Dynamic Characterization of Pulse Combustion by Image Series Processing

Identification of CO$$_{\mathrm{2}}$$ and O$$_{\mathrm{2}}$$ emissions dynamics in a natural gas furnace through flame images, ARMAX models, and Kalman filtering

“…Since the flames luminosity is affected by the equivalence ratio and their shape essentially reflects a stochastic process, distinct threshold values are used to carry out the segmentation process, that was aimed at building a "mask" to distinguish between the objects of interest (flames) from the furnace background. Then, towards reducing the computational load and translating the images into quantitative descriptors, 10 geometrical and luminosity characteristics which represent operational parameters of flame images usually mentioned in the literature [18,22,23] are extracted. These 10 features, calculated over the time series of the images, were submitted to a correlation analysis with the collected O 2 and CO 2 emissions, in order to discard those that presented poor relation with the emitted gases, i.e., the ones that do not go significantly outside the confidence region.…”

“…• Centroid ( u 4 , u 5 ) defined as the point C(C x , C y ) with the coordinates in which the origin of Oxy reference frame is the pixel located at the center of the burner, R f is the flame region, x i is the distance from the origin to the center of the pixel (i, j) in the direction of the Ox axis and y j in the direction of the Oy axis. • [22] ( u 3 ) dispersion of the distribution of the flame area is indicated by the standard deviation of the location of each pixel in relation to its centroid. It is given by…”