支持向量机回归SVR（Support Vector Regression）方法作为叶面积指数反演的一种新思路，在LAI反演中具有一定的应用价值和前景，但SVR算法中惩罚系数C、核函数宽度参数g、不敏感损失函数参数ε的取值对回归精度有显著的影响。本文提出了一种基于人工蜂群算法ABC（Artificial Bee Colony）优化SVR参数的遥感影像叶面积指数反演方法。研究数据为美国土壤水分实验（SMEX02）2002年LAI实测数据和同期的Landsat 7 ETM+地表反射率数据，为了验证ABC算法优化SVR各个参数对反演精度的影响，建立了未优化参数（SVR）、优化单个参数（ABC-SVR-C，ABC-SVR-g，ABC-SVR-ε）、优化3个参数（ABC-SVR）的3类LAI反演模型，并比较了其回归拟合精度。在此基础上，分析了3个关键参数对LAI反演模型精度的敏感性，并对ABC算法优化SVR模型的精度进行显著性检验。研究表明：（1）相比未优化参数模型，ABC算法优化模型具有更高的反演精度，优化3个参数优于优化单个参数，回归直线斜率k达到0.797、决定系数r²达到0.775。（2）SVR的3个关键参数对模型精度都有影响，相较参数C和g，参数ε引起模型精度的不确定性更高。（3）95%的置信区间下，ABC-SVR模型与SVR模型的回归直线斜率k、r²、RMSE的差异显著性检验P值均小于0.005，ABC算法显著改善了SVR模型的精度。

Support Vector Regression (SVR) method as a new idea in LAI inversion has certain application value and prospect. However, the value of penalty coefficient C, width parameter g of kernel function and insensitive loss function parameterεin the SVR algorithm have a significant impact on regression accuracy. This paper proposed a method for Leaf Area Index (LAI) inversion using remote sensing images based on ABC (Artificial Bee Colony) algorithm to optimize SVR parameters. In addition, the LAI measurement values were from the Soil Moisture Experiment 2002 in US (SMEX02) and Landsat 7 ETM + surface reflectance data at the same time. In order to verify the effect of SVR optimized by ABC, this paper established three types of LAI inversion models with non-optimized parameters(SVR), optimized single parameter(ABC-SVR-C, ABC-SVR-g, ABC-SVR-ε), and optimized three parameters (ABC-SVR),and compared the accuracy of the three kinds of models. Based on this, we analyzed the sensitivity of LAI inversion model of three key parameters of SVR, and did a significant test on the accuracy of the ABC algorithm optimized SVR model. The study showed: (1) Compared with the model without optimizing parameters, the four models with the SVR parameters optimized by ABC algorithm had higher accuracy, and the optimized three parameters model had better accuracy than the model with optimizing single parameter, the slope of regression straight line reaching 0.797 and decision coefficient reaching 0.775. (2) The three key parameters of SVR have an influence on the accuracy of the LAI model, and compared with the parameters C and g, the parameter ε is more uncertain to the accuracy of the model. (3) At the confidence interval of 95%, the P value of difference significance test on the slope k, r², and RMSE between ABC-SVR model and SVR model all less than 0.005, indicated that the ABC algorithm significantly improved the accuracy of the SVR model.