yanshil / video-apriltags

Camera Calibration (for video)

1. Download Checkerboard for Camera Calibration: https://www.mrpt.org/downloads/camera-calibration-checker-board_9x7.pdf

   • Don't forget to print it using A4 Paper! Otherwise the size of squares is not corret.
   • Instead of number of cells, the 9x7 here actually means the points of intersection. So there is actually 10x8 cells, but the parameters filled in the code is (9,7)

2. Attach the printed checkerboard to a hardboard.

3. Hold the hardboard, record a video: Slowly move the board upside-down, left-right, turn around, turn with some angles etc... (just make sure all the cells are visible)

4. Once you get the video (i.e. data/90fov.mp4), you can run the calibration code

   

   • GUI has already been tested under Windows and Linux, but not Mac.
   • The recorded video will store the sampling frames inside dir_name.
   • Current sampling size is random_num = 20.

5. The code will then output the camera matrix. The format already been tidied as capable for copy and paste. This matrix will be used in next step

   

Tag Detection

1. Requirements:

   • duckietown-Apriltags bindings https://github.com/duckietown/dt-apriltags
     • pip install dt_apriltags
   • Others: pip install opencv-python numpy pandas scipy matplotlib

2. Print April Tags and attach it to the object you want to detect.

   • Pre-generated tag families: apriltag-imgs
   • The pre-generated tag is png format, and relatively small. You can scale it with the scriptscale_tags.sh. The script will scale all the png files under current folder.
   • You would need to modify tag_size (m) to the actual size of your printed tag. (Suggestion: You can use Word to accurately modify tag size)

3. Get Video Tags Trajectory

   • Modify run.py as your needs. For first-time running, set loadInfoDictFromPickle as False, and all the other config as True.
   • 

4. Run python run.py. You will get the trajectory image, txt and pkl

   

   • pickle file (pkl) will be useful when debugging and applying some other transformations.

OSError?

If met with OSError as the following during funning, add a parameter "searchpath" in Detector() like this

OSError: /home/yanshimsi/.local/lib/python3.8/site-packages/dt_apriltags/libapriltag.so: cannot open shared object file: No such file or directory

Try to find where your libapriltag.so is located first (Mine is in /usr/local/lib after doing sudo make install), and then use the Detector as

from dt_apriltags import Detector
at_detector = Detector(searchpath=['/usr/local/lib'],
                        families='tag36h11',
                        nthreads=1,
                        quad_decimate=1.0,
                        quad_sigma=0.0,
                        refine_edges=1,
                        decode_sharpening=0.25,
                        debug=0)