Class TCamera

Defined in File TCamera.h

Inheritance Relationships

Base Type

public mrpt::serialization::CSerializable

Class Documentation

class TCamera : public mrpt::serialization::CSerializable

Intrinsic parameters for a pinhole or fisheye camera model, along with the associated lens distortion model.

The camera projection follows the standard pinhole model:

3D point in camera frame: P_cam = (X, Y, Z)
Normalized image coordinates: x = X/Z, y = Y/Z
Apply distortion to get: (x_d, y_d)
Project to pixel coordinates: u = fx*x_d + cx, v = fy*y_d + cy

The distortion model is specified by the distortion field and can be:

DistortionModel::none: No distortion (already rectified images)
DistortionModel::plumb_bob: Standard radial-tangential distortion (OpenCV model)
DistortionModel::kannala_brandt: Fish-eye camera model

Parameters calibrated for one camera resolution can be scaled to other resolutions using TCamera::scaleToResolution(), which preserves the aspect ratio.

See also

mrpt::img::camera_geometry for projection and undistortion functions

Camera parameters

uint32_t ncols = 640: Image resolution in pixels

uint32_t nrows = 480

mrpt::math::CMatrixDouble33 intrinsicParams

Camera intrinsic matrix (calibration matrix) in the standard form:

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

where:

fx, fy: focal length in pixels (horizontal and vertical)
cx, cy: principal point coordinates in pixels

See: https://www.mrpt.org/Camera_Parameters

DistortionModel distortion = DistortionModel::none : The lens distortion model type. Determines how to interpret the dist coefficients.

std::array<double, 8> dist{{.0, .0, .0, .0, .0, .0, .0, .0}}

Distortion coefficients. Interpretation depends on distortion:

DistortionModel::none: Coefficients ignored (rectified image)
DistortionModel::plumb_bob: Standard radial-tangential model (OpenCV compatible)
- Format: [k1, k2, p1, p2, k3, k4, k5, k6]
- k1, k2, k3, k4, k5, k6: Radial distortion coefficients
- p1, p2: Tangential distortion coefficients
- Distortion equations:
  - r² = x² + y²
  - x_d = x(1 + k1·r² + k2·r⁴ + k3·r⁶)/(1 + k4·r² + k5·r⁴ + k6·r⁶) + 2p1·xy + p2(r² + 2x²)
  - y_d = y(1 + k1·r² + k2·r⁴ + k3·r⁶)/(1 + k4·r² + k5·r⁴ + k6·r⁶) + p1(r² + 2y²) + 2p2·xy
DistortionModel::kannala_brandt: Fish-eye model
- Format: [k1, k2, *, *, k3, k4, *, *] (indices 2,3,6,7 unused)
- Distortion equations:
  - θ = atan(r), where r = sqrt(x² + y²)
  - θ_d = θ(1 + k1·θ² + k2·θ⁴ + k3·θ⁶ + k4·θ⁸)
  - x_d = (θ_d/r)·x, y_d = (θ_d/r)·y

Recommended: Use getter/setter methods k1(), k2(), p1(), p2(), etc. to avoid indexing errors.

Default: all zeros (no distortion)

double focalLengthMeters = .0: Physical focal length in meters (optional). Can be used with intrinsicParams to compute physical pixel size.

std::string cameraName = "camera1": Optional descriptive camera name (e.g., “left_camera”, “front_cam”).

Intrinsic parameter accessors

Convenience methods for getting/setting individual intrinsic parameters

inline double cx() const

inline double cy() const

inline double fx() const

inline double fy() const

inline void cx(double val)

inline void cy(double val)

inline void fx(double val)

inline void fy(double val)

Distortion coefficient accessors

Convenience methods for getting/setting individual distortion coefficients. Interpretation depends on the distortion model.

inline double k1() const

inline double k2() const

inline double p1() const

inline double p2() const

inline double k3() const

inline double k4() const

inline double k5() const

inline double k6() const

inline void k1(double val)

inline void k2(double val)

inline void p1(double val)

inline void p2(double val)

inline void k3(double val)

inline void k4(double val)

inline void k5(double val)

inline void k6(double val)

Public Functions

TCamera(): Default constructor: all intrinsic parameters set to zero.

mrpt::containers::yaml asYAML() const: Export camera parameters as YAML in OpenCV calibration format. Refer to this example.

void scaleToResolution(unsigned int new_ncols, unsigned int new_nrows)

Scale all parameters to a new image resolution. Throws an exception if the aspect ratio changes, as this would invalidate the calibration.

Parameters:

new_ncols – New image width in pixels
new_nrows – New image height in pixels

void saveToConfigFile(const std::string &section, mrpt::config::CConfigFileBase &cfg) const

Save camera parameters to a configuration file section in the format:

[SECTION]
resolution = [NCOLS NROWS]
cx         = CX
cy         = CY
fx         = FX
fy         = FY
dist       = [K1 K2 P1 P2 K3 K4 K5 K6]
camera_name = camera1
distortion = {none|plumb_bob|kannala_brandt}
focal_length = FOCAL_LENGTH  [optional]

void loadFromConfigFile(const std::string &section, const mrpt::config::CConfigFileBase &cfg)

Load camera parameters from a configuration file section. See saveToConfigFile() for the expected format.

Throws:: std::exception – on missing required fields

inline void loadFromConfigFile(const mrpt::config::CConfigFileBase &cfg, const std::string &section): Overload with swapped parameter order (consistent with other MRPT APIs)

std::string dumpAsText() const

Return all parameters as a formatted text block in INI file format.