Main visual odometry class. More...

Public Member Functions
bool	getChangeReferenceFrames () const

int	getFastThreshold () const

const Eigen::Matrix3d &	getInitialHomography () const

const Eigen::Isometry3d &	getMotionEstimate () const

const Eigen::MatrixXd &	getMotionEstimateCov () const

MotionEstimateStatusCode	getMotionEstimateStatus () const

const MotionEstimator *	getMotionEstimator () const

const VisualOdometryOptions &	getOptions () const

const Eigen::Isometry3d &	getPose ()

const OdometryFrame *	getReferenceFrame () const

const OdometryFrame *	getTargetFrame () const

void	processFrame (const uint8_t gray, DepthSource depth_source)

void	sanityCheck () const

	VisualOdometry (const Rectification *rectification, const VisualOdometryOptions &options)

Static Public Member Functions
static VisualOdometryOptions	getDefaultOptions ()

Detailed Description

Main visual odometry class.

#include <fovis/fovis.hpp>

This is the primary fovis class for estimating visual odometry. To use it, you'll need three things:

a source of grayscale input images.
a DepthSource that can estimate the distance to as many pixels in the input images as possible.
a Rectification object for converting the source image coordinates to a rectified pinhole projection coordinate system. This is typically used to correct radial lens distortion.

A typical use case for the VisualOdometry class is to repeatedly call processFrame() as new image data is available, which estimates the camera motion and makes the resulting estimation data available via accessor methods.

Options to control the behavior of the visual odometry algorithm can be passed in to the constructor using a VisualOdometryOptions object.

Constructor & Destructor Documentation

VisualOdometry	(	const Rectification *	rectification,
		const VisualOdometryOptions &	options
	)

Constructs a new visual odometry estimator.

Parameters

rectification	specifies the input image dimensions, as well as the mapping from input image coordinates to rectified image coordinates.
options	controls the behavior of the estimation algorithms. This is specified as a key/value dictionary.

Member Function Documentation

void processFrame	(	const uint8_t *	gray,
		DepthSource *	depth_source
	)

process an input image and estimate the 3D camera motion between gray and the frame previously passed to this method. The estimated motion for the very first frame will always be the identity transform.

Parameters

gray	a new input image. The image dimensions must match those passed in to the constructor, and the image data must be stored in row-major order, with no pad bytes between rows. An internal copy of the image is made, and the input data is no longer needed once this method returns.
depth_source	a source of depth information that can either provide a depth estimate at each pixel of the input image, or report that no depth estimate is available.

const Eigen::Isometry3d& getPose ( )

inline

Retrieves the integrated pose estimate. On initialization, the camera is positioned at the origin, with +Z pointing along the camera look vector, +X to the right, and +Y down.

const OdometryFrame* getReferenceFrame ( ) const

inline

Retrieve the current reference frame used for motion estimation. The reference frame will not change as long as new input frames are easily matched to it.

const OdometryFrame* getTargetFrame ( ) const

inline

Retrieve the current target frame used for motion estimation.

bool getChangeReferenceFrames ( ) const

inline

If this returns true, then the current target frame will become the reference frame on the next call to processFrame().

MotionEstimateStatusCode getMotionEstimateStatus ( ) const

inline

Returns: whether motion estimation succeeded on the most recent call to processFrame(), or provides a rough failure reason.

const Eigen::Isometry3d& getMotionEstimate ( ) const

inline

Returns: the estimated camera motion from the previous frame to the current frame.

const Eigen::MatrixXd& getMotionEstimateCov ( ) const

inline

Returns: the covariance matrix resulting from the final nonlinear least-squares motion estimation step.

const MotionEstimator* getMotionEstimator ( ) const

inline

Returns: the MotionEstimator object used internally.

int getFastThreshold ( ) const

inline

Returns: the threshold used by the FAST feature detector.

const Eigen::Matrix3d& getInitialHomography ( ) const

inline

Returns: the 2D homography computed during initial rotation estimation.

const VisualOdometryOptions& getOptions ( ) const

inline

Returns: the options passed in to the constructor.

static VisualOdometryOptions getDefaultOptions ( )

static

Returns: a reasonable set of default options that can be passed in to the constructor if you don't know or care about the options.

void sanityCheck ( ) const

Performs some internal sanity checks and aborts the program on failure. This is for debugging only.

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

visual_odometry.hpp

Public Member Functions

Static Public Member Functions

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation