CameraCalibration

Contains intrinsic and extrinsic calibration parameters for a camera.

Panel support

CameraCalibration is supported in the following panels: 3D and Image.

Schema

Field	Type	Description
`timestamp`	`time`	Time at which the calibration parameters are valid.
`frame_id`	`string`	Reference frame of the camera. The origin is the optical center. +X points right, +Y points down, and +Z points into the image plane.
`width`	`uint32`	Image width in pixels.
`height`	`uint32`	Image height in pixels.
`distortion_model`	`string`	Distortion model used for the camera lens.
`D`	`float64[]`	Distortion coefficients, format depends on the chosen model.
`K`	`float64[9]`	Intrinsic matrix of the raw (distorted) image, in 3x3 row-major format.
`R`	`float64[9]`	Rectification matrix for stereo configurations, 3x3 row-major format.
`P`	`float64[12]`	Projection matrix of the rectified image, 3x4 row-major format.

Distortion models

The following distortion models are supported:

plumb_bob: k1, k2, p1, p2, k3
rational_polynomial: k1, k2, p1, p2, k3, k4, k5, k6
kannala_brandt: k1, k2, k3, k4

The plumb_bob and rational_polynomial models are based on OpenCV’s pinhole model.
The kannala_brandt model corresponds to OpenCV’s fisheye model.

K — Intrinsic matrix

A 3x3 matrix used to project 3D camera coordinates to 2D pixel coordinates:

[fx  0  cx]
[ 0 fy  cy]
[ 0  0   1]

Where fx, fy are the focal lengths, and cx, cy are the optical center.

R — Rectification matrix

Used to align the camera’s coordinate frame with a common stereo image plane, ensuring the epipolar lines are parallel in both images.

P — Projection matrix

[fx'  0  cx'  Tx]
[ 0  fy' cy'  Ty]
[ 0   0   1    0]

This matrix projects 3D points into rectified 2D image space using possibly updated intrinsics (fx', fy', cx', cy').
For monocular setups, Tx = Ty = 0.

In stereo configurations:

The first (left) camera has Tx = Ty = 0.
The second (right) camera usually has Ty = 0, and Tx = -fx' * B, where B is the baseline distance.

Given a 3D point [X Y Z]', its projection [x y] on the image is computed as:

[u v w]' = P * [X Y Z 1]'
x = u / w
y = v / w

This applies to both images in a stereo pair.

Reference implementations

Foxglove schemas are implementation-independent and can be used with any of the supported message encodings:

Encoding	Schema
ROS 1	`foxglove_msgs/CameraCalibration`
ROS 2	`foxglove_msgs/msg/CameraCalibration`
JSON	`foxglove.CameraCalibration`
Protobuf	`foxglove.CameraCalibration`
FlatBuffers	`foxglove.CameraCalibration`
OMG IDL	`foxglove::CameraCalibration`

Lichtblick Documentation