由于热红外多波段数据间具有高度的相关性和混合像元的大量存在，使得多波段陆面温度反演精度难以提高，并且难以得到组分温度信息。在连续植被热辐射方向性规律上的基础上，以喜直型连续植被为例，进行了大量的Monte-Carlo模拟，建立了组分有效比辐射率与土壤表面比辐射率和植被叶面积指数之间的经验函数关系，并以此构造目标函数，采用遗传算法，从热红外多角度数据中，同时反演混合像元组分温度和土壤比辐射率以及叶面积指数。通过对模拟的观测数据进行遗传算法反演的大量试验，结果表明，遗传算法反演组分温度非常稳健，在宽松的先验知识条件下，遗传算法可以解决不确定性反演问题。遗传算法反演结果和野外实测数据作了比较，证实了反演原理的正确性，为基于热红外方向性辐射模型反演组分温度，提供了新方法。

Due to high correlation coefficients among multi-channel thermal infrared data and mixed pixels widely existed, it is difficult to improve the accuracy of retrieved land surface temperature; further more, component temperature can not be retrieved from multi-channel thermal infrared data. In this paper, taken erectophile type continuous vegetation as an example, we did many Monte-Carlo simulations, and established empirical analytic expressions of component effective emissivities with soil emissivity and leaf area index. Empirical analytic expressions were used to construct objective function, and genetic algorithm was employed to synchronously retrieve component temperature, soil emissivity and LAI from thermal infrared multi-angle data. Many experiments of genetic algorithm inversion from simulated data were conducted, results show that it is very robust to retrieve component temperature using genetic algorithm, and genetic algorithm can cope with uncertainty inversion problem pretty well if we take full advantage of priori knowledge. Comparison between inversion results and ground-truth data were made. This paper offers a new method to retrieve component temperature from multiangle thermal infrared data based on the model of directionality of thermal radiance