Trejo-Morales, A., Córdova-Esparza, D.M., Rosas-Raya, C., Herrera-Navarro, A.M., Jiménez-Hernández, H.
The background subtraction process locates and labels non-moving and moving areas in an image sequence. The efficiency and performance of a given background subtraction method depends on the variables, model parameters and scene conditions. This paper describes the parametric effects under different operating conditions, for the subtraction Mixture of Gaussians model, applied to road monitoring videos. The parameters analyzed include the size of the noise reduction filters, the granularity of the motion detection and the convergence speed of the background estimation. The results show the considerations to be taken in assigning values to the different parameters, to ensure a correct detection of moving and stationary zones, allowing to optimize motion detection (controlled and uncontrolled), for example outdoor scenarios (uncontrolled) or motion detection in production lines where it is complicated to control light variations.
Key words: Background subtraction, Mixture of gaussian, Parametric fit, Surveillance system.