opentl::modalities::IntensityEdges Class Reference

Class for contour point feature modality computation. ATTENTION: This class is currently not prepared for multi-threaded match...Level() calls related to a single sensor index! More...

Inherits opentl::modalities::Modality.

List of all members.

Public Types


Public Member Functions

virtual modalities::Modality * clone () const
 Clone this class and all potential childs RECURSIVELY!! (deep copy).
opentl::core::cvdata::VisContourSamplePts * getContourData (int s=0)
virtual void init ()
 Allocate data structures required by this class.
 IntensityEdges (const IntensityEdges &c)
 Deep Copy Constructor.
 IntensityEdges (const ObjModelPtrVector &objModels, opentl::modelprojection::Warp *warp, opentl::modelprojection::GLScene *scene, int camIdx, modelprojection::ContourSampler *contourSampler=NULL, opentl::modelprojection::GLRenderer *renderer=NULL)
 Constructor ATTENTION: The sensor vector element size/sequence must be identical for the warps of all object models !!!
void learnBackgroundEdges (const opentl::core::cvdata::Image &imageGray, bool bFirstFrame, double bgDistThreshold, double cannyT1, double cannyT2, double bgRunningAvgCoeff)
 Learn a model of background edges (position and orientation) for subtraction, during pre-processing.
virtual int matchFeatLevel (const TargetPtrVector &targets, T_MEAS_FEATPtrVector &outputMeas, std::size_t partitionIdx)
 Matching on feature level
  • match projected contour points with detected image edges.

virtual int matchPixLevel (const TargetPtrVector &targets, T_MEAS_PIXPtrVector &outputMeas, std::size_t partitionIdx)
 Matching on pixel level
  • chamfer matching of projected silhouette on distance transformed edge image.

virtual int preProcess (const opentl::core::cvdata::Image &image, const std::vector< std::vector< int > > &preProcessROIs)
 Compute the edge map of I (e.g. with Canny), and store it into mImageFeatDetected.
void readSilhouettesFromFile (std::string &silhouetteFile, TargetPtrVector &targets, std::vector< std::vector< int > > &parentIdxVector)
 Read silhouette data from file.
virtual int sampleModelFeatures (const TargetPtrVector &targets)
 Do GPU-based visible edge sample point computation.
void saveSilhouettesToFile (std::string &silhouetteFile, const TargetPtrVector &targets, const std::vector< std::vector< int > > &parentIdxVector)
 Save silhouette data to file.
virtual ~IntensityEdges ()
 Destructor.

Public Attributes

opentl::core::cvdata::Image * m_bgCanny
std::vector< std::vector< int > > mAngle_ratios
 Ratios (vertical/horizontal) for each angle of the oriented DT.
opentl::core::cvdata::Image * mImageDebugOutput
 Debug Image output.
std::vector< double > mNormalsx
 Normal directions for each angle.
std::vector< double > mNormalsy
std::vector< std::vector
< std::vector< std::vector
< int > > > > 		mSilhNorm
opentl::core::cvdata::Image * mSilhOutputImage
 Image for restoring the silhouette.
opentl::core::cvdata::Image * mSilhOutputImageGray
 Image for restoring the silhouette.
std::vector< std::vector
< std::vector< std::vector
< int > > > > 		mSilhPos

Static Public Attributes

static std::vector
< boost::shared_ptr
< opentl::core::cvdata::Image > > 		mPreProcessOutput_Canny
 Canny edge map.
static std::vector
< boost::shared_ptr
< opentl::core::cvdata::Image > > 		mPreProcessOutput_D1
 Sobel X gradient.
static std::vector
< boost::shared_ptr
< opentl::core::cvdata::Image > > 		mPreProcessOutput_D2
 Sobel Y gradient.
static std::vector
< std::vector
< boost::shared_ptr
< opentl::core::cvdata::Image > > > 		mPreProcessOutput_orientedDT
 Oriented distance transform (for each angle).
static std::vector
< boost::shared_ptr
< opentl::core::cvdata::Image > > 		mSilhouetteBinaryImages
 Binary silhouette Image.
static std::vector
< boost::shared_ptr
< opentl::core::cvdata::Image > > 		mSilhouetteImages
 Silhouette Image.

Detailed Description

Class for contour point feature modality computation. ATTENTION: This class is currently not prepared for multi-threaded match...Level() calls related to a single sensor index!

Author:
Erwin Roth <eroth@mytum.de>

Member Enumeration Documentation

enum opentl::modalities::IntensityEdges::FeatDetectFilterType

Possible filter types for image feature detection.

