| Leaf chlorophyll content is an important index for evaluating crop photosynthesis and guiding field management.Its nondestructive monitoring is of great significance for identifying crop health status.At present,many empirical models based on hyperspectral index and wavelet function have been developed and used to estimate chlorophyll content in leaves of many crops.However,so far,most studies have focused on the extraction of sensitive bands,and the importance of hyperspectral index algorithm,wavelet function category and band optimization for improving the estimation accuracy has not been determined.In order to further evaluate these methods to estimate the corn leaf chlorophyll content of anti-interference ability and sensitivity,to build a stronger universality and stability estimating model of corn leaf chlorophyll content.In this study,four spectral indices algorithms including dispersed band spectral indices,red edge position algorithm,shape area spectral indices and wavelet transform are selected.In 2019 to 2021,field trials with three different nitrogen gradients were conducted in three typical regions of maize cultivation in Inner Mongolia to obtain leaf spectral reflectance and chlorophyll content at four critical growth stages and three layers of maize.Firstly,the effects of growth stages and leaf layers on chlorophyll content and hyperspectral characteristics of leaves were analyzed,the influence of the height,it was found that the average value of chlorophyll content in maize layers increased firstly and then decreased with the development of the growth stages,and the middle leaf content was the highest.The overall trend of spectral reflectance in maize leaves at the growth period and under the layers was consistent,and the change of chlorophyll content in maize leaves directly affected the area of spectral reflectance at 400-800 nm.Secondly,the correlation between the spectral indices,red edge position,shape area indices and leaf chlorophyll content were calculated.It was found that the optimized hyperspectral index significantly improved the estimation ability,and the optimized red edge normalized difference index(OPT-m ND705)showed the best performance(R~2=0.96)in the spectral index of the dispersed band.Among the six red edge position algorithms,linear extrapolation(REP-LE)had the highest estimation accuracy(R~2=0.95).Optimized bimodal area index(ONDDA)had the best estimation effect(R~2=0.96).Compared with the published hyperspectral indices,the estimation ability of hyperspectral indices is improved by 6%-76%by band optimization,but the index algorithm has certain influence on the stability of hyperspectral indices and the location of sensitive bands.Then,the spectral reflectance of leaves was analyzed by wavelet transform based on continuous and discrete wavelet functions and the wavelet feature spectrum was constructed to screen the feature bands.It was found that the Mexh wavelet function of continuous wavelet transform had the best performance in estimating the leaf chlorophyll content(R~2=0.96).Finally,two independent experimental datasets and PROSPECT-5B model simulation datasets were used to evaluate and validate the estimation models of chlorophyll content in maize leaves based on OPT-m ND705,REP-LE,ONDDA and Mexh,and it was found that ONDDA index had the best validation performance.The R2 and RE%between predicted and measured values were 0.92～0.96and 5.84%～18.59%,respectively.The estimation performance of hyperspectral indices and the location of sensitive band are greatly affected by the algorithm.The combined evaluation of two independent datasets and PROSPECT-5B model showed that ONDDA had the best predictive performance,with the sensitive bands of ONDDA located at 670,720,and 750 nm.These findings provide a choice of hyperspectral index for estimating maize leaf chlorophyll content,and provide an important theoretical basis for diagnosing maize leaf chlorophyll content in a larger area and accurately grasping the nutritional status of crops in the future. |
PDF Full Text Request