欢迎访问《麦类作物学报》编辑部网站

为探索基于全波段冠层高光谱以及变换光谱的冬小麦地上部生物量的遥感估算方法，以2016、2017年冬小麦田间试验为基础，通过对冠层光谱和地上部生物量的相关性分析，筛选拔节期、抽穗期的冬小麦冠层光谱、一阶导数光谱、对数变换光谱和连续统去除光谱对地上部生物量的敏感波段，并结合偏最小二乘法(PLS)分别建立拔节期和抽穗期基于SPA算法的冬小麦地上部生物量估测模型，再与基于任意两波段组合的最佳归一化光谱指数、比值光谱指数、差值光谱指数和已报道光谱指数的冬小麦地上部生物量估测模型进行比较。结果表明：(1)SPA算法较好地利用了全波段冠层光谱信息，并显著降低了光谱维度，不同变换光谱的地上部生物量敏感波段个数在4~14之间；(2)拔节期和抽穗期冠层光谱与地上部生物量的相关性高于开花期和灌浆期，各生育时期一阶导数光谱与地上部生物量之间的相关性优于连续统去除光谱、对数变换光谱和光谱指数；(3) 利用抽穗期一阶导数光谱敏感波段建立的预测模型和验证模型达到了较高的精度，其预测模型的决定系数和均方根误差分别为0.78和0.87 t·hm^-2，验证模型的决定系数和均方根误差分别为 0.84和0.69 t·hm^-2，预测相对偏差为2.74。这说明，抽穗期是估算地上部生物量的最佳生育时期，且基于冠层一阶导数变换光谱，结合连续投影算法和偏最小二乘回归方法所构建抽穗期地上部生物量估算模型具有最优的精度和预测能力，可用于地上部生物量的定量估算。

英文摘要:

In order to explore the best way for estimating shoot biomass of winter wheat using canopy hyper-spectra and transformation spectra, and provide the scientific shoot biomass data for growth detection and yield prediction, the experiments were carried out based on the winter wheat field trail in 2016 and 2017. With the results of correlation analysis between canopy spectra and shoot biomass, the sensitive bands of the original canopy spectrum, first derivative spectra, logarithmic transformation spectra and continuum removal spectra were successfully extracted. Based on successive projection algorithm(SPA) at jointing and heading stages, the partial least square(PLS) estimation models of winter wheat shoot biomass were established. The PLS models were compared with the estimation models based on any two-band combinations of optimal normalized spectral index, ratio spectral index, difference spectral index and reported spectral index. The results showed that:SPA algorithm made good use of the whole band canopy spectral information, and significantly reduced the spectral dimension. The number of sensitive bands to shoot biomass between different transformation spectra ranged from 4 to 14. The correlation between canopy spectra and shoot biomass at jointing stage and heading stage was higher than that at flowering stage and grain filling stage. The correlation between first derivative spectra and shoot biomass at each growth stage was significantly better than that of continuum removal spectra, logarithmic transformation spectra and spectral index. The prediction model and validation model established with the first derivative spectra of the spectral sensitive band at heading stage obtained the highest estimation accuracy. The determination coefficient and root-mean-square error of the prediction model were 0.78 and 0.87 t·hm^-2, respectively, and the determination coefficient and root-mean-square error of the validation model were 0.84 and 0.69 t·hm^-2, respectively. The predicted relative deviation was 2.74. It is confirmed that heading stage was the best growth stage for estimating shoot biomass, and the estimation model with the first derivative spectra of the spectral sensitive band at heading stage using SPA and PLS methods performed a high prediction capacity of the model to samples and could be used for quantitative estimation of shoot biomass.

查看全文查看/发表评论下载PDF阅读器

关闭