Further calibration problems–green indicator problems and no red crosshairs

Home Forums AR Sandbox Forum Further calibration problems–green indicator problems and no red crosshairs

Viewing 15 posts - 1 through 15 (of 15 total)
  • Author
    Posts
  • #2309
    FOCW
    Participant

    Hello again,

    I’m still trying to achieve a decent projector calibration for our sandbox. I’m having problems with getting the yellow indicator to turn green. A secondary problem is that I don’t get the red crosshairs after the first set of tie points is collected. My guess is that the points are not good enough to do a calculation with, so solving the first problem may solve the second, but I don’t know.

    Onto the first problem. I uploaded some videos of trying to take tie points. The first three are for the first three points and the last two are for a fourth point. Note: A piece of cardboard blocked up to half the depth of our box is currently standing in for sand, which is on order. This cardboard was also used to take the corner points for BoxLayout.txt. They are included at the bottom of the post in case they’re useful. The calibration video (https://www.youtube.com/watch?v=EW2PtRsQQr0&t=610s) says a perfectly level surface is not necessary so I don’t think the sand substitute is the problem(?).

    Description of difficulties:
    In the calibration video, the indicator seems to turn from yellow to green quite easily and remain green over a reasonable height range. Dr. Reylo is able to keep the target parallel with the sand surface and keep it green while taking tie points at various heights. The only adjustment needed was to decrease the sea level for the below plane tie point.

    In contrast, I am generally having to raise the target quite high, tilt it, or both, to find a point where it turns green. When it does turn green it isn’t stable within a range or even a point. Some of the footage I link shows the target stationary while the indicator is flickering between colors.

    This definitely doesn’t seem right, and the calibrations I’ve attempted took over an hour each to get the minimum twelve points, the vast majority of which time was spent moving the target around futilely.

    Tie point 1: https://photos.app.goo.gl/x5bQf52YbgbeV1bR6 (target must be tilted)

    Tie point 2: https://photos.app.goo.gl/V8aLHLGbWxMNNAhE8 (target flickers)

    Tie point 3: https://photos.app.goo.gl/RxmroQR3Wqza8bBn8 (target is over a foot above the rim of the sandbox)

    Tie point 4: https://photos.app.goo.gl/r2RfFDinMfV4QqjT9 (target does not achieve green)

    Tie point 4 (2): https://photos.app.goo.gl/ompWFupoXWttZyyy7 (target won’t stay green unless raised high and tilted)

    Command window outputs for the tie points in the videos:
    cdm@admin-Sandbox ~ $ /home/cdm/src/SARndbox-2.3/bin/CalibrateProjector -s 1024 768
    CalibrateProjector: Capturing 120 background frames…VRWindow::VRWindow: Vertical retrace synchronization requested but not supported
    done
    CalibrateProjector: Capturing 120 background frames… done
    CalibrateProjector: Capturing 60 tie point frames… done
    Tie point: 204.5, 192.5; -17.239, -21.0238, -98.2385
    CalibrateProjector: Capturing 60 tie point frames… done
    Tie point: 409.5, 192.5; -9.21404, -13.2975, -84.4043
    CalibrateProjector: Capturing 60 tie point frames… done
    Tie point: 614.5, 192.5; 12.0585, -7.30491, -81.1033
    CalibrateProjector: Capturing 60 tie point frames… done
    Tie point: 819.5, 192.5; 19.4192, 0.522661, -55.114

    BoxLayout.txt outputs taken right before this calibration attempt (sea level changed from -90 to -105):

    (-0.0216684, -0.0394426, 0.998987), -105.8983
    ( -46.0027, -32.887, -103.57)
    ( 47.7056, -29.9049, -104.008)
    ( -48.19, 39.7827, -104.159)
    ( 48.9422, 38.9245, -101.912)

    I’m not sure what other information might be helpful. The projector is a BenQ MW632ST, the keystone hasn’t been adjusted. I didn’t do the glass checkerboard calibration step for the Kinect, as it said it was optional.

