Multi-spectral Radiation Temperature Inversion Method Based on the Statistical Distribution of the Predicted Temperature

Accurate temperature measurement is essential for quality control and industrial process optimization. However, a major challenge in the temperature measurement arises when the emissivity of the object is unknown. Multi-wavelength temperature measurement has emerged as an effective solution to address this issue, as it allows for the simultaneous consideration of multiple wavelengths, compensating for unknown emissivity variations. In this work, a novel multi-wavelength radiation temperature inversion algorithm is proposed for the simultaneous determination of both the target temperature and spectral emissivity. This method is based on the assumption that the spectral emissivity of the object follows an exponential polynomial model. By utilizing different wavelength combinations, multiple overdetermined equations are constructed and the conjugate gradient least squares method is employed to solve these equations to obtain predicted temperatures. After acquiring multiple predicted temperatures, the adaptive kernel density estimation method is applied to model the probability density distribution of the predicted temperatures. Finally, the final temperature prediction is determined as the weighted average of all temperatures above a predefined threshold and emissivity is further obtained based on the final predicted temperature. The accuracy and reliability of this method are validated through simulation experiments under different emissivity models, as well as actual emissivity measurement data from four samples, and experimental results demonstrate that the method can offer a promising solution for overcoming the challenges associated with the traditional temperature measurement techniques.