Study On Extraction And Inversion Models Of Weak Information In Maize Leaf Polluted Under Copper Stress

The gangue piles,tailings and waste water generated by high-intensity and continuous mining have brought serious challenges to the ecological environment and agricultural security in mining area and its surroundings.The heavy metal pollution is one of important environmental problems.In China,maize is widely planted,especially in mining area,and it is also an important crop.Hyperspectral remote sensing can provide continuous and fine spectra.It has the advantages of fine recognition and non-destructive detection,and can distinguish the diagnostic spectral characteristics of ground objects.In order to timely and effectively use hyperspectral remote sensing to monitor copper(Cu)pollution in maize leaves under heavy metal stress,eleven stress levels were set in control experiment,in which CuSO4·5H2O crystals was used as pollution source.From the perspective of spectral dimension and frequency domain,the extraction and inversion models of the weak information of copper pollution in maize leaves were studied in this paper.The paper mainly included following research contents and results.1.Based on PROSPECT-D,the changes of chlorophyll,mesophyll structure and water content in leaves were analyzed,and the mechanism of spectral response of maize leaves were also discussed.The study concluded that:(I)The spectral response of maize leaves to chlorophyll mainly located in the visible and near infrared bands of 400-800 nm,and the reflectance of 500-650 nm was greatly affected.Under copper stress,the thickness of vascular wall of maize leaves tended to decrease,and leaf internal structure became more disordered.The reflectance from near infrared to short infrared band also increased.Under copper stress,the water balance in maize leaves was also broken,and spectral response of maize leaves to the changes of leaf water was in the range of 900-2500 nm.Due to frequency doubling or combination of water molecular vibration near 970 nm,1200 nm,1450 nm and 1950 nm,water absorption valleys were formed near 970 nm,1200 nm,1450 nm and 1950 nm.(2)The range of 400-900 nm was the important spectral bands for weak information analysis of copper pollution in maize leaves.2.Nine vegetation indices,and six spectral characteristic parameters were selected to analyze the weak information of copper pollution in maize leaves at jointing stage.The index ESDD(Entropy of spectra differential difference,ESDD)was also constructed to analyze and extract weak information of copper pollution.The study concluded that:(1)The correlation coefficients between the vegetation indices of mNDVI705,SR705,NDVI705,MCARI,NPCI,RVSI and copper contents in new and middle leaves were greater than 0.5,among which the vegetation indices of mNDVI705,SR705,NDVI705 and MCARI were greater than 0.7.The correlation coe:fficients between mNDV705,SR705,NDVI705,MCARI and copper contents in maize old leaves were all greater than 0.5.The inversion models of exponential function based on mNDVI705 for copper contents in new,middle and old leaves had the highest accuracy,and R2 was 0.716,0.418 and 0.695,respectively.(2)The correlation coefficients between REP and copper contents in new and middle leaves were all greater than 0.6.It was showed that REP was an important spectral characteristic parameter for copper stress discrimination.For new leaves,the quadratic polynomial model constructed by REP had the highest accuracy,and R2 was 0.498.(3)ESDD(480-670)and ESDD(670-750)could better describe the weak information of copper pollution contained in the spectral differential difference of maize leaves under different stress levels.The exponential function based on ESDD(480-670)and quadratic polynomial function model constructed by ESDD(670-750)were effective in predicting copper contents in maize new leaves.3.The theory of harmonic analysis was introduced into the analysis of copper stress degree on maize plant,and the theoretical basis of applying harmonic analysis to discrimination of copper stress degree on maize plant was introduced.At 480-670nm and 670-750nm,the reflectance spectra simulated under the conditions of normal growth,low concentration stress and high concentration stress were constructed respectively.The spectra simulated were used to verify that the amplitudes of harmonic subsignal were effective in detecting and analyzing the spectral weak distortion in leaf reflectance.The harmonic analysis was used to analyze copper pollution stress based on old,middle and new leaves reflectance at three growth stages:seedling,jointing and seedling stage.(1)The seedling stage and jointing stage was the best growth stages for monitoring copper pollution stress on maize plants by harmonic amplitude characteristics.(2)At seeding stage,the harmonic amplitudes C1 and C2 of 480-670nm of new and middle leaves reflectance could effectively identify and distinguish the degree of copper pollution stress on corn plants.From Cu(100)to Cu(1200),the first harmonic amplitudes C1 of old leaves reflectance could identify and distinguish the degree of copper stress on maize plants.At seeding stage,from Cu(100)to Cu(1200),the harmonic amplitude C1 of 670-750nm of old and middle leaves reflectance was effective in identifying the degree of copper stress.(3)At heading stage,from Cu(50)to Cu(1200),in addition to Cu(1000),the harmonic subsignal amplitude C1 of 480-670nm could better distinguish the degree of copper pollution stress on maize plant.4.The theory of Hilbert marginal spectrum was used to analyze the old maize leaves reflectance of 400-900 nm.It was showed that:(1)The frequency of Hilbert marginal spectrum of 400-900 nm of old leaves reflectance under eleven stress levels mainly located in the range of 0-30 Hz,and there were also some differences.The Hilbert marginal spectrum of 30-100Hz had a little change,but it also carried a certain amount of copper pollution information in maize leaves.(2)The Hilbert marginal spectrum of the 400-900 nm at 0-100 Hz could well characterize the weak information of copper pollution in maize leaves.(3)Four characteristic parameters of the Marginal spectrum Surrounding Area(MSA),Marginal Spectrum Energy(MSE),Marginal Spectrum Mean(MSM)and Marginal Spectrum Amplitude Maximum(MSAM)were defined.Four marginal spectrum characteristic parameter were positively correlated with copper contents in old leaves.The correlation coefficient between MSM and copper contents was greater than 0.8,the analytical ability of MSE and MSM were stronger than MSA,MSAM.(4)The prediction models for copper contents in old leaves based on MSA,MSE,MSM and MSAM were constructed,among which the linear model with one variable of MSE had the highest prediction accuracy with R2 of 0.557 and RMSE of 3.619 ?g/g.5.In order to reduce the uncertainty and instability in the diagnosis of copper pollution caused by single type of parameters in spectral dimension or frequency domain,the innovative collaborative framework and inversion models for copper contents in maize leaves based on characteristic characters of spectral dimension and frequency domain were proposed.Five vegetation indices,ESDD(480-670),ESDD(670-750),the first harmonic amplitude C1 and four characteristic parameters of Hilbert marginal spectrum were selected to establish the prediction models.Six modes of parameters combination were set,and six inversion models for copper contents in maize leaves were constructed based on independent component Analysis and radial basis function neural network.The prediction accuracy from high to low was as follow:plan5,2,6,4,1 and 3.When the spectral dimension was coordinated with the frequency domain,the deficiency of describing the copper pollution information by single type of characteristic parameters could be reduced to a certain extent,which could improve the inversion accuracy of copper contents in maize leaves.