boofcv.alg.background

Class BackgroundModelMoving<T extends boofcv.struct.image.ImageBase,MotionModel extends georegression.struct.InvertibleTransform<MotionModel>>

java.lang.Object

boofcv.alg.background.BackgroundModel<T>

boofcv.alg.background.BackgroundModelMoving<T,MotionModel>

Direct Known Subclasses:

BackgroundMovingBasic, BackgroundMovingGaussian

public abstract class BackgroundModelMoving<T extends boofcv.struct.image.ImageBase,MotionModel extends georegression.struct.InvertibleTransform<MotionModel>>
extends BackgroundModel<T>

Base class for classifying pixels and background based on the apparent motion of pixels when the camera is moving. The camera motion is provided externally.

There are three coordinate systems.

1. World is the background model's coordinate system.
2. Home is the image which all the camera motion is relative to
3. Current is the location of the current image.

The background model is composed of a single fixed sized image. The background image size is specified in the initialize(int, int, InvertibleTransform) function. After that the background model is updated by calling updateBackground(InvertibleTransform, ImageBase). To flag pixels as background/motion call segment(InvertibleTransform, ImageBase, GrayU8).

If a pixel in the current frame has no corresponding pixel in the background or the background pixel hasn't been observed yet then it will be assigned a special value, which is user configurable. The default value is 0, which is a background pixel. See BackgroundModel.setUnknownValue(int)}

Field Summary

Fields

Modifier and Type	Field and Description
protected int	backgroundHeight
protected int	backgroundWidth
protected georegression.struct.point.Point2D_F32[]	corners
protected MotionModel	currentToWorld
protected MotionModel	homeToWorld
protected boofcv.struct.distort.PointTransformModel_F32<MotionModel>	transform
protected georegression.struct.point.Point2D_F32	work
protected MotionModel	worldToCurrent
protected MotionModel	worldToHome

Fields inherited from class boofcv.alg.background.BackgroundModel

imageType, unknownValue

Constructor Summary

Constructors

Constructor and Description
BackgroundModelMoving(boofcv.struct.distort.PointTransformModel_F32<MotionModel> transform, boofcv.struct.image.ImageType<T> imageType) Constructor which provides the motion model and image type

Method Summary

Modifier and Type	Method and Description
protected abstract void	_segment(MotionModel currentToWorld, T frame, boofcv.struct.image.GrayU8 segmented)
abstract void	initialize(int backgroundWidth, int backgroundHeight, MotionModel homeToWorld) Initializes background model.
void	segment(MotionModel homeToCurrent, T frame, boofcv.struct.image.GrayU8 segmented) Invoke to use the background image to segment the current frame into background and foreground pixels
protected abstract void	updateBackground(int x0, int y0, int x1, int y1, T frame) Call to update the background with the frame inside the bounding box.
void	updateBackground(MotionModel homeToCurrent, T frame) Updates the background with new image information.

Methods inherited from class boofcv.alg.background.BackgroundModel

getImageType, getUnknownValue, reset, setUnknownValue

Methods inherited from class java.lang.Object

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

Field Detail

transform

protected boofcv.struct.distort.PointTransformModel_F32<MotionModel extends georegression.struct.InvertibleTransform<MotionModel>> transform

homeToWorld

protected MotionModel extends georegression.struct.InvertibleTransform<MotionModel> homeToWorld

worldToHome

protected MotionModel extends georegression.struct.InvertibleTransform<MotionModel> worldToHome

currentToWorld

protected MotionModel extends georegression.struct.InvertibleTransform<MotionModel> currentToWorld

worldToCurrent

protected MotionModel extends georegression.struct.InvertibleTransform<MotionModel> worldToCurrent

backgroundWidth

protected int backgroundWidth

backgroundHeight

protected int backgroundHeight

corners

protected georegression.struct.point.Point2D_F32[] corners

work

protected georegression.struct.point.Point2D_F32 work

Constructor Detail

BackgroundModelMoving

public BackgroundModelMoving(boofcv.struct.distort.PointTransformModel_F32<MotionModel> transform,
                             boofcv.struct.image.ImageType<T> imageType)

Constructor which provides the motion model and image type

Parameters:

transform - Point transform which can be used to apply the motion model

imageType - Type of input image

Method Detail

initialize

public abstract void initialize(int backgroundWidth,
                                int backgroundHeight,
                                MotionModel homeToWorld)

Initializes background model. Specifies the size of the background image and transform from the "home" image to the background "world"

Parameters:

backgroundWidth - Width of background

backgroundHeight - Height of background

homeToWorld - Transform from home to world.

updateBackground

public void updateBackground(MotionModel homeToCurrent,
                             T frame)

Updates the background with new image information.

Parameters:

homeToCurrent - Transform from home image to the current image

frame - The current image in the sequence

updateBackground

protected abstract void updateBackground(int x0,
                                         int y0,
                                         int x1,
                                         int y1,
                                         T frame)

Call to update the background with the frame inside the bounding box. Implementing class needs to make sure the rectangle is inside the background.

segment

public void segment(MotionModel homeToCurrent,
                    T frame,
                    boofcv.struct.image.GrayU8 segmented)

Invoke to use the background image to segment the current frame into background and foreground pixels

Parameters:

homeToCurrent - Transform from home image to the current image

frame - current image

segmented - Segmented image. 0 = background, 1 = foreground/moving

_segment

protected abstract void _segment(MotionModel currentToWorld,
                                 T frame,
                                 boofcv.struct.image.GrayU8 segmented)