Enumerator:
CANNY 
SOBEL 
CANNY_AND_SOBEL 
DISTANCE_TRANSFORM 

enum opentl::modalities::IntensityEdges::OfflineParameters

Enumerator:
generate_contourSampler  (bool) If true, create the contour sampler internally; else, use a pre-defined one (contourSampler)
construct_loadShaderCodeFromFile  (bool) Load Shader code from File
construct_sampleSilhouetteEdgesOnly  (bool) Sample silhoutte edged only
construct_MaxSamplePointsPerTarget  (int) maximum sample points per target
preProcess_featureFilterType  (FeatDetectFilterType) preprocess filter type
matchP_glRenderer  Renderer needed if pixel level matching with DISTANCE_TRANSFORM is used.
subtract_learnedBackground  (bool) If true, subtract the learned background
OFFLINE_COUNT 

Reimplemented from opentl::modalities::Modality.

enum opentl::modalities::IntensityEdges::OnlineParameters

Enumerator:
sample_discardBackgroundEdgeWithinPixelRange  (unsigned int) range to discard background edge pixel
sample_computeScreenPointJacobians  (bool) compute screen point jacobians
preProcess_cannyThreshold1  (double) canny threshold1
preProcess_cannyThreshold2  (double) canny threshold2
preProcess_bgOrientThresh  (double) background orient threshold
preProcess_numofAngles  (unsigned int) number of angles for computing orientedDT
matchF_alpha  (double) Missing detection rate
matchF_edgeSearchlengthAlongNormalinPixel  (unsigned int) edge search length along the normal in pixel
matchF_useRansac  (bool) use RANSAC
matchF_dotprodThreshold  (double) dot product threshold
matchF_angleThreshold  (double) angle threshold
matchF_enableDebugOutput  (bool) enable debug output
matchFeatLevel_robustFlag  (bool) robust flag
matchFeatLevel_useFixedCov  (bool) use fixed covariance
matchF_nearestNeighbor  (bool) nearest neighbour
matchF_maxMultiHypoAssoc  (unsigned int) Max. number of associations (for multi-hypotheses matching)
ONLINE_COUNT 

Reimplemented from opentl::modalities::Modality.

Constructor & Destructor Documentation

opentl::modalities::IntensityEdges::IntensityEdges ( const ObjModelPtrVector &  objModels,
opentl::modelprojection::Warp *  warp,
opentl::modelprojection::GLScene *  scene,
int  camIdx,
modelprojection::ContourSampler *  contourSampler = NULL,
opentl::modelprojection::GLRenderer *  renderer = NULL 
)

Constructor ATTENTION: The sensor vector element size/sequence must be identical for the warps of all object models !!!

opentl::modalities::IntensityEdges::IntensityEdges ( const IntensityEdges &  c  ) 

Deep Copy Constructor.

virtual opentl::modalities::IntensityEdges::~IntensityEdges (  )  [virtual]

Destructor.

Member Function Documentation

virtual modalities::Modality* opentl::modalities::IntensityEdges::clone (  )  const [virtual]

Clone this class and all potential childs RECURSIVELY!! (deep copy).

Implements opentl::modalities::Modality.

opentl::core::cvdata::VisContourSamplePts* opentl::modalities::IntensityEdges::getContourData ( int  s = 0  )  [inline]

virtual void opentl::modalities::IntensityEdges::init (  )  [virtual]

Allocate data structures required by this class.

Reimplemented from opentl::modalities::Modality.

void opentl::modalities::IntensityEdges::learnBackgroundEdges ( const opentl::core::cvdata::Image &  imageGray,
bool  bFirstFrame,
double  bgDistThreshold,
double  cannyT1,
double  cannyT2,
double  bgRunningAvgCoeff 
)

Learn a model of background edges (position and orientation) for subtraction, during pre-processing.

virtual int opentl::modalities::IntensityEdges::matchFeatLevel ( const TargetPtrVector &  targets,
T_MEAS_FEATPtrVector &  outputMeas,
std::size_t  partitionIdx 
) [virtual]

Matching on feature level

  • match projected contour points with detected image edges.

Parameters:
targets Vector of targets
outputMeas Feature-space measurements and residuals (one per target and internal state)

Reimplemented from opentl::modalities::Modality.

virtual int opentl::modalities::IntensityEdges::matchPixLevel ( const TargetPtrVector &  targets,
T_MEAS_PIXPtrVector &  outputMeas,
std::size_t  partitionIdx 
) [virtual]

Matching on pixel level

  • chamfer matching of projected silhouette on distance transformed edge image.

