase, the OCR algorithm recognizes the properties of the marked objects and verifies them using the non- marked image areas. The Classifier concentrates only on the properties which describe the object, leading to high recognition speed and reliability. The algorithm is extremely insensitive to variations in object illumination. Noise interfering with the object also represents no obstacle to recognition. Objects can be automatically learned at different angles of rotation and at different sizes.
Special functionality is provided for the creation of very efficient OCR/OCV Classifiers. The system can be taught to read all standard fonts including OCR-A, OCR-B, E13B, Gothic and CMC-7 as well as barcodes and symbols. New or non-standard fonts can also be quickly taught and stored on the system hard disk drive for later use.
A variety of calculations can be required for variable data verification. These include:
Other requirements may be to check for OCR Quality, OCR / symbol print contrast (for over and under inking trends), alignment and check digit. In practice, the high speed triggered camera(s) capture images and record data from multiple fields for analysis. Ideally, each field can be assigned to search for a different style font, barcode or unique symbol, pattern or logo.
Figure 2 shows a screenshot from a system measuring 'banknotes' with the live image feed from the cameras. Figure 3 shows a typical report screen for this inspection, in a format representing the layout of the product. The blue text is the calculated number and the green text the actual number extracted from the images. Figure 4 shows another live feed, but this time, one of the serial numbers is positioned incorrectly, and this is flagged red on the image, even though the number itself is correct. Figure 5 shows a results screen highlighting that the errors have occurred only on Stream 1.
To enable the appropriate interfacing of many different types of presses and transport systems, either Web or Sheet fed, provision must be made for all the necessary timing and interface signals between the system and the customer application.
Camera trigger signals are normally generated from Document sensors and/or an optical shaft encoder. These signals are used to trigger the camera at the correct place on the web/sheet.
The system should also provide output signals for alarms and lamps etc. and interface with customer stop circuits and PLC's. A signal can be generated for each fault detected which can be used to reject the sheet and mark the defective document.
Reports should show the time and date of all sequences inspected and details of all faults detected. By recording the data for each fault, document re- makes can be run at a later time. In the event of a serious printing error, an alarm signal should be generated, typically using an alarm beacon to warn the operator.
STEMMER IMAGING has been one of the leaders in the machine vision market since 1987. It is one of Europe's largest technology providers in this field. In 1997 STEMMER IMAGING presented Common Vision Blox (CVB), a powerful programming library for fast and reliable development and implementation of vision solutions, which has been deployed successfully throughout the world in more than 40,000 imaging applications in various industries.