The study of methods for analysis burning torch infrared images

Abstract

The paper describes the results of the study of methods for analysis burning torch infrared images obtained by an infrared camera in the band of electromagnetic wavelengths of 1.5-5.1 μm. It was shown that the known infrared image analysis methods cannot provide the quantitative parameters extraction that could describe combustion process. In addition, it was figured out that the known methods are time-consuming and cannot run in real time. As a result, nowadays the combustion control system that uses optical control of torch parameters in infrared band cannot be designed. In our study we analyzed the pixels quantity distribution density in the range of [520,560] relative Celsius degrees on each frame of the initial infrared sequence of burning torch. It was shown that the pixels quantity distribution has the bimodal distribution law and can be described by three local extremes coordinates: two maximums and a minimum located between them. The pixels that have relative degrees values in the range from 520 degrees to the value of the minimum’s abscissa and from the value of the minimum’s abscissa to 560 degrees relatively form two separate zones on the burning torch visualization. It was demonstrated that time-domain series constructed from frame-by-frame calculated local extremes coordinates of the P(T) distributions are stationary random sequences. This result allows to use these time-domain series as quantitative parameters of the torch combustion. It was shown that the local minimum’s abscissa value of the P(T) distribution with a relative error of 2.8 % is a constant value equal to 536.3 relative degrees. This allows to count the pixels quantity of each of the separate zones without using time-consuming Rosenblatt – Parzen estimation and run data processing in real time. © 2020 National Research Nuclear University. All rights reserved.