Flame recognition in video

作者: PI Walter , M Shah , NV Lobo

DOI: 10.1016/S0167-8655(01)00135-0

关键词:

摘要: This paper presents an automatic system for fire detection in video sequences. There are several previous methods to detect fire, however, all except two use spectroscopy or particle sensors. The that visual information suffer from the inability cope with a moving camera scene. One of these is not able work on general data, such as movie other too simplistic and unrestrictive determining what considered fire; so it can be used reliably only aircraft dry bays. We propose uses color motion computed sequences locate fire. done by first using approach based upon creating Gaussian-smoothed histogram fire-colored pixels, then temporal variation pixels determine which actually pixels. Next, some spurious automatically removed erode operation, missing found region growing method. Unlike vision-based detection, our method applicable more areas because its insensitivity motion. Two specific applications possible algorithms recognition presence global scene movies rating system. show works variety conditions, when has insufficient information.

参考文章(4)
R. Kjeldsen, J. Kender, Finding skin in color images Proceedings of the Second International Conference on Automatic Face and Gesture Recognition. pp. 312- 317 ,(1996) , 10.1109/AFGR.1996.557283
G. Healey, D. Slater, T. Lin, B. Drda, A.D. Goedeke, A system for real-time fire detection computer vision and pattern recognition. pp. 605- 606 ,(1993) , 10.1109/CVPR.1993.341064