Background subtraction

This Background subtraction algorithm is a more advanced method in comparison to the Differential images method. The key difference is that it uses and calculates a background image. Images taken from the capture device are then subtracted from the background image, resulting in a more robust segmentation of the foreground. As this method is more accurate, it also consumes a lot more CPU power and memory; the background model is adapted by every iteration.

The algorithm is described in following paper.

Parameters

The parameters of the Background images algorithm can be found in the config/algorithm.xml file, but you can also use the web to modify the parameters. Below you see a default configuration file.

<algorithms>
	<BackgroundSubtraction>
		<shadows type="text">true</shadows>
		<history type="number">50</history>
		<nmixtures type="number">5</nmixtures>
		<ratio type="number">1</ratio>
		<erode type="number">5</erode>
		<dilate type="number">7</dilate>
    	<threshold type="number">20</threshold>
    </BackgroundSubtraction>
</algorithms>

Shadows

The shadows parameter is a true or false. By using this parameter you can make the Background subtraction a little bit more intelligent; to neglect shadows in the foreground model.

History

This is an integer which tells the system, the amount of images the system should take into account (history), to calculate the background model. The higher the history integer, the more images it will use to calculate the background model. By increasing the history you get a more stable background model, but a less flexible. As more images from the past are taken into account, the slower the background model will adapt to a new scenario. E.g. when a car drives into your street, and stands still for 5 minutes.

Finding the ideal history parameters, depends heavily in the application you want to use it. In practice this is achieved by trial and error.

NMixtures

Sets the number of gaussian components in the background model.

Ratio

If a foreground pixel keeps semi-constant value for about backgroundRatio*history frames, it’s considered background and added to the model as a center of a new component.

Erode

Find a more detailed explanation at the Differential images section.

Dilate

Dilate is exactly the same as the erode parameter, except that it has the reverse behaviour; instead of removing isolated pixels it will connect them. The dilate function is mainly used in segmentation techniques, to create a more complete segment; it fills gaps. The higher the number of the erode parameter, the more aggressive the dilating will be.

Threshold