基于发射率模型约束的多光谱辐射真温反演法

温度测量是工业生产或科学实验中保证产品质量、降低生产成本和确保实验安全的重要因素之一。目前非接触的测温方法以辐射测温法为主，二次测量法是多光谱辐射测温中一种常用的方法。但是，二次测量法不适用于实时数据处理。针对此问题，基于多光谱亮度温度数学模型引入了发射率模型约束条件，提出了一种多光谱辐射真温快速反演法。对于非黑体，根据不同波长下的亮度温度的关系，得出当亮度温度在一个区间内是增函数或者常数函数时，发射率在该区间内是增函数;当亮度温度在一个区间内是减函数时，则发射率在该区间内满足一个关于发射率和波长的不等式。该发射率模型约束条件根据亮度温度的信息，将发射率假设值的构建由多类减少到一类，避免了不必要的发射率的构建。实验分别采用实际发射率随波长单调下降、单调上升、先下降再上升、先上升再下降和随机变化的具有代表性的五个被测目标，针对两个被测温度点进行了仿真对比分析。结果表明，与二次测量法相比，对于同一个被测目标，在相同的温度初值和相同的发射率搜索范围下，新算法在保证精度的情况下，不仅所得结果相同，而且处理速度提升了19.16%～43.45%。

Multispectral Radiation Algorithm Based on Emissivity Model Constraints for True Temperature Measurement

Temperature measurement is one of the important factors for ensuring product quality, reducing production cost and ensuring experiment safety in industrial manufacture and scientific experiment. Radiation thermometry is the main method for non-contact temperature measurement. The second measurement (SM) method is one of the common methods in the multispectral radiation thermometry. However, the SM method cannot be applied to on-line data processing. To solve the problems, a rapid inversion method for multispectral radiation true temperature measurement is proposed and constraint conditions of emissivity model are introduced based on the multispectral brightness temperature model. For non-blackbody, it can be drawn that emissivity is an increasing function in the interval if the brightness temperature is an increasing function or a constant function in a range and emissivity satisfies an inequality of emissivity and wavelength in that interval if the brightness temperature is a decreasing function in a range, according to the relationship of brightness temperatures at different wavelengths. The construction of emissivity assumption values is reduced from multiclass to one class and avoiding the unnecessary emissivity construction with emissivity model constraint conditions on the basis of brightness temperature information. Simulation experiments and comparisons for two different temperature points are carried out based on five measured targets with five representative variation trends of real emissivity: decreasing monotonically, increasing monotonically, first decreasing with wavelength and then increasing，first increasing and then decreasing and fluctuating with wavelength randomly. The simulation results show that compared with the SM method, for the same target under the same initial temperature and emissivity search range, the processing speed of the proposed algorithm is increased by 19.16%~43.45% with the same precision and the same calculation results.