Represents a pinhole camera model containing intrinsic and extrinsic parameters, used for projection, rectification, and transformation. More...

#include <CameraModel.h>

Public Member Functions
	CameraModel ()
	Default constructor.

	CameraModel (const std::string &name, const cv::Size &imageSize, const cv::Mat &K, const cv::Mat &D, const cv::Mat &R, const cv::Mat &P, const Transform &localTransform=opticalRotation())
	Constructor using full camera parameters.

	CameraModel (double fx, double fy, double cx, double cy, const Transform &localTransform=opticalRotation(), double Tx=0.0f, const cv::Size &imageSize=cv::Size(0, 0))
	Minimal constructor using intrinsic parameters. This assumes the images are already rectified.

	CameraModel (const std::string &name, double fx, double fy, double cx, double cy, const Transform &localTransform=opticalRotation(), double Tx=0.0f, const cv::Size &imageSize=cv::Size(0, 0))
	Minimal constructor with name for saving.

virtual	~CameraModel ()
	Destructor.

bool	initRectificationMap ()
	Initializes the rectification maps used to undistort and rectify images.

bool	isRectificationMapInitialized () const
	Checks if the rectification map has been initialized.

bool	isValidForProjection () const
	Checks if the model is valid for 2D->3D projection.

bool	isValidForReprojection () const
	Checks if the model is valid for 3D->2D reprojection.

bool	isValidForRectification () const
	Checks if the model has sufficient data for image rectification.

void	setName (const std::string &name)
	Sets the camera name, used to set a camera name when saving to a file.

const std::string &	name () const
	Returns the camera name.

double	fx () const
	Returns focal length in x.

double	fy () const
	Returns focal length in y.

double	cx () const
	Returns principal point x.

double	cy () const
	Returns principal point y.

double	Tx () const
	Returns the x translation (usually fx * baseline in case of stereo, otherwise would be 0).

cv::Mat	K_raw () const
	Returns the raw intrinsic matrix (before rectification).

cv::Mat	D_raw () const
	Returns the raw distortion coefficients (before rectification).

cv::Mat	K () const
	Returns the rectified camera intrinsic matrix if the projection matrix P exists, otherwise returns the raw intrinsic matrix.

cv::Mat	D () const
	Returns the rectified distortion coefficients (1x5 filled with zeros) if the projection matrix P exists, otherwise returns the raw distortion coefficients.

cv::Mat	R () const
	Returns the rectification matrix.

cv::Mat	P () const
	Returns the projection matrix.

void	setLocalTransform (const Transform &transform)
	Sets the local transform of the camera (base frame to optical frame).

const Transform &	localTransform () const
	Returns the local transform (base frame to optical frame).

void	setImageSize (const cv::Size &size)
	Sets the image size of the camera model and updates the principal point if undefined.

const cv::Size &	imageSize () const

int	imageWidth () const

int	imageHeight () const

double	fovX () const
	Returns the horizontal field of view (FoV) in radians.

double	fovY () const
	Returns the vertical field of view (FoV) in radians.

double	horizontalFOV () const
	Returns the horizontal field of view in degrees.

double	verticalFOV () const
	Returns the vertical field of view in degrees.

bool	isFisheye () const
	Checks if the distortion model is fisheye (6 coefficients: k1,k2,0,0,k3,k4).

bool	load (const std::string &filePath, bool initRectificationMaps=true)
	Loads the camera model parameters from a YAML calibration file.

bool	load (const std::string &directory, const std::string &cameraName, bool initRectificationMaps=true)
	Loads the camera model by constructing a file path from a directory and camera name.

bool	save (const std::string &directory) const
	Saves the camera model parameters to a YAML calibration file in ROS format.

std::vector< unsigned char >	serialize () const
	Serializes the camera model to a binary format.

unsigned int	deserialize (const std::vector< unsigned char > &data)
	Deserializes a camera model from a byte vector.

unsigned int	deserialize (const unsigned char *data, unsigned int dataSize)
	Deserializes a camera model from a raw byte buffer.

