Build a camera calibration model from intrinsic coefficients

Assembles a [CalModel-class] from the focal lengths, principal point, and distortion coefficients reported by a camera-calibration toolbox. This is the low-level constructor used by [read_calibration()]; call it directly when you already have the numbers in hand.

Usage

cal_model(fx, fy, cx, cy, k1 = 0, k2 = 0, p1 = 0, p2 = 0, k3 = 0, F = NULL)

Arguments

fx, fy

Focal lengths in pixels.

cx, cy

Principal point in pixels, in the same coordinate convention as the track coordinates the model will be applied to (0-based for most tracking tools).

k1, k2, k3

Radial distortion coefficients. Default to `0` (no radial distortion).

p1, p2

Tangential distortion coefficients. Default to `0`.

F

Scale factor controlling the units of the corrected output, passed to [scaled_xy2mm()]. When `NULL` (the default) it is set to `c(fx, fy)`, which leaves the output in undistorted pixels. Supply the physical focal length in millimetres to obtain a metric result, or any length-2 vector for custom scaling.

Value

A [CalModel-class] object.

Details

radiatR stores the intrinsic matrix in MATLAB's *transposed* convention: the focal lengths sit on the diagonal (`K[1, 1]`, `K[2, 2]`) and the principal point occupies the bottom row (`K[3, 1:2]`). `cal_model()` owns this layout so callers never have to assemble `K` by hand. Axis skew is not used by the downstream correction and is therefore dropped.

The distortion coefficients follow the Brown-Conrady model shared by MATLAB (`RadialDistortion` + `TangentialDistortion`) and OpenCV (`[k1, k2, p1, p2, k3]`): radial terms `k1`, `k2`, `k3` and tangential terms `p1`, `p2`.

See also

[read_calibration()], [calibrate_positions()]

Other calibration: CalModel-class, calibrate_positions(), read_calibration()

Examples

model <- cal_model(fx = 800, fy = 800, cx = 640, cy = 360,
                   k1 = -0.21, k2 = 0.04)
cam_cal_pt(100, 100, model@K, model@k, model@F)
#> [1] -624.1998 -300.5406