Chip Detection Algorithm

[Flo93]

Hello,

i would like to inspect surfaces and find anomalies. Pattern matching doesn't work well. Which tool would you suggest to use?

I attached original images and ones with a filter i would use labeld with "good" and "bad".

Thanks for your ideas

Sincerely

bad1_threshold.png

bad2_threshold.png

bad3_threshold.png

bad4_original.png

bad5_original.png

bad6_original.png

Good1_threshold.png

Good2_threshold.png

Good3_threshold.png

Good4_original.png

Good5_original.png

Good6_original.png

            

[1100fps]

From the set of images, it appears that the images that pass are (more or less) symmetric about the center of the disk.  One approach would be to unwrap the image  (Imaq unwrap) and then run a sliding average across the resulting image, comparing adjacent windows looking for changes above a preset threshold.  It's a guess.. Nothing is ever that easy.. 🙂  

[AmitShachaf]

I was working on semiconductors to detect all kind of issues like that.

I usually had good success with texture analysis. 

LabVIEW has good set of VIs for texture analysis with very little documentation.

 https://www.ni.com/docs/en-US/bundle/ni-vision-labview-api-ref/page/imaqvision/texture_pal.html

I strongly recommend to read some theory on the subject. But start with NI documentation.

I was using texture analysis to extract features. 

Then train classifier to detect the bad images.

All the best - Amit