CameraModel	scaled (double scale) const
	Returns a new camera model with all intrinsic parameters scaled by a given factor.

CameraModel	roi (const cv::Rect &roi) const
	Returns a new camera model adjusted for a given region of interest (ROI).

cv::Mat	rectifyImage (const cv::Mat &raw, int interpolation=cv::INTER_LINEAR) const
	Rectifies a raw image using the precomputed rectification maps.

cv::Mat	rectifyDepth (const cv::Mat &raw) const
	Rectifies a raw depth image using the precomputed rectification maps.

void	project (float u, float v, float depth, float &x, float &y, float &z) const
	Projects a 2D pixel and depth value into a 3D point in the camera coordinate frame (/camera_link).

void	reproject (float x, float y, float z, float &u, float &v) const
	Reprojects a 3D point in the camera frame (/camera_link) into 2D image coordinates (floating-point).

void	reproject (float x, float y, float z, int &u, int &v) const
	Reprojects a 3D point in the camera frame (/camera_link) into 2D image coordinates (rounded to int).

bool	inFrame (int u, int v) const
	Checks if a given pixel coordinate lies within the image bounds.

Static Public Member Functions
static Transform	opticalRotation ()
	Returns the default optical rotation to convert image coordinates to robot coordinates.

Detailed Description

Represents a pinhole camera model containing intrinsic and extrinsic parameters, used for projection, rectification, and transformation.

This class encapsulates camera calibration data, including intrinsic parameters (fx, fy, cx, cy), distortion coefficients, rectification and projection matrices. It provides utility functions for image rectification, projection from 2D to 3D, and vice versa.

This class supports the 4 to 14 parameters Radial Tangential distortion model (also called Plumb Bob or Brown-Conrady model) and 4 parameters Fish Eye model (also known as Equidistant model).

See also: OpenCV's calib3d module for all supported camera models.

Definition at line 52 of file CameraModel.h.

Constructor & Destructor Documentation

◆ CameraModel() [1/2]

rtabmap::CameraModel::CameraModel	(	const std::string &	name,
		const cv::Size &	imageSize,
		const cv::Mat &	K,
		const cv::Mat &	D,
		const cv::Mat &	R,
		const cv::Mat &	P,
		const Transform &	localTransform = `opticalRotation()`
	)

Constructor using full camera parameters.

Parameters

name	Camera name or ID.
imageSize	Size of the image (width x height).
K	Intrinsic matrix (3x3).
D	Distortion coefficients, 1xN matrix where N is between 4 and 14 parameters: k1,k2,p1,p2[,k3[,k4,k5,k6[,s1,s2,s3,s4[,tx,ty]]]]). To set Fish Eye / Equidistant model, it is implicitly used if we provide 6 values like this: [k1,k2,0,0,k3,k4], where you only need to fill "k" parameters.
R	Rectification matrix (3x3).
P	Projection matrix (3x4).
localTransform	Local transform to apply to the camera frame.

◆ CameraModel() [2/2]

rtabmap::CameraModel::CameraModel	(	double	fx,
		double	fy,
		double	cx,
		double	cy,
		const Transform &	localTransform = `opticalRotation()`,
		double	Tx = `0.0f`,
		const cv::Size &	imageSize = `cv::Size(0, 0)`
	)

Minimal constructor using intrinsic parameters. This assumes the images are already rectified.

Parameters

fx	Focal length x.
fy	Focal length y.
cx	Principal point x.
cy	Principal point y.
localTransform	Local transform to apply to the camera frame.
Tx	Baseline * fx (optional). Mainly used in case of stereo pair.
imageSize	Image size (optional).

◆ ~CameraModel()

virtual rtabmap::CameraModel::~CameraModel ( )

inlinevirtual

Destructor.

Definition at line 128 of file CameraModel.h.

Member Function Documentation

◆ opticalRotation()

static Transform rtabmap::CameraModel::opticalRotation ( )

inlinestatic

Returns the default optical rotation to convert image coordinates to robot coordinates.

Image frame: x -> right, y -> down, z -> forward
Robot frame: x -> forward, y -> left, z -> up

Returns: Transform rotation matrix.

