12.07.2015 Views

Background Subtraction Using Ensembles of Classifiers with an ...

Background Subtraction Using Ensembles of Classifiers with an ...

Background Subtraction Using Ensembles of Classifiers with an ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

CHAPTER 2RELATED WORKS2.1 Early Work in <strong>Background</strong> ModelingBecause <strong>of</strong> the computational dem<strong>an</strong>ds that dynamic background modeling incurs, progress inthe field has mainly been over the past decade, paralleling the explosion <strong>of</strong> faster computers. Priorto the current robust models, simple background estimation was performed where the backgroundwas assumed to be static. In these schemes <strong>an</strong> algorithm similar to the one in Equation 2.1 wasused, where B(p) is the estimated background pixel value at pixel p, n is the number <strong>of</strong> framesused to build the background model, <strong>an</strong>d I t (p) is the value <strong>of</strong> pixel p at time t.B(p) =n∑i=0I t (p)n(2.1)For a future frame t, pixel p is then predicted to be foreground (FG) or background (BG)based on Equation 2.2, where τ is the threshold (typically set around 50).if |I t (p) − B(p)| ≥ τ then F G else BG (2.2)The largest adv<strong>an</strong>tage <strong>of</strong> this technique is that it is fast <strong>an</strong>d has very low memory requirementsbecause the images may be deleted after the average is calculated. The major problem <strong>with</strong> such<strong>an</strong> algorithm is that it does not adapt to its environment, <strong>an</strong>d it does not have the ability todetect dynamic background regions. The failure to adapt is eliminated if a running average istaken instead. The running average uses a learning rate α (where 0 < α < 1), <strong>an</strong>d updates thebackground model <strong>with</strong> a new frame I new using Equation 2.3:B new (p) = α · I new (p) + (1 − α) · B old (p) (2.3)7

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!