Positioning Pinmame/FlexDMD in FullDMD image

[francisdb]

There is an existing tool that does this:
https://github.com/Le-Syl21/dmd-config

It uses Open Computer Vision to find the hole. However we probably don't want to depend on that and could try a more naive implementation:

• Divide the image in squares of equal aspect ratio as the image, to be seen how many divisions are needed
• For each square start at the center, read a pixel, move up down left right (cross) until you find a pixel that does not equal the original pixel + threshold.
• For all the resulting squares take the biggest one (there will probably be duplicates)
• Do some checks to make sure we have a square att least x % of the original image

Then for config

• read vpinball ini
• find b2s fulldmd position
• get the scale for the fulldmd image by comparing dimensions
• scale the biggert square we found
• configure PinmameDMD and FlexDMD windows according to the found scaled square.

see also vpinball/b2s-backglass#70

[francisdb]

vpin-studio also uses computer vision: https://github.com/lessthanoptimal/BoofCV

https://github.com/syd711/vpin-studio/blob/main/vpin-studio-server/src/main/java/de/mephisto/vpin/server/dmd/DMDPositionDetector.java