Just as human inspectors working on assembly lines visually inspect parts to determine the quality of work manship, so too can machine vision systems.
The technology is becoming increasingly popular among man ufacturing sectors which require visual inspections with high-mag nification, 24-hour operation and repeatability of measurements — most notably the food and bever age and automotive industries.
SAGE Automation senior sys tems engineer and vision team leader, Mark Dankiw, says a vision system essentially pro vides an unbiased and repeat able inspection but it will only be as good as the lighting and the rules established for it.
“The camera will only do exactly what you specify it to do. It won’t do anything more or any thing less,” Dankiw told Manufacturers’ Monthly.
“Companies will need to deter mine what criteria and toler ances are required in order to build up some rules which then become the program.”
He adds vision systems work best and are most reliable in batch processes where variables do not often change.
“You really need to contain the environment which means not changing anything, including lights. If the cell is open, you can’t even put up an extra little fluorescent light next to your cell,” Dankiw said.
“Because essentially it’s been programmed to certain rules and as soon as you change that envi ronment and those rules don’t deal with that, you’re going to be affecting reliability.”
Dankiw says when looking to install a vision system, compa nies first need to identify what they actually need from the solu tion which can often be done by talking to the process and quali ty assurance departments.
“They need to look at where the rejects and defects are and need to determine how much revenue is being lost in the process due to labour, material and lost production,” Dankiw said.
It’s all in the lighting
According to Dankiw, lighting is one of the most critical components involved in setting- up and operating a successful vision system.
There are three main lighting techniques: dark field illumina tion which is used for inspecting defects that are either raised or lowered; bright field illumina tion which works like a spot light to illuminate the object and pro vide an even light; and back lighting which is used to show the outline of a part. Backlighting is typically used in bottle cap detection or gauging applications.
“Dark field illumination requires the light to be situated at 90 degrees to where the cam era is looking and quite low to the part, meaning anything that is smooth on the object will look quite dark and any bumps or dints will look bright,” he said.
“You can still look for dints and bumps using bright field illumination but they will be dark instead of light.”
Dankiw also says speed and depth of field play an important role in achieving a reliable vision system.
“By depth of field I really mean how well the camera can keep a large object in focus. If you’ve got a stack of boxes on a pallet and you’ve got a camera on top, the top box might be in focus but by the time you get to the bottom layer it might be out of focus,” he said.
“You really have got to be aware of limitations and that’s where experience comes in: knowing what sort of equipment you need to select to actually solve the solution or problem,” Dankiw said.
Vision systems are also suitable for production lines which con tain product variables, but Dankiw says the technology nor mally has to be integrated into some sort of external input for it to work.
“Because not all production lines are batch product lines there might be instances where there are different parts so you might need to give the system some specific information about the part that it’s inspecting,” he said.
One such example is John Hart Automation and Robotics’ new multi-view integrated vision system which combines articulated robots with camera technology. The offering is an Integrated Vision (iR Vision) suite installed on Fanuc’s R-30iA robot controller.
According to John Hart Automation and Robotics automation sales manager, Matthew Plant, the system is suitable for applications including: pick and place; machine loading; palletising; part assembly; and welding.
He says generally one integrat ed camera will be sufficient but the robot controller does have the capability to connect to up to 32 CCD cameras or four 3D laser sensors.
“For larger parts the multi-view option will combine the images from two separate cameras and calculate an accurate part position, which cannot be achieved by just looking at one small area of the part through one camera,” he said.