Definition at line 63 of file CameraModel.h.

◆ initRectificationMap()

bool rtabmap::CameraModel::initRectificationMap ( )

Initializes the rectification maps used to undistort and rectify images.

This function prepares the mapX_ and mapY_ lookup tables used for image rectification. It supports both standard radial-tangential distortion and fisheye/equidistant distortion models.

If the distortion model is fisheye (indicated by D_.cols == 6), it uses cv::fisheye::initUndistortRectifyMap() to create the rectification maps. This requires OpenCV ≥ 2.4.10.
Otherwise, it uses the standard cv::initUndistortRectifyMap() for plumb bob or rational polynomial models.

Precondition

The camera model must be valid for rectification:

imageSize_ must be non-zero.
D_ must be a 1-row matrix with an accepted number of columns (4, 5, 6, 8, 12, or 14).
R_ must be a 3x3 rectification matrix.
P_ must be a 3x4 projection matrix.

Returns: true if the rectification maps were successfully initialized (mapX_ and mapY_ are not empty), false otherwise.

See also: isRectificationMapInitialized(), rectifyImage(), rectifyDepth()

Warning: Requires OpenCV 2.4.10 or newer for fisheye support. If the version is older, fisheye rectification will not work.

◆ isRectificationMapInitialized()

bool rtabmap::CameraModel::isRectificationMapInitialized ( ) const

inline

Checks if the rectification map has been initialized.

Returns: True if both mapX_ and mapY_ are initialized.

Definition at line 159 of file CameraModel.h.

◆ isValidForProjection()

bool rtabmap::CameraModel::isValidForProjection ( ) const

inline

Checks if the model is valid for 2D->3D projection.

Definition at line 164 of file CameraModel.h.

◆ isValidForReprojection()

bool rtabmap::CameraModel::isValidForReprojection ( ) const

inline

Checks if the model is valid for 3D->2D reprojection.

Definition at line 168 of file CameraModel.h.

◆ isValidForRectification()

bool rtabmap::CameraModel::isValidForRectification ( ) const

inline

Checks if the model has sufficient data for image rectification.

Definition at line 172 of file CameraModel.h.

◆ setName()

void rtabmap::CameraModel::setName ( const std::string & name )

inline

Sets the camera name, used to set a camera name when saving to a file.

Definition at line 183 of file CameraModel.h.

◆ name()

const std::string & rtabmap::CameraModel::name ( ) const

inline

Returns the camera name.

Definition at line 185 of file CameraModel.h.

◆ fx()

double rtabmap::CameraModel::fx ( ) const

inline

Returns focal length in x.

Definition at line 188 of file CameraModel.h.

◆ fy()

double rtabmap::CameraModel::fy ( ) const

inline

Returns focal length in y.

Definition at line 190 of file CameraModel.h.

◆ cx()

double rtabmap::CameraModel::cx ( ) const

inline

Returns principal point x.

Definition at line 192 of file CameraModel.h.

◆ cy()

double rtabmap::CameraModel::cy ( ) const

inline

Returns principal point y.

Definition at line 194 of file CameraModel.h.

◆ Tx()

double rtabmap::CameraModel::Tx ( ) const

inline

Returns the x translation (usually fx * baseline in case of stereo, otherwise would be 0).

Definition at line 196 of file CameraModel.h.

◆ K_raw()

cv::Mat rtabmap::CameraModel::K_raw ( ) const

inline

Returns the raw intrinsic matrix (before rectification).

Definition at line 199 of file CameraModel.h.

◆ D_raw()

cv::Mat rtabmap::CameraModel::D_raw ( ) const

inline

Returns the raw distortion coefficients (before rectification).

Definition at line 201 of file CameraModel.h.

◆ K()

cv::Mat rtabmap::CameraModel::K ( ) const

inline

Returns the rectified camera intrinsic matrix if the projection matrix P exists, otherwise returns the raw intrinsic matrix.

Definition at line 203 of file CameraModel.h.

