多光谱辐射测温技术获取真实温度时, 目标发射率信息是温度求解的关键。 一般解决的方法是基于发射率与波长或温度之间的函数关系建立发射率假设模型。 然而, 当假设模型与实际情况存在偏差时, 会造成较大的温度测量误差。 因此, 消除目标未知发射率的干扰, 减少对发射率模型的依赖, 增加测温算法的通用性, 是多光谱辐射测温技术亟需解决的难题。 提出了改进的粒子群与遗传混合优化算法（HPSOGA）, 算法的核心思想是将多波长辐射测温问题转化为约束优化问题。 首先根据约束条件所设置的范围, 在可行域内生成若干个群, 每个种群对应一组满足条件的光谱发射率, 然后通过HPSOGA算法不断地进化、 迭代操作, 最终寻得最优适应度值的对应解。 该算法实现了在不需要假设发射率模型的情况下, 同时反演出目标的光谱发射率和真实温度。 通过对六种典型的发射率模型进行仿真, 验证了新算法对不同分布趋势的光谱发射率反演的适应性。 结果表明, 在真温800和900 K的情况下, 反演温度的平均相对误差小于0.73%。 最后, 将该算法应用于火箭发动机羽焰温度测量数据的处理。 结果表明, 当设计温度为2 490 K时, 反演温度的相对误差均小于0.65%。 仿真与实验均表明, 新算法可求解出满足一定精度要求的发射率和真温。 因此, 提出的HPSOGA算法是可靠的、 有效的, 为多光谱辐射测温技术测量目标真实温度提供了一种新的思路。

When obtaining the real temperature by multi-spectral radiometric thermometry, the target emissivity information is the key to calculating the temperature. The general solution is to establish an emissivity model based on the function between emissivity and wavelength or temperature. However, when the assumed model deviates from the actual situation, it can cause large temperature measurement errors. Therefore, eliminating the interference of the unknown emissivity of the target, reducing the reliance on the emissivity model, and increasing the universality of the temperature measurement algorithm are the urgent challenges to be solved in multi-spectral radiometric thermometry. This paper propose an improved hybrid optimization algorithm, particle swarm optimization and genetic algorithm(HPSOGA). The core idea of the algorithm is to transform the multi-wavelength radiometric thermometry problem into a constrained optimization problem. Firstly, a group of spectral emissivity satisfying the constraint is initialized, constituting a population. The fitness value is calculated after taking the emissivity into the objective function established by the reference temperature model of multi-spectral radiometric thermometry. The population continuously evolves and iterates in the feasible domain by HPSOGA algorithm until the fitness value is the smallest. The corresponding temperature of each spectral channel is approximately equal. In this algorithm, the spectral emissivity and the real temperature of the target can be inverted simultaneously without assuming an emissivity model. Simulating six typical emissivity models verifies the new algorithms adaptability to the inversion of spectral emissivity with different distribution trends. The results show that the average relative error of the inversion temperature is less than 0.73% for the cases of true temperature 800 and 900 K. Finally, the algorithm is applied to process rocket motor plume flame temperature measurement data. The results show that when the design temperature is 2 490 K, the relative errors of the inverse temperature are less than 0.65%. Both simulation and experiment show that the new algorithm can solve the emissivity and true temperature to meet certain accuracy requirements. Therefore, the HPSOGA algorithm proposed in this paper is reliable and effective and provides a new way formulti-spectral radiometric thermometry to measure the true temperature of the target.