How viable is vision technology on the plant floor? On the surface, when you look at the technology and the capabilities it seems like vision should be as common as the programmable logic controller (PLC) and the human-machine interface (HMI). If you think about the technology and its ability to “see” the environment and make decisions based on what it sees, the applications are boundless. Despite its clear advantages, the use of vision technology on the plant floor is not as commonplace as most people would imagine. Why is that?
I believe the main reasons are: 1.) The supporting technology behind the camera; and 2.) Camera installations are often viewed as being not very robust.
If you know vision, you know that lighting and lensing are the keys to a successful camera installation. Get that right and things work marvelously. Get it wrong, and everything grinds to a halt.
Therein lies the problem. Generally, when you buy a vision system, the brilliant folks you bought it from will come out, set it up, and everything works great. They leave and the camera gets bumped, or parts start feeding differently, and everything goes haywire. Then come the questions: Did we get trained on how to calibrate the camera? Do we have the software tools to adjust for this shadowing? Did anyone grab the business card from the guy who installed the camera? We need him out here, now!
To avoid this scenario, it's important to take the steps necessary to put a solid system in place behind the camera to build robustness into the vision system. This typically involves going beyond the software package and tools that may come with the camera system and that you can program yourself. I'm talking about ensuring that the camera is tailored for your application and environment.
This level of robustness, however, does come at a cost. More advanced analysis tools will buy you more flexibility in part and camera placement, as well as more tolerance to variations in lighting. All of these factors lead to a more robust system, which means your system is up and running and bringing the return on investment you were counting on when you bought the system.
With this end goal in mind, where do you start? How do you gauge what is a good fit for vision?
Here is a quick list of some applications to get the wheels turning:
• Part and package inspection: Look for the presence or absence of specific features for this type of inspection. Do not let part presentation be a showstopper for you. With the advanced tools of vision, many of these issues can be overcome even with randomly placed parts.
• Part identification: The system should be able to identify the part type in mixed-feed applications to sort them.
• Part location and orientation: The camera should be able to identify the part in 2D or even 3D space to be acted upon by an ancillary system, e.g., drilling, grinding, painting, picked up (bin-picking) or rejected.
• Part state: The ability to identify the heat signature of a part is important. Can the vision system determine: If the part has been consistently heated? How long it has been out of the oven? Did the part reach curing temperature?
Another critical factor to consider is the environment. It is important with vision technology that there be a certain level of cleanliness, or at least the ability to keep the camera lens clean so that it can "see" the part. If in doubt, contact a system integrator with vision expertise who will know what to look for in qualifying an application. A good source for qualified integrators is the CSIA Exchange.
If you are not yet considering the use of vision technology, there is a good chance your competition is and gaining a competitive advantage over you. With the advance of laser scanning, near-IR and millimeter wave camera technology coming into the industrial scene, the competitive gap between those using vision technology and those that are not only promises to widen.