• reduced size of A-area scan to the bottom part

• new A-area detection to detect only big cracks

• changed preparation

• Contrast from 5 to 8
• critical top from 9*9 to 11*11
• critical bottom from 11*11 to 15*15

• Unpump detection, thin fakes filtering

• ability to diusable analysis trough “error_analysis” parameter(“should be set to 0 to disable analysis)

• decrease unpump area on bottom
• connect walls trough gaps of 100 px during ridge localization

• Better corner detection

• overview detection. more strict rules
• better ridge detection

• More universal overview detection

• OCR stability improved for cases where ridge protect is too close to a die number

• OCR tweeking

• Fix OCR symbol overlap

• MLP-based OCR made specificly for LE font

• OCR improvement

• OCR improvement

• Include blurrness around the defect in C-area

• Added additional check for ridge height

• New ridge detection

• Overcoming complementary region bug again

Intersection of the region should be always 1px smaller on both axies

• New ridge detection
• new preparation

• ridge detection auto threshold

• preparation auto threshould adjustment

• Halcon bug workaround

During finding complementary region, halcon sometimes recognizes image size incorrectly. Added additional intersection to ensure that complementary region has correct size.

• improved corner detection

• new ridge detection

• new ridge detection

• Rescale parameters(sys_v_scale and sys_h_scale) added for rescaling the image during preparation
• scaleV parameter removed

parameters file 'parameters.json' should be placed under 'Documents\LaserVision\Recepies\G09E_gelpack'

content of the file should be

{  
"sys_v_scale": 0.998,
  "sys_h_scale": 0.986
}

• scaleV parameter for vertical unpump area scale

• Fixed 'red' graphical bug

• OCR changed font to Non-rejection

• OCR modifications

• brougth back the old C-area
• maderidge detection and preparation more resistant to blured areas

• cut small artefacts around chars

• Split characters into grid. Hashtag added is char is not recognized

• number analysis alowed characters check added

• number analysis added

• Top unpump ignore area is tied to bottom unpump area on fixed distance of 17750 px (+-20px window)

• fixed error in debug information

• Removed BTop ignore area for B-area
• increased defect windof from 8,4 to 9,4 for C-area

• added histogram equalization to get rid of brightness influence

• Changed defect windows from 7,4 to 8,3 for C area
• Chenged sensetivity of vertical convolution from -7 to -5 for C area
• Changed minimal defect size from 200 to 260 for B area

• Changed defect closing from 8 to 4 on C-area
• Added debug output for filtered defects

• Fixed ignorance area of prepared image
• Changed ridge detection to more brigther color
• unpump area detection added

• Overview die detection will give no results if die has non-proper cut

• New C-area algorighm

• Added additional check for the rare case if origin was found, but ridge don't

• Fixed B2 area problem
• changed top minimal size to 9×9
• changed bot minimal size to 11×11

• Fixed preparation

• Copy analysis from G08 to G09 for B-area

• Changed calculations of ignorance areas to base on origin point

• B1 changed to 14px from inner ridge edge
• B2 changed to 39px from center
• C1 changed to 24 from center
• C2 changed to 54 from center

• Changed the order of operations in C-area. Defects found, then C3 sustracted, then filter is applyed.

• Creterias for C2 changed from 12×12 to 36×36(+ area 200px)
• Width of C2 changed from 20px to 28px

22.11.18 - Top part size changed to 8172 px

19.11.18 - ridge detection correction. Find biggest piece, assume that it is part of the ridge, find the rest of the pieces vertucally aligned to it.

08.11.18 -

A area. Broaght back stain filter.

B-area. Criteria increased 9⇒10, 11⇒12

B-area. Split moved up from 2.4 to 2.7 mm

C-area. Criteria increased 11⇒12

26.10.18 - Added prev change to GIT and pushed

19.10.18 - Removed stain filter from A-area. Added LongLines