基于敏感光谱波段图像特征的冬小麦LAI和地上部生物量监测

叶面积指数(LAI,leaf area index)和地上部生物量是评价冬小麦长势的重要农学参数,其实时动态监测对冬小麦的长势诊断、产量预测和管理调控等具有重要意义。该研究通过分析叶面积指数、地上部生物量与冬小麦冠层光谱参数的相关性,筛选出冬小麦长势指标敏感波段及最佳带宽范围;基于敏感光谱波段下图像的彩色因子,构建冬小麦叶面积指数和地上部生物量监测模型。结果表明,叶面积指数、地上部生物量长势指标的敏感波段及最佳带宽范围为(560±6)和(810±10)nm。敏感波段560、810 nm波段下获得的图像特征因子中,RGB颜色空间R810、G560、B810对叶面积指数的拟合效果最好,决定系数高达0.989;HSI颜色空间H810、S810、I560对地上部生物量的拟合效果最好,决定系数为0.937。试验数据检验表明,叶面积指数、地上部生物量监测模型的均方根误差RMSE分别为0.4515、3.3556,相对误差分别为15.7%、15.9%,所构建监测模型的精确度较高。因此,基于敏感光谱波段及相应图像特征构建的监测模型可有效对冬小麦叶面积指数、地上部生物量进行实时、快速、准确监测与诊断。

Monitoring leaf area index and biomass above ground of winter wheat based on sensitive spectral waveband and corresponding image characteristic

Abstract: Leaf area index and above ground biomass are important parameters for evaluating winter wheat growth status Monitoring of them in real-time is great tool to diagnose growth, yield prediction, field management and regulation. Through the correlation analysis of leaf area index, above ground biomass with canopy spectral parameters, in this study, we screened the sensitive spectral waveband to growth index of winter wheat and the optimal bandwidth range. Based on the image characteristics extracted from the image of sensitive waveband, monitoring models of winter wheat leaf area index and above ground biomass were established. The results showed that when the wavelength was smaller than 700 nm, leaf area index and above ground biomass were negatively correlated with the canopy reflectance. Meanwhile, an obvious trough appeared in the 560 nm waveband or so. Between 800 nm waveband and 1040 nm waveband, a high stable platform appeared. Therefore, the sensitive wavebands and optimal bandwidth ranges of leaf area index and above ground biomass were (560±6) nm waveband and (810±10) nm waveband. The correlation analysis results of leaf area index, above ground biomass with single image parameters (R, G, B, L, H, S, I) showed that the correlations between image parameters G, L, I of 560 nm waveband with leaf area index and above ground biomass were poor and all of R2 were less than 0.5. Furthermore, even though coefficients of determination between image parameters H, S with leaf area index were both higher than 0.85, the correlations between image parameters R, G, B, L, I of 810 nm waveband with leaf area index and above ground biomass were also poor. Therefore, except for image parameters H and S, other image parameters were not quite fit to leaf area index and above ground biomass. Only through a single image parameter of 560 nm waveband and 810 nm waveband, we can not build satisfying monitoring models for leaf area index and above ground biomass. Then, in this study, we built leaf area index and above ground biomass monitoring models in different color space of RGB and HIS. It showed that the monitoring model of leaf area index built in RGB color space was better than that built in HSI color space and it was opposite for above ground biomass. Among the image parameters obtained in 560 nm waveband and 810 nm waveband, R810, G560 and B810 of RGB color space were the best fitting to leaf area index and coefficient of determination was as high as 0.989. G810, S810 and I560 of HSI color space were the best fitting to above ground biomass and coefficient of determination was 0.937. After the verification of experimental data from the same year at different experimental field, root mean square errors of leaf area index and above ground biomass monitoring models were 0.4515 and 3.3556, and relative errors were 15.7% and 15.9%. So, the accuracy of monitoring models was high. Therefore, based on sensitive spectral waveband and corresponding image characteristic, monitoring models established can monitor and diagnose leaf area index and above ground biomass of winter wheat in real-time quickly and accurately.