Fire Detection Algorithms Using Multimodal ... - Bilkent University

List of Figures2.1 Flicker frequency distribution for a pixel at the boundary of a flameregion in a color-video clip recorded at 25 fps.This frequencydistribution is obtained by analyzing the temporal variations inthe red channel value of the pixel. . . . . . . . . . . . . . . . . . . 102.2 (a) A sample flame pixel process in RGB space, and (b) the spherescentered at the means of the Gaussian distributions with radiustwice the standard deviation. . . . . . . . . . . . . . . . . . . . . 152.3 A two-stage filter bank. HPF and LPF represent half-band highpassand low-pass filters, with filter coefficients {− 1 4 , 1 2 , − 1 4 } and{ 1, 1, 1 }, respectively. This filter bank is used for wavelet analysis. 184 2 42.4 (a) Temporal variation of image pixel at location x = (111, 34),Ĩ n (x).The pixel at x = (111, 34) is part of a flame for imageframes I(x, n), n=1, 2, 3, 19, 23, 24, 41 and 50. It becomes partof the background for n = 12,..., 17, 20, 21, 26, 27, 31,..., 39, 45,52,..., and 60. Wavelet domain subsignals (b) d n and (c) e n revealthe fluctuations of the pixel at location x = (111, 34). . . . . . . 202.5 (a) Temporal history of the pixel at location x = (18, 34). It ispart of a fire-colored object for n = 4, 5, 6, 7, and 8, and it becomespart of the background afterwards. Corresponding subsignals (b)d n and (c) e n exhibit stationary behavior for n > 8. . . . . . . . . 21xiii