Note: In order to maximize calibration quality, **do not reuse the same video sequence for both intrinsic and extrinsic parameter estimation**. The intrinsic parameter calibration should be run camera by camera, where each recorded video sequence should be focused in covering all regions of the camera view and repeated from several distances. In the extrinsic sequence, this video sequence should be focused in making sure that the checkboard is visible from at least 2 cameras at the time. So for 3-camera calibration, you would need 1 video sequence per camera as well as a final sequence for the extrinsic parameter calibration.
### General Quality Tips
1. Keep the same orientation of the chessboard, i.e., do not rotate it circularly more than ~15-30 degress. Our algorithm assumes that the origin is the corner at the top left, so rotating the chessboard circularly will change this origin across frames, resulting in many frames being rejected for the final calibration, i.e., lower calibration accuracy.
2. Cover several distances, and within each distance, cover all parts of the image view (all corners and center).
3. Save the images in PNG format (default behavior) in order to improve calibration quality. PNG images are bigger than JPG equivalent, but do not lose information by compression.
4. Use a chessboard as big as possible, ideally a chessboard with of at least 8x6 squares with a square size of at least 100 millimeters. It will specially affect the extrinsic calibration quality.
5. Intrinsics: Recommended about 400 image views for high quality calibration. You should get at least 150 images for a good calibration, while no more than 500. The calibration of a camera takes about 3 minutes with about 100 images, about 1.5h with 200 images, and about 9.5h with 450 images. Required RAM memory also grows exponentially.
6. Extrinsics: Recommended at least 250 images per camera for high quality calibration.
### Step 1 - Distortion and Intrinsic Parameter Calibration
1. Run OpenPose and save images for your desired camera. Use a grid (chessboard) pattern and move around all the image area.
1. Quality tips:
1. Keep the same orientation of the chessboard, i.e., do not rotate it circularly more than ~15-30 degress. Our algorithm assumes that the origin is the corner at the top left, so rotating the chessboard circularly will change this origin across frames, resulting in many frames being rejected for the final calibration, i.e., lower calibration accuracy.
2. Cover several distances, and within each distance, cover all parts of the image view (all corners and center).
3. Save the images in PNG format (default behavior) in order to improve calibration quality. PNG images are bigger than JPG equivalent, but do not lose information by compression.
4. Recommended about 400 image views for high quality calibration. You should get at least 150 images for a good calibration, while no more than 500. The calibration of a camera takes about 3 minutes with about 100 images, about 1.5h with 200 images, and about 9.5h with 450 images. Required RAM memory also grows exponentially.
# - Note: If your computer has enough RAM memory, you can run all of them at the same time in order to speed up the time (they are not internally multi-threaded).
# Potentially more accurate equivalent for the calibration between cameras 1 and 3: If camera 3 and 1 are too far from each other and the calibration chessboard is not visible from both cameras at the same time enough times, the calibration can be run between camera 3 and camera 2, which is closer to 3. In that case, the `combine_cam0_extrinsics` flag is required, which tells the calibration toolbox that cam0 is not the global origin (in this case is camera 1).
# Note: Wait until calibration of camera index 2 with respect to 1 is completed, as information from camera 2 XML calibration file will be used:
3. Hint to verify extrinsic calibration is successful:
1. Translation vector - Global distance:
1. Manually open each one of the generated XML files from the folder indicated by the flag `--camera_parameter_folder` (or the default one indicated by the `--help` flag if the former was not used).
2. The field `CameraMatrix` is a 3 x 4 matrix (you can see that the subfield `rows` in that file is 3 and `cols` is 4).
3. Order the matrix in that 3 x 4 shape (e.g. by copying in a different text file with the shape of 3 rows and 4 columns).
4. The 3 first components of the last column of the `CameraMatrix` field define the global `translation` (in meters) with respect to the global origin (in our case camera 1).
5. Thus, the distance between that camera and the origin camera 1 should be (approximately) equal to the L2-norm of the `translation` vector.
2. Translation vector - Relative x-y-z distances:
1. The 3x1 `translation` vector represents the `x`, `y`, and `z` distances to the origin camera, respectively. The camera is looking along the positive `z` axis, the `y` axis is down, and the `x` axis is right. This should match the real distance between both cameras.
