Release notes

• Problem with scaling fixed

• Preparation bottom ridge end search area. made narrower
• top unpump area. decreased required line width

• Decreased threshould of cross-detection to 40%

• New preparation based on 2 top-ridge parts
• scale parameters added

• Apply ignorance right after defect detection but before creteria filter

The problem was related to ignorance area was applied after creteria search. In simple steps, here is how algorithm was working:

1. Find defects
2. Filter defects by size
3. Remove defects under ignorance areas

This was leading to next situation: Filter was detecting big defect. Ignorance area was cutting a big part of it. Small part, which is left was still considered as a defect. Now I moved ignorance area before creteria. So if there is small part left, creteria will cut it out.

• Top unpump detection now tryes to also find the line instead of just calculating distance

• display ignorance areas

• changed greedenes of the cross-search algorithm

• thread error as full-screen B-defect

• Removed C-area
• Extended B-area to cover old C-area

• Disable D-area

• Alignment changed from left to center

• Increase the size of top unpump ignore area

• Top unpump ignore area is tied to bottom unpump area on fixed distance of 13090 px (+-20px window)

• Decreased cross detection score from 80% to 70%

• increased area size of top extended area to 200
• increased area size of bottom extended area to 300

• added versioning
• broaght back VarThreshold instead of vertival convolution