Machine Vision
Automated barcode recognition from an image with NI Vision.
Hello JY,
Without really delving into it myself, I can only offer a few suggestions.
1. How do I "teach" the machine to recognize barcodes as a "type" of thing I want to recognize in an image (what sort of functions and VI's would I use) so I can do an automated inspection similar to pattern matching and geometric matching?
As for "finding" something that looks like a barcode, I think Pattern Matching is going to be your best bet. But perhaps Shape Matching would work as well.
Once you find what could be a barcode, you could mask that ROI and attempt to read the barcode.
2. With the instances found, how would I retrieve the ROI information from each of the barcodes found so I can separately display them?
Pattern Match will report the X,Y center of your pattern. Of course only a perfect 1000 score match could you expect your X,Y to be exactly in the center of your match. But it should be fairly close. If you know the expected size of your barcode, you could extrapolate it from the center point.
Good luck with your project.
Without really delving into it myself, I can only offer a few suggestions.
1. How do I "teach" the machine to recognize barcodes as a "type" of thing I want to recognize in an image (what sort of functions and VI's would I use) so I can do an automated inspection similar to pattern matching and geometric matching?
As for "finding" something that looks like a barcode, I think Pattern Matching is going to be your best bet. But perhaps Shape Matching would work as well.
Once you find what could be a barcode, you could mask that ROI and attempt to read the barcode.
2. With the instances found, how would I retrieve the ROI information from each of the barcodes found so I can separately display them?
Pattern Match will report the X,Y center of your pattern. Of course only a perfect 1000 score match could you expect your X,Y to be exactly in the center of your match. But it should be fairly close. If you know the expected size of your barcode, you could extrapolate it from the center point.
Good luck with your project.
Pattern matching can be very tricky. Especially if you have inconsistent image quality or poor lighting.
If you can narrow down the search areas that will help. But if you're looking for multiple hard to match patterns in any number, placement, or orientation, you've working on a tall order.
Pattern matching is usually about the highest "level" tool of any of the machine vision solutions I've worked with. Mileage varies.
In past LabVIEW projects I've used a few tricks to improve results. Like using up to a dozen similar images as patterns when results were low. I would run through all of them and pick the one with the highest score. Probably not the most elegant solution, and definitely not the fastest. But eventually it worked well enough for that application.
If you haven't already read through it, I'd suggest having a look here. http://digital.ni.com/manuals.nsf/websearch/80CC7F2386F7701F8625731500701D23
Some good reading there. As well, it really is worth going through the examples both in Vision Assistant and LabVIEW. If you have Vision Builder, go through those examples too. I've found a lot of good stuff there when I needed it.
If you can narrow down the search areas that will help. But if you're looking for multiple hard to match patterns in any number, placement, or orientation, you've working on a tall order.
Pattern matching is usually about the highest "level" tool of any of the machine vision solutions I've worked with. Mileage varies.
In past LabVIEW projects I've used a few tricks to improve results. Like using up to a dozen similar images as patterns when results were low. I would run through all of them and pick the one with the highest score. Probably not the most elegant solution, and definitely not the fastest. But eventually it worked well enough for that application.
If you haven't already read through it, I'd suggest having a look here. http://digital.ni.com/manuals.nsf/websearch/80CC7F2386F7701F8625731500701D23
Some good reading there. As well, it really is worth going through the examples both in Vision Assistant and LabVIEW. If you have Vision Builder, go through those examples too. I've found a lot of good stuff there when I needed it.