◆ D()

cv::Mat rtabmap::CameraModel::D ( ) const

inline

Returns the rectified distortion coefficients (1x5 filled with zeros) if the projection matrix P exists, otherwise returns the raw distortion coefficients.

Definition at line 205 of file CameraModel.h.

◆ R()

cv::Mat rtabmap::CameraModel::R ( ) const

inline

Returns the rectification matrix.

Definition at line 207 of file CameraModel.h.

◆ P()

cv::Mat rtabmap::CameraModel::P ( ) const

inline

Returns the projection matrix.

Definition at line 209 of file CameraModel.h.

◆ setLocalTransform()

void rtabmap::CameraModel::setLocalTransform ( const Transform & transform )

inline

Sets the local transform of the camera (base frame to optical frame).

Definition at line 212 of file CameraModel.h.

◆ localTransform()

const Transform & rtabmap::CameraModel::localTransform ( ) const

inline

Returns the local transform (base frame to optical frame).

Definition at line 214 of file CameraModel.h.

◆ setImageSize()

void rtabmap::CameraModel::setImageSize ( const cv::Size & size )

Sets the image size of the camera model and updates the principal point if undefined.

This function updates the internal image size (imageSize_) with the provided size. If the intrinsic matrices (K_ or P_) are present and the principal point coordinates (cx, cy) are zero, they are set to the image center (width/2 - 0.5, height/2 - 0.5).

This ensures the camera model remains valid and useful even when the calibration file has no principal point set or the image size is updated manually.

Parameters

size	The new image size. It must be either both dimensions zero (clearing) or both positive.

Precondition: size.width > 0 && size.height > 0 or size.width == 0 && size.height == 0

Postcondition: Updates the imageSize_, and adjusts cx and cy in K_ and P_ if they were initially zero.

Warning: If K_ or P_ are not initialized (empty()), no updates will be applied to them.

See also: imageSize(), imageWidth(), imageHeight()

◆ imageSize()

const cv::Size & rtabmap::CameraModel::imageSize ( ) const

inline

Definition at line 236 of file CameraModel.h.

◆ imageWidth()

int rtabmap::CameraModel::imageWidth ( ) const

inline

Definition at line 237 of file CameraModel.h.

◆ imageHeight()

int rtabmap::CameraModel::imageHeight ( ) const

inline

Definition at line 238 of file CameraModel.h.

◆ fovX()

double rtabmap::CameraModel::fovX ( ) const

Returns the horizontal field of view (FoV) in radians.

The FoV is computed using the pinhole camera model as:

$\text{FoV}_x = 2 \cdot \tan^{-1}\left(\frac{\text{image width}}{2 \cdot f_x}\right)$

Returns: Horizontal field of view in radians. Returns 0.0 if image width or focal length is invalid.

◆ fovY()

double rtabmap::CameraModel::fovY ( ) const

Returns the vertical field of view (FoV) in radians.

The FoV is computed using the pinhole camera model as:

$\text{FoV}_y = 2 \cdot \tan^{-1}\left(\frac{\text{image height}}{2 \cdot f_y}\right)$

Returns: Vertical field of view in radians. Returns 0.0 if image height or focal length is invalid.

◆ horizontalFOV()

double rtabmap::CameraModel::horizontalFOV ( ) const

Returns the horizontal field of view in degrees.

Converts the result of fovX() from radians to degrees.

Returns: Horizontal field of view in degrees. Returns 0.0 if the result is invalid.

◆ verticalFOV()

double rtabmap::CameraModel::verticalFOV ( ) const

Returns the vertical field of view in degrees.

Converts the result of fovY() from radians to degrees.

Returns: Vertical field of view in degrees. Returns 0.0 if the result is invalid.

◆ isFisheye()

bool rtabmap::CameraModel::isFisheye ( ) const

inline

Checks if the distortion model is fisheye (6 coefficients: k1,k2,0,0,k3,k4).

Definition at line 283 of file CameraModel.h.

◆ load() [1/2]

bool rtabmap::CameraModel::load	(	const std::string &	filePath,
		bool	initRectificationMaps = `true`
	)

Loads the camera model parameters from a YAML calibration file.

