Apr 19 2006
Until now, only the human eye has been able to detect scratches and marks on ceiling panels or high-quality glossy paper. A rapid surface inspection system for spotting defects is now in use and will be presented at the Control international trade fair for quality assurance.
People spend a great deal of time in offices and ought to feel comfortable. Ceiling panels play an important part in this. They absorb sound, retain heat within the room or provide fire protection. Normally the ceiling panels just go unnoticed. But if the grain of the panel is flawed, the blemish stands out a mile. Manufacturers therefore go to great lengths to ensure panels are perfect when they leave the factory. Until now, only the human eye could spot the defects because it is so difficult to distinguish them from the design. Some cracks and holes are part of the design, other grooves are actual damage. To speed up quality control, a team of researchers from the Fraunhofer Institute for Industrial Mathematics ITWM in Kaiserslautern and experts from Aschaffenburg Technical University have developed a novel surface inspection system that automatically spots defects during production.
The system is made up of eight cameras and analysis computers. As soon as the ceiling panels come out of the dryer, they slip past the cameras in pairs on a conveyor belt. Four cameras "see" each panel: two inspect the surface as it is illuminated with oblique light, detecting deep scratches or cracks. The other two cameras look out for color defects using light striking the object vertically. Each camera is linked to its own computer that analyzes the image. This massive computing power means the image can be analyzed fast enough to keep up with production. The server – the ninth computer – collates the data from the computers, controls the camera system and raises the alarm in the event of a defect. The MASC – Modular Algorithms for Surface Control – image analysis software developed at the ITWM forms the backbone of the system.
"Analyzing the design and detecting flaws in the ceiling panels was a daunting challenge," project manager Markus Rauhut remembers: "Holes may well be part of the design but any bigger than a certain size and they jump out at you. We first needed to teach the system the nuances the eye can pick up." And to make matters worse, the manufacturer produces around 15 different designs. Consequently, the researchers had to load 15 different algorithms onto the system, one for each specific design, so the defects could be spotted. The manufacturer is now able to activate any of these algorithms at the click of a mouse whenever production shifts from one type of panel to another.
http://www.fraunhofer.de