boofcv.alg.tracker.tld

Class TldParameters

java.lang.Object

boofcv.alg.tracker.tld.TldParameters

public class TldParameters
extends java.lang.Object

Configuration file for TLD tracker.

See Also:

TldTracker

Field Summary

Fields

Modifier and Type	Field and Description
double	confidenceAccept The minimum value for a region's confidence that will be accepted.
double	confidenceThresholdLower Lower acceptance threshold for confidence.
double	confidenceThresholdStrong A track must have a confidence above this value to be considered highly confident, allowing learning to be activated again.
double	confidenceThresholdUpper Upper acceptance threshold for confidence.
int	detectMinimumSide The minimum number of pixels along a side in a detection rectangle which will be considered.
int	fernSize Number of sample points pairs.
int	maximumCascadeConsider Maximum number of NCC templates it will examine inside the detection cascade.
double	maximumErrorFB The maximum allowed forwards-backwards error (pixels) for a track.
int	motionIterations Number of iterations in LSMeD to estimate the region's motion.
int	numFerns Number of fern descriptors.
int	numNegativeFerns Number of ferns it will use to train the negative classifier
double	overlapLower If two regions have an overlap less than this value they are considered to be disconnected
double	overlapUpper If two regions have an overlap more than or equal to this value they are considered to be strongly connected
long	randomSeed Random number seed.
double	regionConnect If two regions have an overlap fraction more than or equal to this value then they will be considered connected.
int	scaleSpread Determines the number of scales it will search in increments of powers of 1.2.
double	thresholdSimilarArea How similar the area needs to be for it to be considered a continuation of the previous track and will update the description
KltConfig	trackerConfig Basic parameters for tracker.
int	trackerFeatureRadius Radius of KLT tracks.
int	trackerGridWidth Tracks are spawned in an evenly spaced grid inside the previous region.

Constructor Summary

Constructors

Constructor and Description
TldParameters()

Method Summary

Methods inherited from class java.lang.Object

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

Field Detail

maximumCascadeConsider

public int maximumCascadeConsider

Maximum number of NCC templates it will examine inside the detection cascade. Used to limit the amount of processing used during detection. To disable set to Integer.MAX_VALUE

numNegativeFerns

public int numNegativeFerns

Number of ferns it will use to train the negative classifier

maximumErrorFB

public double maximumErrorFB

The maximum allowed forwards-backwards error (pixels) for a track. Suggested value is 10.

trackerGridWidth

public int trackerGridWidth

Tracks are spawned in an evenly spaced grid inside the previous region. This value specifies the number of rows and columns in the grid.

trackerFeatureRadius

public int trackerFeatureRadius

Radius of KLT tracks. A radius of 5 is recommended.

detectMinimumSide

public int detectMinimumSide

The minimum number of pixels along a side in a detection rectangle which will be considered.

motionIterations

public int motionIterations

Number of iterations in LSMeD to estimate the region's motion.

regionConnect

public double regionConnect

If two regions have an overlap fraction more than or equal to this value then they will be considered connected. Used inside of non-maximum suppression. A value of 0.5 is suggested.

overlapUpper

public double overlapUpper

If two regions have an overlap more than or equal to this value they are considered to be strongly connected

overlapLower

public double overlapLower

If two regions have an overlap less than this value they are considered to be disconnected

thresholdSimilarArea

public double thresholdSimilarArea

How similar the area needs to be for it to be considered a continuation of the previous track and will update the description

confidenceThresholdStrong

public double confidenceThresholdStrong

A track must have a confidence above this value to be considered highly confident, allowing learning to be activated again.

confidenceThresholdUpper

public double confidenceThresholdUpper

Upper acceptance threshold for confidence. Suggested value is 0.65

confidenceThresholdLower

public double confidenceThresholdLower

Lower acceptance threshold for confidence. Used during hypothesis fusion. Suggested value is 0.5

randomSeed

public long randomSeed

Random number seed. Used to create ferns and perform robust model fitting.

numFerns

public int numFerns

Number of fern descriptors. A value of 10 is recommended.

fernSize

public int fernSize

Number of sample points pairs. 0 < N ≤ 32. A value of 10 is recommended.

confidenceAccept

public double confidenceAccept

The minimum value for a region's confidence that will be accepted. When the tracking hypothesis is accepted its value can dip very low.

scaleSpread

public int scaleSpread

Determines the number of scales it will search in increments of powers of 1.2. All scales from 1.2^(-scaleSpread) to 1.2^scaleSpread are checked.

trackerConfig

public KltConfig trackerConfig

Basic parameters for tracker. KltConfig.createDefault() with maxIterations = 50 is suggested.

Constructor Detail

TldParameters

public TldParameters()