11-06-2012 03:38 AM
I want to detect an extra object present in my product but it has the same color as the background. using thresholding and particle analysis i got the detection but the resulting values are too small that sometimes this procedure leads me to make the false rejection for my products.
In the attachment one of the good image and a bad image i.e. image with the extra object are provided.
11-07-2012 02:32 PM
It looks like the defective side has a distinct bright line. If most defects are similar to this, you could use a Find Edge tool and just set the limits so if any edge is found, call it a reject. If no edges are found, then call it good. This really depends upon how consistant your background is.
The other option might be to count pixel values in the 200 to 255 range with a region of interest (ROI) set in this area. Then, set a limit based on a good part background (say no more than 20 pixels found in a ROI that contains 500 or so total pixels). I've used this method in the past, but the set limit is quite sensitive to the area of the ROI is low contrast conditions.