04-21-2008 08:55 PM
04-21-2008 09:04 PM
It depends how "round" you need to check.
Try creating a circle with a 10 pixel diameter, and you will see how rough the edge looks. Is this good enough for your application? If you start with a high resolution image and reduce the resolution, it should give you an idea of what the edge pixels will look like.
Bruce
04-22-2008 09:06 PM
04-22-2008 09:18 PM
If you are looking for roughness on the edges of the holes, it won't be nearly enough resolution. You would need 10x resolution or more to see details along the edges.
If you are just looking for gross errors, such as a partially blocked hole, it will probably be okay. The difficulty is that unless you are looking at the hole straight on, you won't know how big it should be to start with. At the sides, perspective makes the holes look like narrow ellipses. Are you going to rotate it so all the holes are inspected straight on?
Bruce
04-24-2008 02:38 AM
Hi Bruce
Current solution is to rotate the unit by a step motor for 10 steps(two columns per step), I would like to take 10 photos(totally 20 columns), for one photo two columns of holes are approximately regular round because these two columns reside the center of the body. Then to look for lighting holes as objects on ROI(two columns). Then caculate the area of every hole to match the limits(If I do not need to check the edge)? According to your suggestion there are more than 10 pixels on the edge? If yes, my 1024X1360 will not be enough for this I think:(