Under the background of carbon neutrality, the deep peak shaving and flexible operation of coal-fired generating units put forward urgent requirements for real-time monitoring of the three-dimensional combustion conditions in the furnace. This paper summarized the research progress of three-dimensional combustion temperature distribution monitoring of coal-fired power plant boilers and industrial furnaces. As for the radiation imaging model of combustion flame, the DRESOR method for directional radiation intensity calculation based on Monte Carlo method and the recent optimization of DRESOR method were mainly introduced, which laid a good foundation for improving the inversion accuracy of combustion temperature and inversion of the distribution of radiation characteristic parameters of combustion medium. The basic method to solve the simultaneous inversion problem of three-dimensional temperature field and radiation parameters is to reconstruct the temperature distribution in the furnace from the monochromatic radiative intensity images by Tikhonov regularization method. Then the radiative properties of the particle medium are updated with optimization method and solved iteratively. Recently, there have been new developments in the inversion algorithm. The new algorithm can be divided into three stages. Firstly, assuming uniform distribution of absorption coefficient, scattering coefficient and reflectivity of the furnace wall, the optimal radiation parameters and temperature distribution inside the furnace are obtained by the optimization solution. Secondly, on the basis of stage 1, the absorption and scattering coefficients in the furnace are set as second-order polynomial fitting distributions in spatial coordinates, and the walls are still set with uniform reflectivity to further optimize the iterative calculation. Finally, on the basis of the convergence of the calculation in stage 2, the second-order polynomial distribution of the reflectivity of the furnace wall in wall coordinates is further assumed, and then the calculation is optimized iteratively. The latest development of the inversion algorithm has obtained the reconstruction result of the combustion temperature with reconstruction error within 1%, and realizes the reconstruction of the relative distribution of pulverized coal concentration in the furnace based on the radiative properties. The monitoring systems of three-dimensional temperature field in the furnace has been industrially applied in the combustion monitoring of 200, 300 and 600 MW coal-fired power plant boilers and further expanded to oil-fired or gas-fired industrial kilns such as walking furnace in rolling mill, tubular furnace in petrochemical plant, single burner combustion furnace and cracking furnace in chemical plant, showing a good application prospect. In the future, machine learning and artificial intelligence theory need to be adopted to further improve the efficiency of solution of the coupled reconstruction problem, combined with three-dimensional real-time and dynamic modeling of furnace conditions and thermal system, to realize real-time monitoring and diagnosis of the parameters of the three-dimensional furnace conditions distribution (furnace atmosphere, particulate matter, pollutants, furnace heat load, furnace wall heat load distribution, etc.) and modeling and prediction of the parameters of the distribution of hydrodynamic and thermal systems in the boiler water wall, to further build a multi-timescale big data-driven digital twin system for coal-fired generating units,contributing to the development of the smart boiler and furnace optimization control system.