An adaptive learning rate GMM for background extraction

Abstract

The rapidness and stability of background extraction from image sequences are incompatible, that is, when a conventional Gaussian mixture models (GMM)is used to rebuild the background, if the background regions of the scene are changed, the extracted background becomes bad until the transition is over. A novel adaptive method is presented to adjust the learning rate of GMM in a Hilbert space. The background extraction is treated as a process of approaching to a certain point in the Hilbert space, so the real-time learning rate can be obtained by calculating the distance between the two adjacent extracted background images, and a judgment method of the stability of background is got too. Compared with conventional GMM, the method has both high rapidness and good stability at the same time, and it can adjust the learning rate online. The experiment shows that it is better than conventional GMM, especially in the transition process of background extraction.

References

[1] Stauffer C, Grimson W.E.L. Proceedings of 1999 IEEE Computer Society Conference on Computer vision and Pattern Recognition. IEEE Comput. Soc. Part, 2 (1999)

[2] StaufferC., GrimsonW.E.L.. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2000, 22: 747

[3] P.Wayne Power, J. A. Schoonees, J. Proc. of IVCNZ, (2002), 267.

[4] RajaY., McKennaS.J., GongS.. Computer Vision-ECCV’98. 5th European Conference on Computer Vision, 1998, 1: 460

[5] ZhouD., ZhangH.. IEEE International Conference on Systems, Man and Cybernetics, 2005, 3: 2224

[6] KaewTraKulPong P, Bowden R., Proceedings of the 2nd European Workshop on Advanced Video Based Surveillance Systems, (2001)