A Multispectral Thermometry Based on the Self-adaptive Cuckoo Algorithm

Abstract

Multispectral thermometry stands as a prevalent non-contact method utilized for temperature measurement across various applications. To solve the problem that multispectral thermometry cannot obtain accurate temperature of the target under the unknown spectral emissivity, many scholars have proposed various optimization algorithms. However, there are still problems such as the large emissivity search range, uncertain initial solution and long solution time. To solve the above problems, a new objective function and constraint conditions are established. A self-adaptive cuckoo algorithm is proposed. Real number coding is used to improve the convergence ability and robustness of the algorithm. The adaptive function is used to evaluate the quality of the solution, and the result is avoided to fall into the local optimal solution by the random walk mechanism of Lévy flight. The validity of the proposed self-adaptive cuckoo algorithm is verified by inversion calculation of 6 different emissivity models and zirconia samples. The maximum relative error of self-adaptive cuckoo algorithm is 0.41% in the case of no noise interference, and 0.91% in the case of noise interference. The self-adaptive cuckoo algorithm can still inversion the temperature well with a low signal-to-noise ratio. The experimental results show that the inversion temperature error is less than 0.25%. This method provides a new idea for multispectral temperature measurement.