Study On Estimation Of Chlorophyll Content Of Reclaimed Vegetation Leaf In Rare Earth Mining Area Based On Hyperspectral Data

As an industrial"vitamin",As an industrial"vitamin",rare earth is a very important strategic resource in the development of our country.Among them,ion-adsorbed rare earth is medium and heavy rare earths,which occupies an important position because of its large reserves and shallow burial depth.However,the exploitation of ion-adsorbed rare earth ores will cause some damage to the ecological environment.The three mining methods that it has undergone include pool leaching,heap leaching,and in-situ leaching,all of which will lead to large-scale destruction of vegetation in the mining area,and the natural restoration of vegetation is difficult.In order to restore the environment of the mining area,after the end of rare earth mining,artificial reclamation is usually used for ecological management.However,the leaching liquid used in rare earth mining will penetrate into the soil through infiltration during the mining process,destroy the organic matter in the soil,and destroy the soil composition,which makes it difficult to grow reclaimed vegetation.How to dynamically monitor the vegetation growth process,so as to ensure the healthy growth of reclamation vegetation through timely human intervention,and then accelerate the ecological restoration of the mining area has become an extremely important task.Chlorophyll is an frequently-used biochemical parameter of vegetation,and its content can reflect the growth of vegetation.Hyperspectral technology can provide more various spectral information due to its narrow band range and large number of bands,and then explore the ecological characteristics of vegetation from the spectral dimension.Through correlation analysis of vegetation chlorophyll content and hyperspectral data,combined with various regression algorithms,a model of vegetation chlorophyll content estimation can be constructed to achieve dynamic monitoring of vegetation growth process.Therefore,in this paper,the typical reclaimed vegetation of ion adsorption rare earth mining area,such as tung tree,bamboo willow and Photinia rubra,is taken as the research object to obtain hyperspectral data and chlorophyll content in the field.Firstly,the water vapor and human impact are eliminated by denoising,smoothing and averaging.Then,derivative processing and Then derivative processing and continuum removal processing are performed on the original spectrum to further reduce the background interference and baseline drift of spectral data.In order to study the impact of the rare earth mining area's ecological environment on the spectral changes of reclaimed vegetation,the spectral differences between reclaimed vegetation and normal vegetation were compared based on the original spectrum,derivative spectrum,and continuum removal spectrum;Then,the person correlation between chlorophyll content and various spectral parameters and vegetation index extracted from original spectrum and other transform spectrum was calculated,and the spectral characteristic parameters with significant correlation with chlorophyll content were screened out,which provides the basis for the establishment of its estimation model;Finally,using the various sensitive spectral parameters screened,combined with PLS algorithm,BP neural network algorithm,SAE neural network algorithm to construct a chlorophyll content estimation model,and determine the optimal estimation model through model evaluation indicators.the results show:(1)For the original spectrum,derivative spectrum,and continuum removal spectrum,the change trend of the three vegetation spectral curves is roughly the same,but there is a difference between the reclaimed vegetation and the normal vegetation.In the visible range of the original spectrum,the spectral reflectance of the reclaimed vegetation is greater than that of normal vegetation;at the same time,the"red shift"phenomenon of the green peaks and red valleys characteristics moving toward the long wave appears.Among them,the migration degree of tung tree is 5nm,5nm respectively;bamboo willow is 4nm,5nm,Photinia rubra is 56nm,4nm.The Photinia rubra has the largest migration in the green peak position,which should be more sensitive to the environmental response of the mining area.In the derivative spectrum,the reclaimed vegetation showed a consistent blue shift of"red edge position"relatived to normal vegetation.The offset of the tung tree is 20nm,the offset of the bamboo and willow is 7nm,and the maximum offset is 22nm of Photinia rubra.The position of other parameters is not synchronized.In the continuum removal spectrum,the absorption characteristics of reclaimed vegetation and normal vegetation are more obvious.At the absorption symmetry S,the reclaimed vegetation is smaller than normal vegetation.The reason is that due to the influence of redshift,the absorption peak position of the reclaimed vegetation is closer to the right shoulder wavelength of the absorption feature.(2)According to the person correlation calculation of chlorophyll content and spectral curve,"three edges"parameters,absorption parameters and vegetation index of the whole band,there are similarities and differences among the sensitive spectral parameters selected from the three types of reclaimed vegetation.Among them,in the red edge position?r,Red Valley position?v,absorption depth H,absorption slope K,absorption area A,the correlation coefficient all reach the significant correlation;There are differences in the extraction of other spectral parameters and vegetation index.Among them,the minimum correlation parameter screened by the tung tree is Rg-Rv/Rg+Rv,the value is-0.4628;the maximum parameter is SR(713/725),the value is-0.9520;the correlation selected by the bamboo willow is the smallest The parameter is S,the value is 0.4989;the maximum parameter is the derivative spectrum 671nm,the value is 0.8047;the minimum correlation parameter selected by Photinia rubra is the original spectrum 761nm,the value is 0.4918;the maximum parameter is SR(703/1159),The value is-0.9094.(3)Based on the extraction of sensitive spectral parameters of reclaimed vegetation,PLS algorithm,BP neural network algorithm and SAE neural network algorithm were used to build the model of chlorophyll content,and the validation set was used to test the model.According to the evaluation index determination coefficient R2,root mean square error R_Mand relative error R_E,the optimal regression model of chlorophyll content of various vegetation was determined.According to the statistical results:Among the estimation models established by various vegetations,the model accuracy is SAE neural network algorithm>BP neural network algorithm>PLS algorithm.The SAE neural network algorithm has become the optimal estimation model of three reclamation vegetations due to its nonlinear ability and layer-by-layer training characteristics.In the reclaimed tung tree,the R2value is 0.9117,the R_Mvalue is 2.2370,and the R_Evalue is 7.8393%;in the reclaimed bamboo willow,the R2 value is 0.7418,the R_Mvalue is 2.1782,and the R_Evalue is 5.0118%;in the reclaimed Photinia rubra,the R2 value is 0.9815,the R_Mvalue is 1.5602,and the R_Evalue is 4.5859%.