Hi
I'm experimenting with some "green screen" vision to reliably recognize green 3D printed features agains any other objects (MaskHsv).
This works quite well if I switch on the automatic white balance of the camera, where a "natural" color look can be achieved.
Hower the automatic white balance obviously adapts to the scene. If a large area of a certain color is visible, the balance changes. For example if the green object that I want to recognize covers a large area, the neutral metal background becomes slightly magenta-tinted (magenta is the complementary color of green). That's actually good to distinguish the green object. Unfortuately the auto-balance takes time to adapt, so I must use a very large settle time to get acceptable images.
Therefore I'd like to switch off the automatic white balance setting (like I did with exposure) and choose a fixed manual value. However whatever I try I cannot achieve a natural look with the manual White Balance setting. Far from it actually, it's so bad I can forget reliable MaskHsv because the whole picture becomes greenish-yellowish or greenish-blueish. It never even comes close to magenta-tinted.
I'm also confused how white balance can be achieved at all with just one ruler. From a math standpoint, shouldn't there be three?
I also tried to combine the Hue and White Balance rulers but I never achieved a good color look. Tests with OSB Studio resulted in the same limitation.
Is this a problem of the ELP camera? Or do other cams exhibit the same limitation?
Any tips?
Thanks,
_Mark