An adaptive background reconstruction algorithm based on inertial filtering

Abstract

To improve the detecting effects of moving objects, an adaptive background reconstruction algorithm based on inertial filtering is proposed in this paper. According to different properties of the moving foreground and ever-changing background, the current frame is added to the background with a specific weight value. So the background can not only keep steady, but also be reconstructed at a specific speed. Experimental results show that the algorithm can reconstruct the background quickly and effectively whenever the background changes slowly or suddenly, or the background is mixed with moving foreground, and it can improve the veracity and robustness of objects detection effectively.

Keywords

Gaussian Mixture Model / Background Subtraction / Object Detection / Specific Weight / Current Frame

References

[1] HornB.K., SchunkB.G.. Artificial Intelligence, 1981, 17: 185

[2] SmithS.M., BradyJ.M.. IEEE Trans. on PAMI, 1995, 17: 814

[3] NeriA., ColonneseS., RussoG., TaloneP.. Signal Processing, 1998, 66: 219

[4] MeierT., NganK.N.. IEEE Trans. on Circuits and Systems for Video Technology, 1998, 8: 525

[5] Ridder C, Munkelt O and Kirchner H, Proc. of the Int’l Conf. on Recent Advances in Mechatronics, ICRAM’95, 1995, 193.

[6] Friedman N and Russell S, Proceedings of the 13th Conference on Uncertainty in Artificial Intelligence, 1997, 175.

[7] ZhangM. J., LiC. H., LiuM. Y.. Computer Engineering and Application, 2005, 41: 27

[8] LongW., YangY.. Pattern Recognition, 1990, 23: 1351

[9] KornprobstP., DericheR., AubertG.. Journal of Mathematical Imaging and Vision, 1999, 11: 5

[10] GloyerB., AghajanH.K., SiuK.Y., KailathT.. Proc. of the IS&T-SPIE Symp.on Electronic Imaging:Image and Video Processing, 1995, 2421: 173

[11] HouZ. Q., HanC. Z.. Journal of Software, 2005, 16: 1568

[12] ZhaoY., ChangF. L.. Computer Engineering and Application, 2008, 44: 104