BackgroundSubtractorPAWCS Class Reference

#include <BackgroundSubtractorPAWCS.h>

Inheritance diagram for BackgroundSubtractorPAWCS:

Classes
struct	ColorLBSPFeature
struct	GlobalWord
struct	GlobalWordBase
struct	LocalWord
struct	LocalWordBase
struct	PxInfo_PAWCS

Public Member Functions
	BackgroundSubtractorPAWCS (float fRelLBSPThreshold=BGSPAWCS_DEFAULT_LBSP_REL_SIMILARITY_THRESHOLD, size_t nDescDistThresholdOffset=BGSPAWCS_DEFAULT_DESC_DIST_THRESHOLD_OFFSET, size_t nMinColorDistThreshold=BGSPAWCS_DEFAULT_MIN_COLOR_DIST_THRESHOLD, size_t nMaxNbWords=BGSPAWCS_DEFAULT_MAX_NB_WORDS, size_t nSamplesForMovingAvgs=BGSPAWCS_DEFAULT_N_SAMPLES_FOR_MV_AVGS)
	full constructor
virtual	~BackgroundSubtractorPAWCS ()
	default destructor
virtual void	initialize (const cv::Mat &oInitImg, const cv::Mat &oROI)
	(re)initiaization method; needs to be called before starting background subtraction
virtual void	refreshModel (size_t nBaseOccCount, float fOccDecrFrac, bool bForceFGUpdate=false)
	refreshes all local (+ global) dictionaries based on the last analyzed frame
virtual void	apply (cv::InputArray image, cv::OutputArray fgmask, double learningRateOverride=0)
	primary model update function; the learning param is used to override the internal learning speed (ignored when <= 0)
virtual void	apply (cv::InputArray image, cv::OutputArray fgmask, cv::OutputArray _featureMap, cv::OutputArray _reliabilityMap, cv::OutputArray _foregroundRaw, double learningRateOverride=0)
virtual void	getBackgroundImage (cv::OutputArray backgroundImage) const
	returns a copy of the latest reconstructed background image
virtual void	getBackgroundDescriptorsImage (cv::OutputArray backgroundDescImage) const
	returns a copy of the latest reconstructed background descriptors image
Public Member Functions inherited from BackgroundSubtractorLBSP
	BackgroundSubtractorLBSP (float fRelLBSPThreshold, size_t nLBSPThresholdOffset=0)
	full constructor
virtual	~BackgroundSubtractorLBSP ()
	default destructor
virtual void	initialize (const cv::Mat &oInitImg)
	(re)initiaization method; needs to be called before starting background subtraction
virtual cv::Mat	getROICopy () const
	returns a copy of the ROI used for descriptor extraction
virtual void	setROI (cv::Mat &oROI)
	sets the ROI to be used for descriptor extraction (note: this function will reinit the model and return the usable ROI)
void	setAutomaticModelReset (bool)
	turns automatic model reset on or off

Protected Types
typedef LocalWord< ColorLBSPFeature< 1 > >	LocalWord_1ch
typedef LocalWord< ColorLBSPFeature< 3 > >	LocalWord_3ch
typedef GlobalWord< ColorLBSPFeature< 1 > >	GlobalWord_1ch
typedef GlobalWord< ColorLBSPFeature< 3 > >	GlobalWord_3ch

Protected Member Functions
void	CleanupDictionaries ()
	internal cleanup function for the dictionary structures

Static Protected Member Functions
static float	GetLocalWordWeight (const LocalWordBase *w, size_t nCurrFrame, size_t nOffset)
	internal weight lookup function for local words
static float	GetGlobalWordWeight (const GlobalWordBase *w)
	internal weight lookup function for global words