This method attempts to read camera intrinsic/extrinsic parameters and image size from a YAML file, typically in the ROS calibration format. If the distortion model is "fisheye" or "equidistant", we expect 4 coefficients, which are converted to a 6-coefficient format for internal representation.

Fields loaded (if present):

camera_name
image_width (pixels)
image_height (pixels)
camera_matrix (K, 3x3 double matrix)
distortion_coefficients (D, 1xN double matrix)
distortion_model (string: name of the model)
rectification_matrix (R, 3x3 double matrix)
projection_matrix (P, 3x4 double matrix)
local_transform (camera pose w.r.t robot frame)

On success, the internal matrices and settings of the camera model are updated. If the model is valid for rectification, the rectification map is initialized.

Parameters

filePath	Absolute or relative path to the YAML file.
initRectificationMaps	Set to false to skip building the (potentially large) rectification maps when rectification won't be used (saves time and memory).

Returns: True if the file was successfully loaded and parsed, false otherwise.

Warning: Logs warnings if any fields are missing. If file does not exist or parsing fails, returns false.

See also: initRectificationMap()

◆ load() [2/2]

bool rtabmap::CameraModel::load	(	const std::string &	directory,
		const std::string &	cameraName,
		bool	initRectificationMaps = `true`
	)

Loads the camera model by constructing a file path from a directory and camera name.

This is a convenience wrapper around load(filePath) that constructs the file path as: directory + "/" + cameraName + ".yaml".

Parameters

directory	Path to the folder containing the camera YAML file.
cameraName	Base name of the camera file (without extension).
initRectificationMaps	Set to false to skip building the (potentially large) rectification maps when rectification won't be used (saves time and memory).

Returns: True if loading from the constructed path succeeds, false otherwise.

◆ save()

bool rtabmap::CameraModel::save ( const std::string & directory ) const

Saves the camera model parameters to a YAML calibration file in ROS format.

The file will include the following fields if they are not empty:

camera_name
image_width
image_height
camera_matrix (K)
distortion_coefficients (D)
distortion_model (auto-detected based on number of distortion coefficients)
rectification_matrix (R)
projection_matrix (P)
local_transform (camera pose w.r.t robot frame)

If the distortion matrix contains 6 coefficients (used for fisheye), it is converted to a standard 4-coefficient format for ROS compatibility.

Parameters

directory Path to the folder where the YAML file will be saved.

Returns: True if saving was successful, false otherwise.

Note: If name_ is empty, "camera.yaml" is used as the default filename.

Warning: Returns false and logs an error if none of the matrices are set.

◆ serialize()

std::vector< unsigned char > rtabmap::CameraModel::serialize ( ) const

Serializes the camera model to a binary format.

The serialization includes the camera intrinsics (K_, D_), rectification matrix (R_), projection matrix (P_), image size, and the local transform. The format is compact and suitable for file storage or transmission over a network.

Data layout:

Header (11 integers):
- [0-2] RTAB-Map version (major, minor, patch)
- [3] Camera type (0 = mono, 1=stereo)
- [4-5] Image width, height
- [6-9] Element counts for K, D, R, P matrices
- [10] Size of localTransform (0 if null)
Data section (in order): raw memory blocks for K, D, R, P (double values), followed by float values for localTransform

Returns: A byte vector containing the serialized data. The format is compatible with deserialize().

Note: This is a custom binary format, not meant to be human-readable.

See also: deserialize(), StereoCameraModel

◆ deserialize() [1/2]

unsigned int rtabmap::CameraModel::deserialize ( const std::vector< unsigned char > & data )

Deserializes a camera model from a byte vector.

This is a convenience wrapper around deserialize(const unsigned char*, unsigned int) that takes a std::vector<unsigned char> instead of a raw buffer.

Parameters

data	Byte vector containing data serialized by `serialize()`.

Returns: The number of bytes successfully read and parsed. Returns 0 on failure.

See also: serialize(), deserialize(const unsigned char*, unsigned int)

◆ deserialize() [2/2]

unsigned int rtabmap::CameraModel::deserialize	(	const unsigned char *	data,
		unsigned int	dataSize
	)

Deserializes a camera model from a raw byte buffer.

