Fire Detection Algorithms Using Multimodal ... - Bilkent University

CHAPTER 3. FLAME DETECTION IN INFRA-RED (IR) VIDEO 38background image and thresholding. A recursive adaptive threshold estimationis described in [19] as well. Other methods can be also used for moving objectestimation. After moving object detection, it is checked whether the object ishotter than the background, i.e., it is verified if some of the object pixels arehigher in value than the background pixels.Hot objects and regions in IR video can be determined in moving cameras aswell by estimating local maxima in the image. Contours of these high temperatureregions can be determined by region growing.The next step of the proposed method is to determine the center of massof the moving bright object.A one-dimensional (1-D) signal x(θ) is obtainedby computing the distance from the center of mass of the object to the objectboundary for 0 ≤ θ < 2π. In Fig. 3.1, two FLIR (forward looking infra-red) imageframes are shown. Example feature functions for walking man pointed with anarrow and the fire region in Fig. 3.1 are shown in Fig. 4.5 for 64 equally spacedangles x[l] = x(lθ s ), θ s = 2π . To determine the high-frequency content of a curve,64we use a single scale wavelet transform shown in Fig. 4.2. The feature signal x[l]is fed to a filterbank shown in Fig. 4.2 and the low-band signalc[l] = ∑ mh[2l − m]x[m] (3.1)and the high-band subsignalw[l] = ∑ mg[2l − m]x[m] (3.2)are obtained. Coefficients of the lowpass and the highpass filters are h[l] ={ 1, 1, 1} and g[l] = {− 1, 1, − 1 }, respectively [34], [11], [45].4 2 4 4 2 4The absolute values of high-band (wavelet) w[l] and low-band c[l] coefficientsof the fire region and the walking man are shown in Figs. 4.6 and 4.7, respectively.The high-frequency variations of the feature signal of the fire region isclearly distinct from that of the man. Since regular objects have relatively smoothboundaries compared to flames, the high-frequency wavelet coefficients of flameboundary feature signals have more energy than regular objects. Therefore, the