Protected Attributes
const size_t	m_nMinColorDistThreshold
	absolute minimal color distance threshold ('R' or 'radius' in the original ViBe paper, used as the default/initial 'R(x)' value here)
const size_t	m_nDescDistThresholdOffset
	absolute descriptor distance threshold offset
size_t	m_nMaxLocalWords
	max/curr number of local words used to build background submodels (for a single pixel, similar to 'N' in ViBe/PBAS, may vary based on img/channel size)
size_t	m_nCurrLocalWords
size_t	m_nMaxGlobalWords
	max/curr number of global words used to build the global background model (may vary based on img/channel size)
size_t	m_nCurrGlobalWords
const size_t	m_nSamplesForMovingAvgs
	number of samples to use to compute the learning rate of moving averages
float	m_fLastNonFlatRegionRatio
	last calculated non-flat region ratio
int	m_nMedianBlurKernelSize
	current kernel size for median blur post-proc filtering
cv::Size	m_oDownSampledFrameSize_MotionAnalysis
	specifies the downsampled frame size used for cam motion analysis & gword lookup maps
cv::Size	m_oDownSampledFrameSize_GlobalWordLookup
cv::Mat	m_oDownSampledROI_MotionAnalysis
	downsampled version of the ROI used for cam motion analysis
size_t	m_nDownSampledROIPxCount
	total pixel count for the downsampled ROIs
size_t	m_nLocalWordWeightOffset
	current local word weight offset
LocalWordBase **	m_apLocalWordDict
	word lists & dictionaries
LocalWord_1ch *	m_aLocalWordList_1ch
LocalWord_1ch *	m_pLocalWordListIter_1ch
LocalWord_3ch *	m_aLocalWordList_3ch
LocalWord_3ch *	m_pLocalWordListIter_3ch
GlobalWordBase **	m_apGlobalWordDict
GlobalWord_1ch *	m_aGlobalWordList_1ch
GlobalWord_1ch *	m_pGlobalWordListIter_1ch
GlobalWord_3ch *	m_aGlobalWordList_3ch
GlobalWord_3ch *	m_pGlobalWordListIter_3ch
PxInfo_PAWCS *	m_aPxInfoLUT_PAWCS
cv::Mat	m_oIllumUpdtRegionMask
	a lookup map used to keep track of regions where illumination recently changed
cv::Mat	m_oUpdateRateFrame
	per-pixel update rates ('T(x)' in PBAS, which contains pixel-level 'sigmas', as referred to in ViBe)
cv::Mat	m_oDistThresholdFrame
	per-pixel distance thresholds (equivalent to 'R(x)' in PBAS, but used as a relative value to determine both intensity and descriptor variation thresholds)
cv::Mat	m_oDistThresholdVariationFrame
	per-pixel distance threshold variation modulators ('v(x)', relative value used to modulate 'R(x)' and 'T(x)' variations)
cv::Mat	m_oMeanMinDistFrame_LT
	per-pixel mean minimal distances from the model ('D_min(x)' in PBAS, used to control variation magnitude and direction of 'T(x)' and 'R(x)')
cv::Mat	m_oMeanMinDistFrame_ST
cv::Mat	m_oMeanDownSampledLastDistFrame_LT
	per-pixel mean downsampled distances between consecutive frames (used to analyze camera movement and force global model resets automatically)
cv::Mat	m_oMeanDownSampledLastDistFrame_ST
cv::Mat	m_oMeanRawSegmResFrame_LT
	per-pixel mean raw segmentation results (used to detect unstable segmentation regions)
cv::Mat	m_oMeanRawSegmResFrame_ST
cv::Mat	m_oMeanFinalSegmResFrame_LT
	per-pixel mean raw segmentation results (used to detect unstable segmentation regions)
cv::Mat	m_oMeanFinalSegmResFrame_ST
cv::Mat	m_oUnstableRegionMask
	a lookup map used to keep track of unstable regions (based on segm. noise & local dist. thresholds)
cv::Mat	m_oBlinksFrame
	per-pixel blink detection map ('Z(x)')
cv::Mat	m_oDownSampledFrame_MotionAnalysis
	pre-allocated matrix used to downsample the input frame when needed