Reads the camera intrinsics, distortion, rectification, projection matrices, image size, and local transform from a serialized binary format previously created with serialize().

Parameters

data	Pointer to the binary data buffer.
dataSize	Size of the data buffer in bytes.

Returns: The number of bytes successfully read. Returns 0 on error or if the format is invalid.

Warning: If the buffer format does not match the expected layout or version, an error is logged and the camera model remains in a default-initialized state.

Note: Assumes little-endian architecture and strict size/type matching. The serialized format must be created by CameraModel::serialize(). Non-mono camera types are not supported. See StereoCameraModel to serialize/deserialize stereo models.

See also: serialize()

◆ scaled()

CameraModel rtabmap::CameraModel::scaled ( double scale ) const

Returns a new camera model with all intrinsic parameters scaled by a given factor.

This method scales the camera's intrinsic matrix (K_) and projection matrix (P_), as well as the image size, by the given scale factor. The distortion and rectification matrices are left unchanged.

Only valid camera models (i.e., those for which isValidForProjection() returns true) are scaled. If the model is invalid, a warning is issued and the original model is returned unchanged.

Parameters

scale Scaling factor (> 0). For example, use 0.5 to downscale or 2.0 to upscale.

Returns: A scaled copy of the camera model with updated intrinsics and image size.

Warning: If the camera model is not valid for projection, the scale operation is ignored.

◆ roi()

CameraModel rtabmap::CameraModel::roi ( const cv::Rect & roi ) const

Returns a new camera model adjusted for a given region of interest (ROI).

This method shifts the principal point (cx, cy) in the intrinsic matrix (K_) and projection matrix (P_) by subtracting the ROI’s top-left (x, y) offset. The image size is also set to the ROI size.

Only valid camera models (i.e., those for which isValidForProjection() returns true) can be adjusted. If the model is invalid, a warning is issued and the original model is returned unchanged.

Parameters

roi	Region of interest defined as a rectangle (typically a subwindow of the full image).

Returns: A new camera model adapted to the ROI with adjusted intrinsics and image size.

Warning: If the camera model is not valid for projection, the ROI operation is ignored.

◆ rectifyImage()

cv::Mat rtabmap::CameraModel::rectifyImage	(	const cv::Mat &	raw,
		int	interpolation = `cv::INTER_LINEAR`
	)		const

Rectifies a raw image using the precomputed rectification maps.

This function applies geometric correction (rectification) to an image using the camera model's mapX_ and mapY_ rectification maps. It is typically used to correct lens distortion in images based on the calibration parameters.

Parameters

raw	Input raw image (e.g., from camera). Must be a valid `cv::Mat`.
interpolation	Interpolation method to use. Typically `cv::INTER_LINEAR` or `cv::INTER_NEAREST`.

Returns: Rectified image. If the rectification maps are not initialized, the function logs an error and returns a clone of the original image.

Precondition: mapX_ and mapY_ must be initialized using initRectificationMap().

Note: Works for color and grayscale images of any valid type.

See also: initRectificationMap(), rectifyDepth()

◆ rectifyDepth()

cv::Mat rtabmap::CameraModel::rectifyDepth ( const cv::Mat & raw ) const

Rectifies a raw depth image using the precomputed rectification maps.

This function applies geometric correction (rectification) to a 16-bit unsigned depth image. It performs a pixel-by-pixel bilinear interpolation, only if all neighboring pixels have valid (non-zero) depth values, and the variation among them is within 1% of their average.

The method is optimized to avoid introducing noise in regions of high depth variance.

Parameters

raw	Input raw depth image (`CV_16UC1`). Must contain 16-bit unsigned depth values.

Returns: Rectified depth image. If the rectification maps are not initialized or the input image is not of type CV_16UC1, the function logs an error and returns a clone of the input.

Precondition: Input image must be of type CV_16UC1. mapX_ and mapY_ must be initialized.

Note: Inspired by the Kinect2 CPU depth registration implementation from: https://github.com/code-iai/iai_kinect2

Warning: Invalid or noisy regions are skipped in interpolation to maintain depth consistency.

