WPIcal documentation initial draft #2691
Conversation
|
Great stuff so far overall! Great to see this coming together. |
source/docs/software/wpilib-tools/wpical/images/ChArUcoDetection.png
Outdated
Show resolved
Hide resolved
| .. image:: images/Calibrate.png | ||
| :alt: Calibrate | ||
|
|
||
| There are two options for calibrating your camera, :guilabel:`OpenCV` and :guilabel:`MRcal`. It is generally recommended to calibrate with :guilabel:`MRcal`, as it has slightly better accuracy and is more robust against bad calibration data. WPIcal uses a ChArUco board for camera calibration. You can generate a ChArUco board here: `calib.io <https://calib.io/pages/camera-calibration-pattern-generator>`_ |
There was a problem hiding this comment.
Any particular recommendations on board width/height? I noticed the gui defaults to 12x8. And whitch dictionary should I be using?
There was a problem hiding this comment.
(wpical seems to be hard-coded to 5x5?)
There was a problem hiding this comment.
Yes the calibration is hard coded to 5x5. I was testing with a 12x8 board and it seemed to work well. I haven't tested other widths or heights too thoroughly. The optimal dimensions will vary camera to camera right?
|
|
||
| Camera Calibration | ||
| ------------------ | ||
| To measure the distance between tags in a video, you need the camera's intrinsics to account for things like distortion. WPIcal allows users to upload a video file to calibrate the camera in the tool, or upload a JSON file with your camera's intrinsics |
There was a problem hiding this comment.
How should I capture a video file? I was able to do it on linux with gstreamer, but it was pretty non-trivial.
gst-launch-1.0 -vvv v4l2src device=/dev/video0 extra-controls="c,gain=25,exposure_auto=1" ! image/jpeg,format=MJPG,width=1600,height=1200,framerate=50/1 ! jpegdec ! videoconvert ! tee name=t t. ! queue ! autovideosink t. ! queue ! videoconvert ! avimux ! filesink location=output.avi
There was a problem hiding this comment.
I've been using a phone camera and then transferring the file to a computer to calibrate. For linux It will need to be transcoded. Do you think we should document how to capture the video for linux machines too?
There was a problem hiding this comment.
OBS also works well
- Add a V4L2 video capture device
- Change properties:
- Under settings->video change the base resolution
- Settings->output, not sure why i don't see avi on linux but it should be an option
|
|
||
| .. warning:: CalibDB calibrations have been known to fail when bad snapshots are taken. If your calibration data looks suspect, repeat calibration process again. | ||
|
|
||
| Custom JSON |
There was a problem hiding this comment.
Do you think there's any value in supporting the photonvision JSON format directly? It's fairly similar
{
"resolution":{"width":1600.0,"height":1200.0},
"cameraIntrinsics":{"rows":3,"cols":3,"type":6,"data":[894.6473376999364,0.0,793.0568907406183,0.0,894.5611200668925,667.7549963404546,0.0,0.0,1.0]},
"distCoeffs":{"rows":1,"cols":8,"type":6,"data":[-0.04029854641739141,-0.03601719361588292,-0.0016269824645254384,-6.609780019207605E-4,0.14767766830679485,-0.09429815218469685,0.057468937992229234,0.11574241865507603]},
"calobjectWarp":[6.226950423907065E-4,6.020992541813042E-5]
}
There was a problem hiding this comment.
You mean if teams want to use their PV calibrations and cameras for field calibration?
| Linux Users Only: | ||
| ***************** | ||
| Linux users must transcode all videos to MJPEG codec without an audio stream in an .avi file before uploading to WPIcal. This can be done with an ffmpeg command: | ||
| ``ffmpeg -i <video_file.mp4> -f avi -c:v mjpeg -b:v 4000k -an <video_file.avi>`` |
There was a problem hiding this comment.
Note that -b:v 4000k is totally arbitrary
There was a problem hiding this comment.
Do you think this is a good default?
There was a problem hiding this comment.
I have no idea, I just stole it from stackoverflow
Adds initial draft of documentation for new WPIcal tool: wpilibsuite/allwpilib#6915