Ask the Experts

Bill, I think it depends on to what extent you want a DIY project. If you buy a camera from a vendor with a basic SDK they'll likely have video viewer sample code. I know Basler and the Imaging Source do. To implement zooming and draw the target I'd do it in software. You can stitch in an external library to your application and leverage that code. I'd suggest OpenCV or Emgu CV (an OpenCV wrapper) to do that heavy lifting.

You could build it yourself or you could hire us (or someone else) to do it. For basic zooming and a target, it'd take a software guy about 5 days to put together in C++ or C#. You'd likely want to modify and rebuild OpenCV or Emgu CV to always put your target in the middle of your zoomed image.

You could also buy something more off the shelf like John suggests below and set that up. I really don't know if the exact product you want exists though. If it's a fixed station you need something by Scienscope could work.

Duane, That is possible. The system would need to be programmed to recognize a critical feature on each part, and then perform a set of inspections based on which critical feature was found. I don't know exactly what parts you have, however let's say you had gaskets of various sizes you wanted to check for cracks, and couplings you wanted to check for wear and cracks and fuel valve assemblies you wanted to check for completeness. A system could be programmed to recognize certain gaskets and look for edges and irregularities (cracks), recognize certain couplings and measure wear (thickness) and recognize a fuel valve assembly and check each component was in the proper place.

One place you may run into issues is scaling this up as each part and the corresponding inspection(s) would have to be programmed separately. How many different potential parts are you talking about? Feel free to call or email.

Sanaulla, It sounds like the overall assembly is large but there are two smaller areas of interest. If this is the case your best solution is likely a two camera system. One camera aimed at one "toe" and other at the other "toe". A vision inspection running on a nearby computer would analyze both images and output the difference in the two alignments along with whatever else you needed. That said, we'd need to see pictures of the toe, the alignment to be measured and have some idea of the precision you need. Let me know if we can help you out. In general large assemblies aren't a problem, it's the size of the "area of interest" relative to the defect that matters most. If there are multiple areas of interest far apart from each other a multi camera system is likely best.

Tom

Marios, This is something we could write for you. We've built a handful of custom camera viewer applications for other customers.

We also have a GigE vision viewer we've built. We've not used it with Imaging Source GigE cameras but as it's a common standard it should work equally well.

Tom

Carl, We're always interested in expanding our machine vision lab. What do you have? Can you send pictures and a list?

Thanks,

Tom

Greg, If there are more than two or three types of blocks the easiest way to ID the blocks is going to be by adding a barcode or datamatrix to the blocks (if that's possible). This will also give you downstream serialization and traceability. Pattern matching and Identification via vision can get finicky when there's a lot of variety. Do you have any pictures of these blocks you can share as far as the other defects? Depending on the pictures I would recommend shipping samples to 1-2 integrators for some testing and take things from there.

David,

AIA offers professional certifications to integrators. I would select someone who has received an advanced certification and ask for customer references.

We'd hope you'd select us, but anyone on the list of Advanced integrators can help you.

Tom

Mike, Are there other types of green plant matter flowing through and the system needs to distinguish invasive from non-invasive? Alternatively can the system detect green matter and apply a rule of thumb for how much is likely invasive?

Jana, How many different products are to be identified? Would the device need to be handheld or are these products in board form on a conveyor?

We'd be happy to talk through a solution directly. I'm not aware of an off the shelf product.

Tom

John - This sounds like an interesting problem that would likely need an engineered solution. Would the end goal to be to get to a handheld unit an operator could use in field or would this unit analyze specimens taken to a lab.

To do accurate color analysis we'll want to use full spectrum white light. For the voids a low angle of light could be best as that's typically used to pick up scratches, dents and other non-uniformities in surfaces. We'd probably also need to come up with a calibration to make sure we're providing consistent readings over time.

Happy to discuss further. These are always interesting problems to talk through. Send me an email or call at 303-832-1111x101.