基于HJ-CCD遥感数据和DK-SVR算法的小麦生物量估算研究

Estimation of wheat biomass based on HJ-CCD remote sensing data and DK-SVR algorithm

基于2010～2013年江苏地区小麦环境减灾卫星(HJ-CCD)的影像数据,提取拔节、孕穗和开花3个生育期的卫星植被指数并作为模型的输入参数,分别利用双核支持向量回归算法(DK-SVR)、MLR和PLS方法构建各生育期小麦生物量估测模型,比较各模型的预测性能。结果表明:可以使用DK-SVR算法遥感估算小麦生物量,基于该算法构建的模型预测性能在3个生育期均优于MLR模型和PLS模型,各期实测值与模型预测值之间决定系数R^2分别为0.50、0.67和0.65,相应的均方根误差RMSE为506、1 389和2 058kg·hm^(-2)。

Based on China’s environmental satellite charge-coupled device(HJ-CCD)image data of wheat from test sites in Jiangsu province of China during 2010-2013,vegetation indices were respectively calculated at the jointing,booting,and anthesis stages.Then,through utilizing DK-SVR,MLR and PLS algorithm,the biomass estimating models for each stage were respectively established based on its vegetation indices and corresponding in-situ wheat biomass measured during the HJ-CCD data acquisition.The results indicated that DK-SVR outperformed MLR and PLS at each stage.For DK-SVR models,coefficients of determination(R2)for the estimated-versus-measured biomass values for the three stages were 0.50,0.67,and 0.65,respectively,and the corresponding root mean square errors(RMSE)were 506,1 389 and 2 058 kg·hm-2.Thus,the DK-SVR algorithm provides an effective way to improve the prediction accuracy of biomass in wheat on a large scale.