Package edu.gsu.cs.dmlab.tracking.appearance

Class SparseHistoAppearanceModel

java.lang.Object

edu.gsu.cs.dmlab.tracking.appearance.SparseHistoAppearanceModel

All Implemented Interfaces:

ISTAppearanceModel

public class SparseHistoAppearanceModel
extends Object
implements ISTAppearanceModel

SparseHistoAppearanceModel compares the visual similarity of two tracks based on their sparse histogram similarity.

Author:

Dustin Kempton, Data Mining Lab, Georgia State University

Constructor Summary

Constructors

Constructor	Description
SparseHistoAppearanceModel(ISparseDictionaryLearner dictionaryLearner, ISparseMatrixApproximator coefExtractor, IImgPatchVectorizer patchVectorizer, ISTSparseHistoCreator histoCreator2, ISTSparseCandidateModel candidateModel2, ISTImageDBConnection imageDB2, ImageDBWaveParamPair[] params, int cacheSize)	Constructor

Method Summary

Modifier and Type	Method	Description
double	calcProbAppearance(ISTTrackingTrajectory leftTrack, ISTTrackingTrajectory rightTrack)	Produces a likelihood value for the two input tracks being the same object based upon their appearance at the join point.
void	finalize()

Methods inherited from class java.lang.Object

clone, equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

SparseHistoAppearanceModel

public SparseHistoAppearanceModel(ISparseDictionaryLearner dictionaryLearner,
                                  ISparseMatrixApproximator coefExtractor,
                                  IImgPatchVectorizer patchVectorizer,
                                  ISTSparseHistoCreator histoCreator2,
                                  ISTSparseCandidateModel candidateModel2,
                                  ISTImageDBConnection imageDB2,
                                  ImageDBWaveParamPair[] params,
                                  int cacheSize)

Constructor

Parameters:

dictionaryLearner - The class that is used to learn a sparse dictionary based upon the target area input.

coefExtractor - The class used to generate the sparse approximation of the areas of interest using the learned dictionary from the target area.

patchVectorizer - The class used to extract the column vectors of patches within the areas of interest.

histoCreator2 - The class used to create a histogram of the coefficients from the sparse approximation of the areas of interest.

candidateModel2 - The class used to determine the likelihood of a signal being generated at random from the learned dictionary.

imageDB2 - The connection to the image database to pull the image parameters for each area of interest.

params - A set of image parameter wavelength pairs that tell this object which ones to use from the database.

cacheSize - How many target dictionaries to store in the cache of this object, used so the dictionary doesn't need to be calculated every time another candidate is processed for a given target.

Method Detail

finalize

public void finalize()
              throws Throwable

Overrides:

finalize in class Object

Throws:

Throwable

calcProbAppearance

public double calcProbAppearance(ISTTrackingTrajectory leftTrack,
                                 ISTTrackingTrajectory rightTrack)

Description copied from interface: ISTAppearanceModel

Produces a likelihood value for the two input tracks being the same object based upon their appearance at the join point.

Specified by:

calcProbAppearance in interface ISTAppearanceModel

Parameters:

leftTrack - The track that is earlier in time.

rightTrack - The track that is later in time.

Returns:

The likelihood value for their similarity.