EoW July 2012

Technical article

For print quality detection, it utilises a method called pattern matching. This method requires a template of a good print on that specific cable. The template is compared to each captured image during a run. If the two images differ beyond the desired threshold, an alarm will notify the operator and an error report is saved to the ongoing data log. The print defect will also show up on the display for operators to clarify. OCV and OCR algorithms are more complex and are an extension to the pattern matching technique. For OCV, the software will verify each letter of the print legend. Alphabets of different fonts are stored in the database. The desired print legend is broken down into individual letters of the same font. Each captured image is also broken down into individual letters. Each letter will be compared to the desired letter. If there are differences beyond the threshold, it is a print defect. However, the software does not know what the misprinted letter is. OCR is the most advanced. The algorithm is very similar to what an operator would be able to detect with the eyes and mind. Not only would the software find out if a letter is different from the desired print legend, but it will be able to recognise the misprinted letter(s). This helps identify exactly what the mistake is and immediate actions to fix the system can be done. Since the PV system uses machine vision, it is able to inspect the print legend for any type of products. With proper set-up of the optical components, the system can be used on bare wire, cables, pipe, tube, tapes, flat sheets, etc. Suitable for any Type of Printers/Fonts The PV system uses image templates as a reference. The references fonts or print legends are predefined at the start-up. Regardless if the print legend or fonts are done by an inkjet, inkwheel, hot foil or laser printer, the software will be able to compare the print legends to the references. Reduce Defects On-The-Fly The PV system provides the operator with a view of the current print legend as well as the most recent print defects. This continuous stream of real time information – especially failure data – allows the operator to isolate the causes of print defects while the production line is still running. If the print quality is degrading below the threshold, the operator can immediately adjust the printer to improve the quality. Also, operators can use the PV system at the beginning of the run to set-up the printer and check to ensure 3.3 Improving Quality Assurance Suitable for any Type of Products

Multi-camera systems may be needed to handle cables with several print legends. For example, wires with print legends on opposite sides require a two-camera system. The camera and lens provide a field of view to cover the print legend ( Figure 2 ). Image capture is synchronised with the printer and line speed. The image will be used for further software processing. The lighting can vary depending on the application. The factors in consideration are the reflectivity of the material, surface profile, colour, line speed, and lens aperture. To ensure consistency in the images, the setup features an enclosure to prevent other lighting, dust particles, and other factors from affecting the results. Vibration of the cable is also minimised by wire guides to prevent images from being out of focus. One other challenge is the location of the PV system and to ensure the print stays within the camera’s field of view. The recommended location for the PV system is directly after the printer, where the print will be in the correct spot consistently. Placing the PV system further down the line will result in a high chance of the cable twist; which causes the print to be anywhere around the cable and outside the field of view. 3.2 Software Advance algorithm is used to process the images for print defects inspection. There are a number of algorithms: print quality detection, optical character verification (OCV), and optical character recognition (OCR). Different algorithms are implemented depending on the requirements and budget.

verify lengthy print legends until the next print shows up. As well, the flashing light strobes can be a nuisance to the operator. 2.2 Offline Viewing through Magnifying Lens One technique to detect print defects is having the operators to look through a magnifier. Although the operator can see the print defects and categorise them accordingly, there are a lot of limitations. The operator can only read the print up to a certain speed. This technique is done offline where the line speed is slower or zero. This process is very tedious and time consuming; so it is usually for customers who require this type of inspection. As print legends can be lengthy, there is a high chance of human error during the inspection. Also, the threshold of each operator is different and print quality is judged differently. 2.3 Image Capture and Sampling This process involves a camera acquiring an image based on an encoder signal and sampling rate. The image is then shown on a display for the operator to process a sample of images to determine if there are any print defects. Some main limitations are fatigue, operator’s thresholds, and slower processing speed. There is also a chance of missing some print defects depending on the sampling rate.

3 Machine Vision

Inspection System

3.1 Optical Mechanical Design The print verification system ( Figure 1 ) consists of a camera and adequate lighting to capture images of print legend moving at speeds more than 1,400ft (460m) per minute.

▼ ▼ Figure 2 : Print View SystemOverview

Alarm

Software

Lighting

Camera

Display

Cable

Direction

64

EuroWire – July 2012