Low-carbon chemical fire accidents have a high risk and great harm. Exploring the flame spectrum characteristics of low-carbon chemicals is of great significance in detecting and identifying such fire hazards and pollution. However, domestic and foreign large-scale low-carbon chemical fire accidents produce toxic and harmful sulfur Research on SO_X and NO_X gases is rare. In this paper, by building an experimental platform for flame spectrum testing in the 1.2~12 μm infrared band, the flame spectrum test of carbon disulfide, 92# gasoline and alcohol at three different combustion scales of 5, 14 and 20 cm is carried out to explore the effect of flame combustion scale on high-temperature flame molecular radiation The influence of the spectrum. As the combustion scale increases, the flame radiation intensity increases, and the characteristic waveband appears to broaden. Analyze the different flame spectrum characteristics of liquefied natural gas (LNG), acrylonitrile, acetonitrile and 95# gasoline in the four typical chemicals at the 5 cm combustion scale. Using Fourier transform infrared spectrometer to measure the different temperatures of the high-temperature blackbody furnace, the flame spectrum signal is radiated calibration, and the accurate radiant calibration coefficient is obtained, thereby obtaining the radiance value emitted by the high-temperature flame molecules. Moreover, compared with the HITRAN database simulated atmospheric pressure 1 atm, temperature 1 300 K single SO₂, H₂O, CO₂, NO₂ molecular radiation spectrum for comparative analysis. Among them, the high temperature flame molecular spectrum mainly has 7.3~7.6, 8.7 and 4.0 μm SO₂ bands, 1.8~2.1 and 6.4 μm H₂O bands, 4.2~4.6 μm CO₂ bands, and 2.5~2.9 μm H₂O and CO₂ common bands. The high temperature NO₂ gas did not reach the detection limit of the infrared spectrometer, and the 6.0~6.4, 3.4 and 2.4 μm NO₂ bands can be known through the HITRAN database simulation. In order to further distinguish the flame spectra of various chemicals, the calibrated flame spectrum signal is normalized, and the db2 wavelet basis function is used for 6-layer decomposition to obtain the approximate coefficients of the high frequency part and the detail coefficients of the low-frequency part, by comparing different chemistry, the approximation and detail coefficient difference of the high-temperature flame spectrum. The results show that the flame spectrum characteristics of carbon disulfide and the chemical flame spectrum characteristics of wavelet analysis can be used as an important basis for distinguishing low-carbon chemicals from oils and for subsequent remote sensing detection of low-carbon chemical characteristic pollutants, component concentration inversion and identification evaluation Its pollution hazards lay an important foundation.