Parameters:
targets Vector of targets
outputMeas Pixel-space measurement (a unique residual value - chamfer distance)

Reimplemented from opentl::modalities::Modality.

virtual int opentl::modalities::IntensityEdges::preProcess ( const opentl::core::cvdata::Image &  image,
const std::vector< std::vector< int > > &  preProcessROIs 
) [virtual]

Compute the edge map of I (e.g. with Canny), and store it into mImageFeatDetected.

Parameters:
image Input sensor data (e.g. camera image in RGB)
preProcessROIs Regions of interest (x0,y0,width,height), per target

Reimplemented from opentl::modalities::Modality.

void opentl::modalities::IntensityEdges::readSilhouettesFromFile ( std::string &  silhouetteFile,
TargetPtrVector &  targets,
std::vector< std::vector< int > > &  parentIdxVector 
)

Read silhouette data from file.

Parameters:
silhouetteFile name of the input file
targets Output target vector, containing the grid states (one target per level)
parentIdxVector Output indices of parent cells in the upper level grid (at l=0 it is empty)

virtual int opentl::modalities::IntensityEdges::sampleModelFeatures ( const TargetPtrVector &  targets  )  [virtual]

Do GPU-based visible edge sample point computation.

Parameters:
targets Vector of targets

Reimplemented from opentl::modalities::Modality.

void opentl::modalities::IntensityEdges::saveSilhouettesToFile ( std::string &  silhouetteFile,
const TargetPtrVector &  targets,
const std::vector< std::vector< int > > &  parentIdxVector 
)

Save silhouette data to file.

Parameters:
silhouetteFile name of the output file
targets Input target vector, containing the grid states (one target per level)
parentIdxVector Input indices of parent cells in the upper level grid (at l=0 it is empty)

Member Data Documentation

opentl::core::cvdata::Image* opentl::modalities::IntensityEdges::m_bgCanny

std::vector<std::vector<int> > opentl::modalities::IntensityEdges::mAngle_ratios

Ratios (vertical/horizontal) for each angle of the oriented DT.

opentl::core::cvdata::Image* opentl::modalities::IntensityEdges::mImageDebugOutput

Debug Image output.

std::vector<double> opentl::modalities::IntensityEdges::mNormalsx

Normal directions for each angle.

std::vector<double> opentl::modalities::IntensityEdges::mNormalsy

std::vector<boost::shared_ptr<opentl::core::cvdata::Image> > opentl::modalities::IntensityEdges::mPreProcessOutput_Canny [static]

Canny edge map.

std::vector<boost::shared_ptr<opentl::core::cvdata::Image> > opentl::modalities::IntensityEdges::mPreProcessOutput_D1 [static]

Sobel X gradient.

std::vector<boost::shared_ptr<opentl::core::cvdata::Image> > opentl::modalities::IntensityEdges::mPreProcessOutput_D2 [static]

Sobel Y gradient.

std::vector<std::vector<boost::shared_ptr<opentl::core::cvdata::Image> > > opentl::modalities::IntensityEdges::mPreProcessOutput_orientedDT [static]

Oriented distance transform (for each angle).

std::vector<std::vector<std::vector<std::vector<int> > > > opentl::modalities::IntensityEdges::mSilhNorm

std::vector<boost::shared_ptr<opentl::core::cvdata::Image> > opentl::modalities::IntensityEdges::mSilhouetteBinaryImages [static]

Binary silhouette Image.

std::vector<boost::shared_ptr<opentl::core::cvdata::Image> > opentl::modalities::IntensityEdges::mSilhouetteImages [static]

Silhouette Image.

opentl::core::cvdata::Image* opentl::modalities::IntensityEdges::mSilhOutputImage

Image for restoring the silhouette.

opentl::core::cvdata::Image* opentl::modalities::IntensityEdges::mSilhOutputImageGray

Image for restoring the silhouette.

std::vector<std::vector<std::vector<std::vector<int> > > > opentl::modalities::IntensityEdges::mSilhPos

Silhouette pixels 1st index: target MODEL (NOTE: two targets with the same model share the same silhouettes) 2nd index: level of the grid 3rd index: index of state cell (at this level) 4th index: mSilhPos: pixel coordinate (as a single array index) / mSilhNorm: normal to the silhouette (index in mAngle_ratios)

Generated on Thu Jun 10 21:08:10 2010 for OpenTL by  doxygen 1.5.8