Spectral Characteristics And Estimating Models About SPAD Values Of Maize Leaves

Maize is important food crop.It is very important to monitor the growth status of maize quickly,accurately,nondestructively and in a large area with the help of hyperspectral remote sensing technology.Chlorophyll plays a key role in the conversion of plant light energy to chemical energy(substance accumulation).Chlorophyll content can be used to evaluate the growth of maizes.In order to achieve accurate and effective hyperspectral estimation of chlorophyll content in maize leaves,taking maize leaves as the research object,this paper based on the field experiments of maize from jointing stage to ripening stage.Analyzing the space-time change rules of chlorophyll content and high spectral characteristics maize leaves,using the traditional regression methods,partial least squares(PLS)and BP neural network to construct high spectral estimate models based on the characteristic wavelengths,vegetation indexes and "trilateral" parameters,and the models were verified and evaluated in detail.The main conclusions are as follows:(1)the SPAD value and spectral characteristics of maize leaves changed with the growth and development of plants.With the passage of the growth period,the SPAD value of maize leaves presented a trend of first increasing and then decreasing,and gradually increased from jointing stage to filling stage,and then decreased after filling stage.In the visible light bands,the spectral reflectance of the leaves gradually decreased from jointing stage to filling stage,and then increased at wax ripening stage.It first increased and then decreased in the near infrared bands,and reached the peak reflectance at filling stage.From jointing stage to filling stage,the position of red edge moves towards the long-wave direction,producing "red shift",After filling stage,position of red edge moves towards the short-wave direction,producing "blue shift".The variation trends of red edge amplitude and red edge area is same as red edge location,which first increased and then decreased,decreased after filling period.(2)SPAD values and spectral characteristics of maize leaves were also different in different spatial distributions.The SPAD values of the leaves with different leaf positions at the same growth stage were different,which were manifested as the middle leaves > the bottom leaves > the upper leaves(the jointing stage was the bottom leaves > upper leaves).The spectral characteristics of the leaves at different leaf positions in the same growth period showed that the reflectance of the upper leaves in visible and near-infrared bands was the highest,and the red edge position of the upper leaves was closest to the short-wave direction and the red edge amplitude was the lowest.The spectral characteristics of different parts of the same leaf showed that the parts close to the leaf vein had the highest reflectance in visible and near-infrared bands,and the red edge parameters were the minimum.The parts near the leaf edge had the lowest reflectance in visible and near-infrared,and the red edge parameters were the maximum.(3)the SPAD value of maize leaves is closely related to the spectral characteristics of the leaves.The higher the SPAD value is,the smaller the spectral reflectance of the leaves is in the visible bands,and the larger it is in the near infrared band.The position of the red edge moves towards the direction of long wave,and the amplitude and area of the red edge are also larger.Features of hyperspectral bands were selected and extracted by using characteristic wavelengths,vegetation indexes and first-order differential spectral "tripartite" parameters.A single-variable model and a multivariate statistical analysis model for predicting SPAD values were constructed by using traditional regression models,partial least square(PLS)and BP neural network,respectively.The prediction effects of multivariate statistical analysis models were better than that of single variable models.In the multivariate statistical analysis models,SPA-PLS had the best prediction effect,whose R~2 is 0.799,RMSE is 2.753 and RPD is 1.989.