    #2316
    Sean Robinson
    Participant

    Thank you for sharing the videos. That makes it much faster to see possible causes.

    I believe the source of your calibration problems is that the handle on your target is too wide. The calibration software needs to “see” a circle floating in space and will not project green if fingers, box walls, sand, or other things are too close. Assuming the circle has a 12 cm diameter (CD size), your handle is about 1.6 cm across where it intersects the circle. Try replacing your wide handle with a stiff wire and keep at least a 5 cm gap between the circle and your fingers.

    Let me know if this works or not.

    #2317
    FOCW
    Participant

    Hello!

    This was the problem. I replaced the pipe strapping with a thin allthread rod and was able to get a green target. I took tie points at four height levels and two base plane levels (two heights for each) and got what seems like good calibration according to the red crosshairs following and lining up with the target throughout the sandbox. When running the sandbox program contour lines matched up with the objects I placed in the box.

    However, when I exited the calibration, the command window had the message:

    “Calibration error: Some tie points have negative projection weights. Please start from scratch”

    This message was also on all previous calibration attempts, but I had hoped fixing the target problems would resolve it. What does negative projection weights mean and how do I fix it?

    #2321
    Sean Robinson
    Participant

    Do you have sand? Or are you still using cardboard?

    I suspect the calibration error message is because all your tie points are above the “sand”. Once you are able to capture tie points below the sand surface level, I expect the error messages will stop.

    Look for the “calibration-error-some-tie-points-have-negative-projection-weights” topic on the old forum on the Wayback Machine for a hint from okreylos about tie points.

    (I would show the direct link, but my post fails when I include it.)

    #2324
    FOCW
    Participant

    The sand just arrived today, actually. Checking the archived forum now.

    #2344
    FOCW
    Participant

    Thank you for the tip, the old forum was very useful, although it looks like not all of it wound up archived. I had been taking each set of tie points at the same height. I will try varying within the set and hopefully that plus having our sand will do it.

    (This OP went into mod queue because of the links. Anti-spam measure, I guess.)

    #2345
    FOCW
    Participant

    Test comment. Will this post?

    #2492
    squigs
    Participant

    Im having the same issue above as FOCW. I cannot for the life of me get the target to be green. I have a new laptop I have set up everything up exactly as the guide goes and an older desktop that was working around 2018, but the desktop has the issue now as well. Nothing has been updated on the desktop and the calibration disc is the same one I created in 2015.

    Any help would be great!

    #2493
    Sean Robinson
    Participant

    Are you keeping your calibration disc away from other objects, including your fingers?

    Can you post photos or video a a calibration attempt? Seeing it might make it easier for others to see a problem.

    • This reply was modified 11 months, 4 weeks ago by Sean Robinson.
    #2500
    squigs
    Participant
    #2503
    Sean Robinson
    Participant

    Thank you for the video. I estimate your full disk has a diameter of about 15.25 cm.

    The target disk needs to be close to CD/DVD size (12 cm diameter) before the DiskExtractor will mark it as a valid target. If I’m reading the code right, any diameter between 10.9 cm and 13.2 cm should work.

    I suspect your target disk is too large. Try trimming the paper part of your disk to CD/DVD size.

    #2505
    squigs
    Participant

    Sean

    That was absolutly it! Thank you so much.

    When did this change, I used that disc for calibrating for years and years.

    thanks!!

    #2506
    squigs
    Participant

    Sean

    That was absolutly it! Thank you so much.

    When did this change, I used that disc for calibrating for years and years.

    thanks!!

    #2508
    Sean Robinson
    Participant

    It’s hard to put dates on releases for the Kinect package, but it looks like this was changed in 2015 for v3.2.

    Was your previous sandbox running SARndbox v1.6? Or had you used a 2.x version before mothballing?

    #2510
    squigs
    Participant

    That must have been the difference, I did switch to 2.0 at some point.

    Again, thanks for the help!!

Viewing 15 posts - 1 through 15 (of 15 total)
  • You must be logged in to reply to this topic.