Machine Vision
set coordinate system
Hello,
please be a little bit more clear about your problem:
- what is your goal?
- what have you tried so far?
Best regards,
K
Hello,
an example image would be good... Are the grayscale and binary image the same? Or is the position of the centroid always the same, and only the positions of the circles are shifed? Are the circles already detected?
If I understand correctly, just use the (u,v) coordinate of the centroid as the origin. You can simply calculate the distance between the centroid and the circle centers (subtraction of the horizontal/vertical coordinates).
Best regards,
K
Hello,
again, please post the images, since this would consideraby help finding a solution. You can create the sample images using an image editor, if you don't want to post the original images. It is hard to understand your problem in words. And picture is worth a 1000 of them. 🙂
Best regards,
K
Hello,
your images have a good contrast, so you can try using pyramid based pattern matching. I have tried this in the Vision Assistant and get good correlation scores (> 800) for all holes with the match time of ~50 ms and ~150 ms (with rotation enabled).
After you detect the position of the circles, you can perform additional processing in order to refine the accuracy. Or just threshold the area of the match and find the center.
It would also make your detection more accurate if you place the camera perpendicular to the inspection area and additionally calibrate for perspective distortion.
Also, check some Vision processing examples that can be found under the Help->Find examples in Labview.
You could also
Additionally, your images have only two circular holes, so another possible (roboust) way to detect them is the Hough circle transform. Vision does not have such library, but it is possible to implement this as a dynamic link library (using OpenCV for example: https://decibel.ni.com/content/blogs/kl3m3n/2014/05/08/detect-hough-circles-in-labview-using-opencv)...
Best regards,
K
