PARP Research Group Universidad de Murcia


QVKLTTracker Class Reference
[GPU-KLT+FLOW wrapper functions and classes]

Wrapper class to perform image feature tracking with the GPU-KLT library. More...

#include <QVKLTTracker>

List of all members.

Public Member Functions

 QVKLTTracker ()
 Default constructor Do nothing.
 QVKLTTracker (int width, int height, bool trackWithGain=true, int featuresWidth=64, int featuresHeight=64, int nIterations=10, int nLevels=4, int levelSkip=1, int windowWidth=7, float trackBorderMargin=20.0f, float detectBorderMargin=20.0f, float convergenceThreshold=0.1f, float SSD_Threshold=5000.0f, int minDistance=7, float minCornerness=50.0f)
 Allocation constructor.
 ~QVKLTTracker ()
 Deallocate the tracker from the GPU.
void detect (const QVImage< uChar, 1 > &image, QHash< int, QVKLTTrackerFeature > &features)
 Detect all features in image from scratch.
void redetect (const QVImage< uChar, 1 > &image, QHash< int, QVKLTTrackerFeature > &features)
 Detect new features while respecting old ones.
void track (const QVImage< uChar, 1 > &image, QHash< int, QVKLTTrackerFeature > &features)
 Track features in the hash.

Detailed Description

Wrapper class to perform image feature tracking with the GPU-KLT library.

Definition at line 77 of file qvklttracker.h.


Constructor & Destructor Documentation

QVKLTTracker::QVKLTTracker (  )  [inline]

Default constructor Do nothing.

Definition at line 84 of file qvklttracker.h.

QVKLTTracker::QVKLTTracker ( int  width,
int  height,
bool  trackWithGain = true,
int  featuresWidth = 64,
int  featuresHeight = 64,
int  nIterations = 10,
int  nLevels = 4,
int  levelSkip = 1,
int  windowWidth = 7,
float  trackBorderMargin = 20.0f,
float  detectBorderMargin = 20.0f,
float  convergenceThreshold = 0.1f,
float  SSD_Threshold = 5000.0f,
int  minDistance = 7,
float  minCornerness = 50.0f 
)

Allocation constructor.

This method initializes and configures a new tracker with parameters below.

Parameters:
width Number of cols of the images.
height Number of rows of the images.
trackWithGain Wether the tracker uses the gain in the tracking method or not.
featuresWidth featuresWidth*featuresHeight gives the maximum number of features to track.
featuresHeight featuresWidth*featuresHeight gives the maximum number of features to track.
nIterations Number of iterations.
nLevels Number of pyramid levels to skip. Reasonable values are 1 (all levels) and nLevels-1 (2 levels only)
levelSkip Number of pyramid levels to skip.
windowWidth Size of KLT window.
trackBorderMargin Size of the border image margin when tracking.
detectBorderMargin Size of the border image margin when redetecting.
convergenceThreshold Convergence threshold for KLT.
SSD_Threshold SSD threshold for KLT.
minDistance Minimum distance among tracked features.
minConnerness Minimum degree of cornerness for detecting new features.

Definition at line 27 of file qvklttracker.cpp.

QVKLTTracker::~QVKLTTracker (  ) 

Deallocate the tracker from the GPU.

Definition at line 58 of file qvklttracker.cpp.


Member Function Documentation

void QVKLTTracker::detect ( const QVImage< uChar, 1 > &  image,
QHash< int, QVKLTTrackerFeature > &  features 
)

Detect all features in image from scratch.

This method builds a hash with freshly detected features.

Parameters:
image Input image.
features Pointer to created hash.

Definition at line 100 of file qvklttracker.cpp.

void QVKLTTracker::redetect ( const QVImage< uChar, 1 > &  image,
QHash< int, QVKLTTrackerFeature > &  features 
)

Detect new features while respecting old ones.

This method updates the feature hash with new features while respecting old ones.

Parameters:
image Input image.
features Pointer to hash that will be updated.

Definition at line 110 of file qvklttracker.cpp.

void QVKLTTracker::track ( const QVImage< uChar, 1 > &  image,
QHash< int, QVKLTTrackerFeature > &  features 
)

Track features in the hash.

This method updates the hash with the new positions of the features, deleting the lost ones.

Parameters:
image Input image
features Pointer to hash that will be updated.

Definition at line 120 of file qvklttracker.cpp.


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



QVision framework. PARP research group. Copyright © 2007, 2008, 2009, 2010, 2011.