Application Of Continuous Wavelet Analysis To Reflectance Spectra For The Spectroscopic Estimation Of Wheat Growth Indicators

With the increase of population and the decrease of cultivated land area in China,the problem of food security becomes increasingly important.The grain yield and quality are closely related to the crop growth status over the growing season.The accurate acquisition of the crop growth indicators is important for the prediction of grain yield and quality.Remote sensing technique provides a non-destructive way to accurately monitor crop growth indicators.The development of hyperspectral remote sensing provides useful data for monitoring the growth indicators of crops.By applying continuous wavelet analysis to wheat reflectance spectra at leaf and canopy levels,this thesis carries out the investigation on multi-scale reflectance responses of wheat growth indicators,the construction of wavelet-based empirical models,and the retrieval of wheat growth indicator from a radiation transfer model.The main content of the thesis lies in two aspects.First,a continuous wavelet transform was applied to each reflectance spectrum to generate a wavelet power scalogram composed of a number of wavelet coefficients or wavelet features as a function of scale and wavelength.Then the wavelet features sensitive to wheat growth indicators including biochemical parameters and leaf area index(LAI)were obtained from the wavelet power scalogram so as to generate correlation scalograms for each parameter.While it is feasible to delineate the sensitive wavelet features for each biochemical parameter at leaf level,it is difficult to do so at canopy level due to the covariance of biochemical parameters with LAI.The non-overlapping wavelet features that were sensitive to biochemical parameters were extracted through mathematical operations of correlation scalograms.The distribution of sensitive wavelet features for leaf-and canopy-level parameters revealed that the low-scale wavelet features captured the local absorption of biochemical parameters and the high-scale wavelet features reflected the influence of LAI on reflectance amplitude.Secondly,a wavelet-based model inversion method was proposed by integrating the continuous wavelet transform of leaf reflectance spectra with a radiative transfer model called PROSPECT.To reduce the ill-posed problem in model inversion,this method used the wavelet coefficient at multiple scales to construct constraint functions for model inversion and determined the optimal scale of inversion for each parameter.Compared to traditional methods(reflectance-based,reflectance-based stepwise and vegetation index-based methods),the wavelet-based method significantly improved the accuracy of model inversion for all leaf biochemical parameters.The comprehensive investigations on the applications of the continuous wavelet analysis in the empirical estimation and the physical inversion methods offer a new methodology for monitoring of the wheat growth indicators with hyperspecral data.The significance of this research lies in two aspects:(1)determination of sensitive wavelet features and empirical estimation models with the removal of covariance between biophysical parameters and LAI;(2)proposal of a physical inversion method with the integration of continuous wavelet analysis of reflectance spectra and a radiative transfer model.The methods in this research have great potential in the spatial mapping of crop growth indicators with airborne and spacebome hyperspectral data,and provide the reliable technical support for modern agriculture.