Byung-eun LEE, Thanh-binh NGUYEN, Sun-tae CHUNG
School of Electronics Engineering, University of Soongsil, Seoul 156-743, Korea
Abstract-Foreground moving object detection is an important pr ocess in various computer vision applications such as intelligent visual surveil lance, HCI, object-based video compression, etc. One of the most successful mov ing object detection algorithms is based on Adaptive Gaussian Mixture Model (AGM M). Although AGMM-based object detection shows very good performance with respe ct to object detection accuracy, AGMM is very complex model requiring lots of fl oating-point arithmetic so that it should pay for expensive computational cost. Thus, direct implementation of the AGMM-based object detection for embedded DS Ps without floating-point arithmetic HW support cannot satisfy the real-time p rocessing requirement. This paper presents a novel real-time implementation of adaptive Gaussian mixture model-based moving object detection algorithm for fix ed-point DSPs. In the proposed implementation, in addition to changes of data types into fixed-point ones, magnification of the Gaussian distribution techniq ue is introduced so that the integer and fixed-point arithmetic can be easily a nd consistently utilized instead of real number and floating-point arithmetic i n processing of AGMM algorithm. Experimental results shows that the proposed imp lementation have a high potential in real-time applications.
Key words-background modeling; real-time computing; ob ject detection
Manuscript Number: 1674-8042(2010)02-0116-05
dio: 10.3969/j.issn.1674-8042.2010.02.04
References
[1]T. Chen, H. Haussecker, A. Bovyrin, R. Belenov, K. Rodyushkin, A. Ku ranov, V. Eruhimov, 2005. Computer vision workload analysis: case study of video surveillance systems. Intel Technology Journal, 9(2): 109-1 12.
[2]P. KadewTraKuPong, R. Bowden, 2009. An Improved Adaptive Background Mixture Model for Real-time Tracking with Shadow Detection. Proc. 2nd European Workshop on Advanced Video-Based Surveillance Systems.
[3]C. Stauffer C, W.E.L Grimson, 1999. Adaptive Background Mixture Mode ls for Real-time Tracking. IEEE CVPR, p. 244 - 252.
[4]M. Onoe, 1981. Real Time Parallel Computing Image Analysis. Plenum P ress, New York.
[5]D. Koller, J. Webber, T. Huang, J. Malik, G. Ogasawara, B. Rao, S. R ussel, 1994. Towards Robust Automatic Traffic Scene Analysis in Real Time. Proce edings of International Conference on Pattern Recognition.
[6]A. Elgammal, R. Duraiswami, D. Harwood, L. Davis, 2002. Background a nd foreground modeling using nonparametric kernel density estimation for visual surveillance. Proceeding of IEEE, p.1151-1163.
[7]K. Toyama, B. Brumitt J. Krumm, B. Meyers, 1999. Wallflower: Princip les and Practices of Background, Greece. Proceeding of the 7the IEEE Internation al Conference on Computer Vision, p.255-261.
[8]O. Javed, M. Shah, 2008. Automated Multi-Camera Surveillance: Algor ithms and Practice, Springer, Heidelberg.
[9]B. Nguyen, C. Sun Tae, 2009. An Improved Real Time Blob Detection fo r Visual Surveillance. CISP.
[full text view]
