I am trying to implement some image recognition on a test machine for an LCD screen, and having trouble recognising defects. Is anyone well versed in vision assistant that may be able to assist? I've attached a template image to this post. The image should display several coloured stripes, but occasionally the screen will malfunction and display completely black or white. I have used a simple color pattern matching step in VA to identify this successfully.
But now I need to add some more steps afterwards that can essentially 'scan' the image for any defects. These defects are usually small blemishes like a dead pixel or small crack. The end aim is to have the script return a couple of scores - one that identifies the striped pattern (and should score high), and one that identifies defects (and should score low, indicating none present).
I think a big problem is the quality of the image. Despite using a good camera, the images of the LCD screens are very pixelated and I'm struggling to get any improvement. You'll see this in the template below. I've tried to add pattern matching steps that look for a defect with an ROI over the entire image, but it is picking up false 'defects' that don't actually exist. I also have the ability to change the screen from striped to a block colour, so I thought that may help, but due to the pixelisation VA is still picking up false defects.
If anyone is able to help, have a look at the attached images and let me know if you can think of ways to improve the quality/uniformity of the image, and to identify these defects?
