Adaptive multi-wavelength temperature retrieval method

This paper addresses the limitations of traditional multi-spectral temperature measurement methods, such as low computational efficiency and sensitivity to noise, by introducing an efficient temperature inversion method based on a collaborative design of an adaptive wavelength selection mechanism and a differential evolution optimization algorithm. By dynamically eliminating wavelength channels with low signal-to-noise ratios, the method reduces ineffective computation by over 30%. Furthermore, the global search and parallel processing capabilities of the differential evolution algorithm significantly enhance the accuracy and efficiency of temperature inversion. Experimental results demonstrate that, when applied to black powder data, the temperature standard deviation is optimized from ±85 K using the traditional dual annealing algorithm to ±23 K. The processing time for a single frame (149×280 pixels) is reduced from 22.5 minutes to 4.1 minutes. This research provides a high-precision, high-efficiency temperature measurement solution for scenarios with strict real-time requirements, such as industrial combustion diagnostics and high-temperature material detection.