OpenCV C++: How to slow down background adaptation of BackgroundSubtractorMOG?

Question

I am using BackgroundSubtractorMOG in OpenCV to track objects. When they appear, it works fine but the background fastly adapts so I cannot track static objects. How can I make the background adaptation slower (I dont want it fully static, just slower)?

Setting the learning rate using the constructor doesn't change that:

BackgroundSubtractorMOG pBSMOG = BackgroundSubtractorMOG(???);

How can I solve this? Thanks!

Answer 1

BackgroundSubtractorMOG pBSMOG = BackgroundSubtractorMOG(int history=200, int nmixtures=5, double backgroundRatio=0.7, double noiseSigma=0);

Where,

• history – Length of the history.
• nmixtures – Number of Gaussian mixtures.
• backgroundRatio – Background ratio.
• noiseSigma – Noise strength (standard deviation of the brightness or each color channel). 0 means some automatic value.

Increasing the history value will slow down the adaptation rate.

There is another function available in OpenCV:

Ptr <BackgroundSubtractorMOG2> createBackgroundSubtractorMOG2(int
history=500, double varThreshold=16, bool detectShadows=true )

This is much faster than the previous one and it can eleminate detecting shadows too.