VOXL2 camera cal looks wierd

  • Is this normal? The video tutorial doesn't look like this at all, with the weird 2-bit colour over-saturated look.
    It's extremely hard to get it to detect the corners and draw those rainbow lines. I might try for like 2 or 3 minutes, holding different angles and positions, then it will flash the rainbow lines for and instant and I lose them again.


  • I should mention that it's not always overexposed like that. Obviously it wouldn't be able to identify the checkerboard with the edges blown out like that, but even when exposure is good, it doesn't work.

    Why is it so overexposed?

  • The dynamic range is a little wide, but the normal tracking camera image looks quite fine


  • That's to help you see what the checkerboard detector sees this "thresholding" takes the 0-255 image and setting each pixel to 1 or 0, making it significantly easier for the CV lib to find the checkerboard.

    The system has always done this, we just recently added it as default on the overlay because it makes it much easier to debug when you can see what the checkerboard detector is actually looking at. If it's having trouble finding the checkerboard, it's usually a lighting condition issue, make sure that your board is relatively matte (standard printer paper is usually pretty good), and you have it well lit but not directly reflecting the light at the camera as you'll run into issues where the center of the light completely washes out a section.

  • Oh I see. Ok it's on standard printer paper, I'll try to get some more diffuse lighting...

  • For instance: the lighting conditions in our office are good and the thresholder ends up showing a very clear image of the checkerboard:
    Screenshot from 2023-01-25 10-00-44.png

  • Yes my image is MUCH noisier than that even when it's not over/underexposed. Lighting is bright though...
    On another topic, should I be using --fisheye for the standard tracking cam?

  • Yes! The standard tracking camera we ship is very fisheye, I'd be incredibly impressed if anyone was able to get a successful cal with the normal pinhole model.

  • Haha ok ^_^

    We moved out of the office to an overcast day outside. The image of the checkerboard looks good, how come it can't detect it?



  • I'd recommend putting a strong white border around your checkerboard (at least half a square). If you look at mine we've just taped over the outer half of the outside squares with white tape, that also significantly helps the checkerboard detector when there's a well defined outer square.

  • Also it looks like you've specified a 6x8 board to the calibrator, but your board looks like it has 6x9 interior corners. This technically wouldn't matter too much since it'll find a valid 6x8 within that, but it'll run faster if it's not given 2 very similar solutions to have to choose between.

  • @Alex-Gardner oops silly me!! The latency has been awful on this latest calibration, on the order of 2s, so that is probably a major contributor! Thanks I'll try with white tape an 6x9

  • yessss 6x9 and a white background both improved detection rate, not just latency. Thanks for your help!

Log in to reply