基于离散小波技术定量反演冬小麦叶片含水量的研究

Study on Quantitative Inversion of Leaf Water Content of Winter Wheat Based on Discrete Wavelet Technique

受大田环境的影响,田间采集的冬小麦冠层光谱内含有大量与目标信息无关的噪声,这制约了高光谱数据对冬小麦植株信息的估测能力。为制约噪声信息对光谱信息的影响,探寻提升光谱对冬小麦植株水分供给信息估测能力的方法,通过野外地面实验获取大田冬小麦高光谱数据及其叶片含水量信息,采用离散小波算法处理分析高光谱数据,结合相关性分析算法、偏最小二乘算珐,定量分析5类小波基对离散小波算法分离光谱信息的影响规律,离散小波算法在分离可用光谱信息与噪声中的普适规律及小波基对信息分离的影响进行探讨,从而为田间光谱数据的处理与分析提供理论与方法支撑。结果表明:(1)与冬小麦含水量敏感的波段多分布于D1-D5尺度,且敏感波段在各小波基内的分布区间相对一致,但波段位置与相关强度均存在一定差异,这表明小波基的选择能影响高频信息与冬小麦叶片含水量的相关强度与波段位置。(2)可用的光谱信息与噪声信息均随分解尺度的增加而呈先升后降的规律,噪声信息对高频信息估测能力的干扰强度随尺度的增加而降低,高频信息对冬小麦叶片含水量的估测能力随尺度的增加而降低。(3)模型的精度与稳定性是可用光谱信息与噪声信息综合作用的结果,其中基于meyer小波基的D5尺度构建的估测模型为最优模型,其建模精度的R 2=0.625、RMSE=1.562,验证精度的R 2=0.767、RMSE=1.828。本研究的结论可为基于离散小波算法的光谱处理与分析提供指导,并为受噪声影响较重的光谱信息的处理与分析提供一定参考,同时也可为我国西南、南部等全年水汽含量较高区域内或北方夏季作物叶片含水量的检测提供基础支撑。

Due to the influence of the field environment,winter wheat canopy spectra collected in the field contain a large amount of noise unrelated to the target information,which limits the ability of hyperspectral data to estimate the information of winter wheat plants.In order to limit the influence of noise information on spectral information and explore the methods to improve the estimation ability of spectral information on the water supply of winter wheat plants,this study obtained the hyperspectral data of winter wheat and its leaf water content information through field experiments,processed and analyzed the hyperspectral data by discrete wavelet algorithm,combined with correlation analysis algorithm and partial least squares algorithm,and quantitatively analyzed the influence of five types of wavelet bases on the discrete wavelet algorithm to separate The results show that:(1)the discrete wavelet algorithm can be used to separate the available spectral information from noise,and(2)the wavelet bases can be used to separate the available spectral information from noise,to provide theoretical and methodological support for the processing and analysis of the spectral data in the field.The results show that(1)the sensitive bands are mostly distributed in D1-D5 scales,and the distribution intervals of sensitive bands are relatively consistent among wavelet bases.However,there are some differences in band positions and correlation strengths,which indicates that the choice of wavelet bases can influence the correlation strengths and band positions of high-frequency information and winter wheat leaf water content.(2)The available spectral information and noise information both show a pattern of increasing and then decreasing with the increase of decomposition scale.The interference strength of noise information on the estimation ability of high-frequency information decreases with the increase of scale,and the estimation ability of high-frequency information on the water content of winter wheat leaves decreases with the increase of scale.(3)The accuracy and stability of the model are the results of the combined effect of available spectral information and noise information,in which the estimation model constructed based on the D5 scale of Meyer wavelet basis is the optimal model with R 2=0.625 and RMSE=1.562 for modeling accuracy and R 2=0.767 and RMSE=1.828 for validation accuracy.Spectral processing and analysis,and provide some reference for the processing and analysis of spectral information that is heavily influenced by noise,and also provide basic support for the detection of the water content of crop leaves within regions with high annual water vapor content,such as southwest and south China,or in the north in summer.