中文摘要：

提出了运用多核支持向量回归（MK-SVR）算法构建小麦叶面积指数（LAI）遥感监测模型。以2010—2013年试验样点小麦拔节、孕穗、开花3期的实测LAI数据为基础,同步获取我国自主研发的环境减灾卫星HJ-CCD对该研究区域的影像数据,分析了各生育期小麦LAI与8种植被指数间的相关性。以显著相关的植被指数作为输入参数,使用MK-SVR算法构建了每个生育期的小麦LAI反演模型,即MK-SVR-LAI模型。为了评价模型,每期使用单一核支持向量回归（SK-SVR）、偏最小二乘（PLS）回归算法构建了SK-SVR-LAI、PLS-LAI模型。将模型估算LAI值和田间观测LAI值进行比对,以决定系数（R2）和均方根误差（RMSE）为指标评价并比较了模型。结果表明：3个生育期MK-SVR-LAI模型的RMSE值均低于参比模型,拔节期为0.293 1,孕穗期为0.466 8,开花期为0.548 6,且该模型的R2也都最高,拔节期为0.762 4,孕穗期为0.801 8,开花期为0.668 9。

英文摘要：

The multi-kernel support vector regression （ MK - SVR） was used to construct remote sensing monitoring algorithmic models for estimating leaf area index （LAI） in wheat. The experiment was carried out during 2010--2013 in Jiangsu Province, China. Based on LAI in wheat and synchronous China＇ s domestic HJ - CCD multi-spectral data at jointing stage, booting stage and anthesis stage respectively, the relationships between LAI and eight vegetation indices were analyzed at corresponding period. Taking these vegetation indices which were significantly related to LNC at 0.01 level as input parameters, the remotely estimating model was established based on MK - SVR to invert LAI, named the MK - SVR - LAI model. Meanwhile, in order to evaluate the MK - SVR - LAI model, single kernel support vector regression （SK - SVR ） and partial least squares （PLS） were employed to establish models at each period, named the SK - SVR - LAI and PLS - LAI models. Comparing predicted LAI by model with actual measured LAI, the coefficient of determination （R2） and root mean square error （RMSE） were used to evaluate models. The results showed that the lowest RMSE and the highest R2 were obtained by using MK -SVR - LAI model at each stage, of which the RMSE and the R2 were 0. 293 1 and 0. 762 4 at jointing stage, 0. 466 8 and 0. 801 8 at booting stage, 0. 548 6 and 0. 668 9 at anthesis stage, respectively.