Package boofcv.alg.similar

Class SimilarImagesFromTracks<Track>

java.lang.Object

boofcv.alg.similar.SimilarImagesFromTracks<Track>

All Implemented Interfaces:

LookUpSimilarImages

Direct Known Subclasses:

SimilarImagesTrackThenMatch

public class SimilarImagesFromTracks<Track>
extends Object
implements LookUpSimilarImages

Processes frames from PointTracker and converts the tracking results into a LookUpSimilarImages. This is intended for scene reconstruction from video sequences.

• Results are ready as soon as processFrame(java.util.List<Track>, long) finishes
• The unique string for each image is set to the frameID
• Frames can only be matched to each other with they are withing searchRadius of each other
• No loop closure is performed

Internally arrays are used as much as possible to reduce memory overhead as higher level objects have a very large amount of overhead.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	SimilarImagesFromTracks.Frame	Observations for a frame.
static class	SimilarImagesFromTracks.Match	Describes how two frames are related to each other through common observations of the same feature
static interface	SimilarImagesFromTracks.TrackToID<Track>	Gets a unique ID from the track
static interface	SimilarImagesFromTracks.TrackToPixel<Track>	Gets the pixel coordinate of the track

Field Summary

Fields

Modifier and Type	Field	Description
final DogArray<SimilarImagesFromTracks.Match>	connections	List of all matches between frames
final Map<String,SimilarImagesFromTracks.Frame>	frameMap	Quick way to retrieve a frame based on its ID
final DogArray<SimilarImagesFromTracks.Frame>	frames	Stores observations and track ID for every frame
int	imageHeight
int	imageWidth
final ConfigLength	minimumCommonTracks	If the number of common tracks between two frames drops below this then they are considered to be disconnected.
protected DogArray<AssociatedIndex>	pairs
int	searchRadius	Maximum number of frames in forwards and backwards direction which will be searched for being related.
protected List<PointTrack>	tracks

Constructor Summary

Constructors

Constructor	Description
SimilarImagesFromTracks(SimilarImagesFromTracks.TrackToID<Track> lookupID, SimilarImagesFromTracks.TrackToPixel<Track> lookupPixel)

Method Summary

Modifier and Type	Method	Description
protected void	connectFrames(SimilarImagesFromTracks.Frame current, SimilarImagesFromTracks.Frame previous, FastAccess<AssociatedIndex> associated)	Connects the two frames and remembers the associated features between them
protected SimilarImagesFromTracks.Frame	createFrameSaveObservations(List<Track> tracks, long frameID)	Create a new frame from the tracker's current observations.
void	findSimilar(String target, BoofLambdas.Filter<String> filter, List<String> similarImages)	Given an imageID, it will find list of similar images.
List<String>	getImageIDs()	Returns the list of image ID's in the same order they were processed.
void	initialize(int width, int height)	Resets all data structures and saves the image size.
boolean	lookupAssociated(String viewB, DogArray<AssociatedIndex> pairs)	Looks up pairs of associated features from a similar image which was returned in the most recent call to LookUpSimilarImages.findSimilar(java.lang.String, boofcv.misc.BoofLambdas.Filter<java.lang.String>, java.util.List<java.lang.String>).
void	lookupPixelFeats(String target, DogArray<Point2D_F64> features)	Looks up pixel observations of features in the specified view.
void	processFrame(List<Track> tracks, long frameID)	Process the latest observations from the tracker.

Methods inherited from class java.lang.Object

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

Field Details

searchRadius

public int searchRadius

Maximum number of frames in forwards and backwards direction which will be searched for being related. If less than zero there are no hard limit.

minimumCommonTracks

public final ConfigLength minimumCommonTracks

If the number of common tracks between two frames drops below this then they are considered to be disconnected. If relative, then it's relative to the minimum number of observed tracks in each frame..

frames

public final DogArray<SimilarImagesFromTracks.Frame> frames

Stores observations and track ID for every frame

frameMap

public final Map<String,SimilarImagesFromTracks.Frame> frameMap

Quick way to retrieve a frame based on its ID

connections

public final DogArray<SimilarImagesFromTracks.Match> connections

List of all matches between frames

imageWidth

public int imageWidth

imageHeight

public int imageHeight

tracks

protected List<PointTrack> tracks

pairs

protected DogArray<AssociatedIndex> pairs

Constructor Details

SimilarImagesFromTracks

public SimilarImagesFromTracks(SimilarImagesFromTracks.TrackToID<Track> lookupID,
 SimilarImagesFromTracks.TrackToPixel<Track> lookupPixel)

Method Details

initialize

public void initialize(int width,
 int height)

Resets all data structures and saves the image size. Must be called before other functions

processFrame

public void processFrame(List<Track> tracks,
 long frameID)

Process the latest observations from the tracker. Save the results and

Parameters:

tracks - (Input) List of active tracks visible in the current frame

frameID - (Input) Identifier for this image/frame

createFrameSaveObservations

protected SimilarImagesFromTracks.Frame createFrameSaveObservations(List<Track> tracks,
 long frameID)

Create a new frame from the tracker's current observations.

connectFrames

protected void connectFrames(SimilarImagesFromTracks.Frame current,
 SimilarImagesFromTracks.Frame previous,
 FastAccess<AssociatedIndex> associated)

Connects the two frames and remembers the associated features between them

getImageIDs

public List<String> getImageIDs()

Returns the list of image ID's in the same order they were processed.

Specified by:

getImageIDs in interface LookUpSimilarImages

findSimilar

public void findSimilar(String target,
 @Nullable
 BoofLambdas.Filter<String> filter,
 List<String> similarImages)

Description copied from interface: LookUpSimilarImages

Given an imageID, it will find list of similar images. Similarity is defined from the perspective of the 'target', so viewA might think viewB is similar to it, but viewB might think viewA is not similar to it.

Specified by:

findSimilar in interface LookUpSimilarImages

Parameters:

target - (Input) ID of target image

filter - (Input) Filter results by ID. true = keep, false = reject. Null means no filter.

similarImages - (Output) Storage for IDs of similar images. Cleared upon each call

lookupPixelFeats

public void lookupPixelFeats(String target,
 DogArray<Point2D_F64> features)

Description copied from interface: LookUpSimilarImages

Looks up pixel observations of features in the specified view.

Specified by:

lookupPixelFeats in interface LookUpSimilarImages

Parameters:

target - ID of target image

features - Storage for pixel observations. Cleared upon each call

lookupAssociated

public boolean lookupAssociated(String viewB,
 DogArray<AssociatedIndex> pairs)

Description copied from interface: LookUpSimilarImages

Looks up pairs of associated features from a similar image which was returned in the most recent call to LookUpSimilarImages.findSimilar(java.lang.String, boofcv.misc.BoofLambdas.Filter<java.lang.String>, java.util.List<java.lang.String>). The src will be the target in 'findSimilar' and the dst will be the requested similar image.

Specified by:

lookupAssociated in interface LookUpSimilarImages

Parameters:

viewB - ID of a similar image to the target when calling LookUpSimilarImages.findSimilar(java.lang.String, boofcv.misc.BoofLambdas.Filter<java.lang.String>, java.util.List<java.lang.String>).

pairs - Storage for associated features. Cleared upon each call

Returns:

true if views are similar and have known associations. False if not and results should be ignored