What to look for with implementing your vision inspection system
There are four main areas to think about when implementing a vision system: image sensor resolution, sensor lens selection, lighting source and predictable part presentation. We take you through them here.
1. Resolution of image sensors
Image sensors convert the light they collect into electrical signals, which are then digitised into pixels (as an array of values). The vision system processes these pixels as it inspects. With smart cameras, image sensors can be integrated into the system; they can also be integrated into a camera that attaches to the system. The number of pixels in the sensor impacts the resolution, hence the precision of the inspection.
Standard VGA (video graphic array) sensors have 640 wide x 480 high physical pixels, and each is some 7.4 microns square. You can estimate “real world” units from these numbers. Vision systems use image sensors that are highly specialised (so not your average web cam, as a contrast). To make measurement calculations more precise and easier, it’s good to have square physical pixels. The vision system can trigger the cameras to take a picture centred on a part-in-place signal; the cameras also have fast electronic shutters and complex exposure capable of freezing most parts’ movement as they go down the production line.
There are lots of different resolutions and interfaces of image sensors, so different ones will suit different application. To inspect large parts, or even different surfaces of the same part, multiple image sensors are often used.
2. Selecting sensor lenses
The next thing to think about is the lens for the image sensor. Sensors need a lens that gathers the light reflected from the item being inspected to actually form the image.
The right lens enables you to see the field of view (FOV) you need; it also allows you to place the camera at a suitable working distance from the item being inspected. (A more precise definition of “working distance” takes the lens structure and camera body into account.)
You can estimate the lens’s focal length of the lens can be from the FOV, working distance and camera specifications. Theoretically, “focal length” is the distance behind the lens where parallel light rays are brought to focus, and is a common way to specify lenses. Other important lens specs include the type and amount of optical distortion the lens introduces and how closely it can focus.
3. Source of lighting
While human eyes can see pretty well in lots of different lighting conditions, machine vision system can’t. That’s why properly lighting the parts being inspected is so important.
Some vision algorithms can tolerate some light variations, but well-designed lighting just removes any uncertainty. Using ambient light (such as overhead lights) is generally a no-no, because it can vary over time, become dim or blocked, or beat and burn out. Light from adjacent windows can negatively impact your system’s robustness.
For more in-depth information about lighting, please see this blog on tips for lighting vision inspection solutions.
4. Presenting parts predictably
The fourth area to consider in implementing your vision system is how items will be presented for inspection. Part not presented consistently reduce the system’s effectiveness, so your system won’t do what you want it to.
As it moves past, the surface of the part you want to inspect must face the sensor. Take into account whether the part will be inspected as it moves, or if it will be stationary. It it’s moving, the motion will probably have to be “frozen”— this can be achieved by using the sensor’s high-speed electronic shutter feature (this is standard on most machine vision sensors) or by turning the light on briefly. There will also need to be a trigger to the sensor, so it knows when to take a picture. Generally, a photo-eye sensor generates the trigger by detecting the part’s front edge as it moves into the inspection area. For stationary parts, the sensor can be triggered to take a picture from a PLC or the robot (if robots are used to position the item).
Parts inspected at very high speed will probably need to be have their position optimised to reduce processing time. Something to be aware of with system design is that everything uses processing bandwidth, so for high-speed inspections, work out which of your needs are critical and which are just “nice to have”.
The cost of implementing your machine vision solutions on the factory floor will be returned multiple times through increased quality, production efficiency and scrap reduction. So work with a reputable supplier and don't be afraid to ask questions of them.
This blog gives an overview of a highly complex area. For more information or if you have any questions, please contact us via email
or call 1300 IQVISION (1300 478 474). We have years of experience customising applications from the simple to the complex.
iQVision also has a host of information in our resource library
, including case studies, whitepapers, FAQ’s, videos, our blog and brochures. They’re all free to download.