See also: initRectificationMap(), rectifyImage()

◆ project()

void rtabmap::CameraModel::project	(	float	u,
		float	v,
		float	depth,
		float &	x,
		float &	y,
		float &	z
	)		const

Projects a 2D pixel and depth value into a 3D point in the camera coordinate frame (/camera_link).

This function uses the camera's intrinsic parameters to compute the 3D point corresponding to the given 2D image coordinates and depth value.

Parameters

	u	Horizontal image coordinate (in pixels).
	v	Vertical image coordinate (in pixels).
	depth	Depth value at (u, v) in meters.
[out]	x	Output X coordinate in 3D space.
[out]	y	Output Y coordinate in 3D space.
[out]	z	Output Z coordinate in 3D space (equals `depth`).

Note: If depth <= 0, the output (x, y, z) will be set to NaN.

See also: reproject()

◆ reproject() [1/2]

void rtabmap::CameraModel::reproject	(	float	x,
		float	y,
		float	z,
		float &	u,
		float &	v
	)		const

Reprojects a 3D point in the camera frame (/camera_link) into 2D image coordinates (floating-point).

This function computes the image plane coordinates for a given 3D point using the camera's intrinsic parameters.

Parameters

	x	X coordinate in camera space.
	y	Y coordinate in camera space.
	z	Z coordinate in camera space (must be non-zero).
[out]	u	Output horizontal image coordinate (float).
[out]	v	Output vertical image coordinate (float).

Precondition: z != 0

See also: project(), reproject(int&, int&)

◆ reproject() [2/2]

void rtabmap::CameraModel::reproject	(	float	x,
		float	y,
		float	z,
		int &	u,
		int &	v
	)		const

Reprojects a 3D point in the camera frame (/camera_link) into 2D image coordinates (rounded to int).

This version of reproject() returns integer pixel indices, computed from the 3D position.

Parameters

	x	X coordinate in camera space.
	y	Y coordinate in camera space.
	z	Z coordinate in camera space (must be non-zero).
[out]	u	Output horizontal image coordinate (integer pixel).
[out]	v	Output vertical image coordinate (integer pixel).

Precondition: z != 0

See also: project(), reproject(float&, float&)

◆ inFrame()

bool rtabmap::CameraModel::inFrame	(	int	u,
		int	v
	)		const

Checks if a given pixel coordinate lies within the image bounds.

Parameters

u	Horizontal image coordinate (in pixels).
v	Vertical image coordinate (in pixels).

Returns: true if the pixel is within the image dimensions, false otherwise.

Note: Inclusive lower bound, exclusive upper bound: [0, width), [0, height)

The documentation for this class was generated from the following file:

corelib/include/rtabmap/core/CameraModel.h

Public Member Functions

Static Public Member Functions

Detailed Description

Constructor & Destructor Documentation

◆ CameraModel() [1/2]

◆ CameraModel() [2/2]

◆ ~CameraModel()

Member Function Documentation

◆ opticalRotation()

◆ initRectificationMap()

◆ isRectificationMapInitialized()

◆ isValidForProjection()

◆ isValidForReprojection()

◆ isValidForRectification()

◆ setName()

◆ name()

◆ fx()

◆ fy()

◆ cx()

◆ cy()

◆ Tx()

◆ K_raw()

◆ D_raw()

◆ K()

◆ D()

◆ R()

◆ P()

◆ setLocalTransform()

◆ localTransform()

◆ setImageSize()

◆ imageSize()

◆ imageWidth()

◆ imageHeight()

◆ fovX()

◆ fovY()

◆ horizontalFOV()

◆ verticalFOV()

◆ isFisheye()

◆ load() [1/2]

◆ load() [2/2]

◆ save()

◆ serialize()

◆ deserialize() [1/2]

◆ deserialize() [2/2]

◆ scaled()

◆ roi()

◆ rectifyImage()

◆ rectifyDepth()

◆ project()

◆ reproject() [1/2]

◆ reproject() [2/2]

◆ inFrame()