Combustion is a very complex chemical reaction, and the flame’s combustion temperature is an important state parameter that reflects the combustion process. It is crucial to study the heat generated by combustion and the heat transfer in space, which affects the dynamics, economy, and emissions of internal combustion engines. Therefore, stable combustion is of great significance to the current topic of energy conservation and emission reduction.
Common methods for measuring the temperature in the cylinder are the contact and non-contact types. When measuring the temperature distribution in the cylinder, the contact temperature measurement method has a large difference from the actual temperature due to the disturbance of the temperature sensing element to the temperature field, the thermal inertia of the sensor and the limitation of the temperature measurement area. Non-contact methods, such as the optical temperature measurement, have the advantage of a wide measuring range, transient response, and high precision, which have made them widely used in temperature measurement in internal combustion engines.
The charge-coupled device (CCD)-based optical temperature measurement technology combines the traditional radiation principle, modern photoelectric conversion principle, and digital image processing technology, and has the advantage of a non-contact, fast response, small error, and real-time dynamic measurement. The working principle of the CCD image sensor is [
1,
2,
3,
4,
5]: When external light illuminates the pixel array of the sensor, and the optical signal is converted into an electrical signal, which is a photoelectric effect. At this time, a moving charge is generated on the internal silicon of the CCD, and a current is formed. The corresponding pixel unit is selected by the logic controller according to the nature of the obtained current, and the overall image is represented by an electrical form signal. The electrical signal is passed to the A/D converter, which, in turn, is converted to a digital signal, a discrete value image signal (color image).
CCD as an important sensor that is easy to connect with a computer and has a wide range of applications in temperature field simulation, velocity measurement, and spectral analysis. In 1932, Hottel and Broughton [
6] used the two-color method to measure flame temperature for the first time. This study pioneered two-color temperature measurement. In this experiment, they tried to get the temperature and total emissivity of the luminous flame. Because of the limitations of the conditions at that time, this method has many shortcomings, such as the inability to realize real-time measurements. In 1979, Cashdollar [
7] used a pyrometer to measure the continuum radiation from particles in a flame at three wavelengths (0.8 μm, 0.9 μm, and 1.0 μm) and calculated the particle temperature from the radiation data using the Planck equation. The HAICS-3000 system was developed by Hitachi in 1985. Flame image recognition technology was used to obtain the distribution of the whole flame temperature field [
8]. Abe et al. [
9] proposed an indirect method to estimate light source color or the correlated color temperature of illuminants. The correlated color temperature of the light source illuminating the color chips is estimated. Ito et al. [
10] presented a quantitative characterization of propane premixed flame color and its applications in their work. Detailed relations between flame colors and flame spectra are investigated in the specific range of the air/fuel ratio. In 1993, BHP Billiton researcher Chen et al. [
11] showed how nearly infrared CCD temperature measurement technology was applied to the steel industry. In 1995, Hsu et al. [
12] used the enhanced CCD to measure the flame temperature of liquid metal and got some results, but the error was large. In 1996, Skarman et al. [
13] used CCD to take a hologram of fluid and reconstruct the three-dimensional temperature field of the fluid by image processing technology. Because of the performance of CCD, the experimental results were not satisfactory. Panagiotou et al. [
14] used a three-color near-infrared optical pyrometer, with wavelengths centered at 998, 810, and 640 nm, to monitor the combustion of polymer particles. Zhao et al. [
15] used laser induced incandescence (LII) to image the two-dimensional soot distribution. Lu et al. [
16] of Greenwich University developed a device to measure the temperature distribution of pulverized coal flame and the concentration of soot in 2001. The experimental results in a furnace show that the device can effectively measure the instantaneous flame temperature and soot concentration. In 2003, Sutter et al. [
17] used CCD to measure the temperature of the radiator by measuring the monochromatic light emitted from the surface of the object. This method was used to measure the temperature of high-speed cutting tools, and the results were satisfactory. Brisley et al. [
18] combined the image-processing techniques and two-color radiation thermometry for temperature measurement of combustion flame in 2005. Lu et al. [
19] presented an imaging-based multicolor pyrometric system for the monitoring of temperature and its distribution in a coal-fired flame. Panditrao and Rege [
20] proposed a novel noncontact temperature measurement technique using a consumer-grade digital still camera. Wu et al. [
21] employed an intensified CCD camera coupled with bandpass filters to capture the quasi-steady state flame emissions of diesel (No. 2) and a diesel-gasoline blend (dieseline: 80% diesel and 20% gasoline by volume) at 430 nm and 470 nm bands in 2016. Liu and Liu [
22] presented an inverse analysis based on the flame emission spectrum technique for the simultaneous reconstruction of two-dimensional (2D) temperature and the concentration fields of soot and metal-oxide nanoparticles in the asymmetric nanofluid fuel flames by means of CCD cameras in 2018. The technique was successfully used for measuring the temperature distribution of different industrial applications, like in a muffle furnace, salt bath furnace, or induction furnace.
However, the above studies used zero-dimensional calibration, in other words, a homogeneous temperature object, to examine the accuracy. There is a difference between the radiation characteristics of the hot metal and the diesel flame. Further, there may be some impacts while processing a two-dimensional image as the program was calibrated zero-dimensionally. The CMOS or CCD sensors with custom filters (often narrow bandwidth filters) are needed for traditional monochrome or bichrome temperature measurement, which would increase the cost of equipment and reduce the generality of data processing. This paper verifies the feasibility of temperature measurement with ordinary cameras, which reduce the cost of measurement and benefit from the two-color method. The calculation program can be used to process diesel flame photographs taken by the same type of cameras in order to obtain more basic data. In addition, the flame temperature characterization of the diesel engine under open-cycle and closed cycle conditions, which represents the working conditions of a submarine, was needed to be further studied and discussed [
23]. In this study, some typical flame images of in-cylinder combustion were recorded using a high-speed CCD camera, and the combustion temperature was calculated and corrected by a three-primary color temperature method. Finally, the temperature of the typical combustion flame images under the open-cycle and closed cycle conditions was compared by the CMS2002 (University of Science and Technology of China, Hefei, China) measurement and the MATLAB (MathWorks, Natick, USA) program, respectively.