cv::Mat	m_oLastRawFGMask
	the foreground mask generated by the method at [t-1] (without post-proc, used for blinking px detection)
cv::Mat	m_oFGMask_PreFlood
	pre-allocated CV_8UC1 matrices used to speed up morph ops
cv::Mat	m_oFGMask_FloodedHoles
cv::Mat	m_oLastFGMask_dilated
cv::Mat	m_oLastFGMask_dilated_inverted
cv::Mat	m_oCurrRawFGBlinkMask
cv::Mat	m_oLastRawFGBlinkMask
cv::Mat	m_oTempGlobalWordWeightDiffFactor
cv::Mat	m_oMorphExStructElement
cv::Mat	featureMap
	Features and reliability Maps.
cv::Mat	reliabilityMap
cv::Mat	foregroundRaw
cv::Mat	reliabilityMap2
cv::Mat	featureMap3
cv::Mat	reliabilityMap3
ReliabilityBg	reliabilityBg
float	feature1
float	reliability1
float	feature2a
float	reliability2a
float	feature2b
float	reliability2b
float	feature3
float	reliability3
float	minScale
float	maxScale
float	threshold
float	scaleFactor
Protected Attributes inherited from BackgroundSubtractorLBSP
cv::Mat	m_oROI
	background model ROI used for LBSP descriptor extraction (specific to the input image size)
cv::Size	m_oImgSize
	input image size
size_t	m_nImgChannels
	input image channel size
int	m_nImgType
	input image type
const size_t	m_nLBSPThresholdOffset
	LBSP internal threshold offset value, used to reduce texture noise in dark regions.
const float	m_fRelLBSPThreshold
	LBSP relative internal threshold (kept here since we don't keep an LBSP object)
size_t	m_nTotPxCount
	total number of pixels (depends on the input frame size) & total number of relevant pixels
size_t	m_nTotRelevantPxCount
size_t	m_nFrameIndex
	current frame index, frame count since last model reset & model reset cooldown counters
size_t	m_nFramesSinceLastReset
size_t	m_nModelResetCooldown
size_t	m_anLBSPThreshold_8bitLUT [UCHAR_MAX+1]
	pre-allocated internal LBSP threshold values LUT for all possible 8-bit intensities
size_t *	m_aPxIdxLUT
	internal pixel index LUT for all relevant analysis regions (based on the provided ROI)
PxInfoBase *	m_aPxInfoLUT
	internal pixel info LUT for all possible pixel indexes
const int	m_nDefaultMedianBlurKernelSize
	default kernel size for median blur post-proc filtering
bool	m_bInitialized
	specifies whether the algorithm is fully initialized or not
bool	m_bAutoModelResetEnabled
	specifies whether automatic model resets are enabled or not
bool	m_bUsingMovingCamera
	specifies whether the camera is considered moving or not
cv::Mat	m_oLastColorFrame
	copy of latest pixel intensities (used when refreshing model)
cv::Mat	m_oLastDescFrame
	copy of latest descriptors (used when refreshing model)
cv::Mat	m_oLastFGMask
	the foreground mask generated by the method at [t-1]

Additional Inherited Members
Public Attributes inherited from BackgroundSubtractorLBSP
int	m_nDebugCoordX
int	m_nDebugCoordY
std::string	m_sDebugName
cv::FileStorage *	m_pDebugFS

Detailed Description

Pixel-based Adaptive Word Consensus Segmenter (PAWCS) change detection algorithm.

Note: both grayscale and RGB/BGR images may be used with this extractor (parameters are adjusted automatically). For optimal grayscale results, use CV_8UC1 frames instead of CV_8UC3.

For more details on the different parameters or on the algorithm itself, see P.-L. St-Charles et al., "A Self-Adjusting Approach to Change Detection Based on Background Word Consensus", in WACV 2015.

This algorithm is currently NOT thread-safe.

---

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

• src/BackgroundSubtractorPAWCS.h
• src/BackgroundSubtractorPAWCS.cpp