| Rice is an important food crop in China,whose planting area accounts for about one-sixth of the total rice area in the world.In the process of rice planting,heavy metal pollution will cause the decline of rice yield and quality result from seriously poisoning rice tissues,stimulating physiological response to resist heavy metal stress,as well as affecting the normal growth of plants.In addition,heavy metals transfer and accumulate from the ecological enviro nment to the root-stem-leaf-grain tissue of rice,then pass through the food chain to the human body,which affects the normal immune and metabolic system,causing food security problems at last.It is not only conducive to timely and effective monitoring of rice growth stress effect and heavy metal pollution control,but also beneficial to the precise management and regulation of green ecological production of rice to rapidly track and detect physiological information changes besides heavy metal accumulation in rice leaves under heavy metal stress,ensuring food security.This study focuses on rapidly detecting the physiological information and accumulation of heavy metals in rice leaves polluted by Cadmium(Cd)and Copper(Cu)based on hyperspectral imaging and laser-induced breakdown spectroscopy,establishing rapid detection methods and visualization techniques about physiological information in the adversity such as ascorbic acid(ASA),glutathione(GSH)and free proline(FP)in rice leaves based on hyperspectral;researching fast quantitative method of heavy metals(Cd,Cu)in rice leaves based on different laser wavelengths,ambient atmosphere and pyrolysis treatment;constructing detection system of heavy metal element characteristic index coefficient;realizing rapid detection of stress physiology and heavy metal accumulation in rice leaves under early stress of heavy metals.The main conclusions are as follows:(1)Rapid detection methods and visual distribution models of physiological information of rice leaves under heavy metal stress based on hyperspectral imaging were established.The changes of ASA,GSH and FP content in Rice Leaves under different heavy metal stress(Cd,Cu),different stress time(5 d,10 d,15 d and 20 d),and different stress concentration(0μM,5μM,25μM,50μM and 100μM)were detected by hyperspectral imaging.It was found that the correlation between ASA and GSH contents increased with the increase of stress time,and the spectral intensity of leaves showed significant difference.The partial least squares(PLS),least squares support vector machine(LS-SVM)and limit learning machine(ELM)models based on full spectrum and characteristic variables were established to realize the rapid quantitative detection of ASA,FP and GSH.The R_p of full spectrum model under Cd stress was 0.9623,0.9566 and0.9190,respectively.The R_p of full spectrum model under Cu stress was 0.8965,0.9501 and0.9082 respectively.Genetic algorithm(GA),competitive adaptive weighted sampling(CARS)and PLS weighted regression coefficient(Bw)were used to select characteristic variables.The R_pwith best model of ASA,FP and GSH under Cd stress was 0.9684,0.9624,0.9426,respectively,and the R_p under Cu stress was 0.9235,0.9576 and 0.9104 respectively.The best detection models were used to realize visual distribution of ASA,FP and GSH on the surface of rice leaves under Cd stress and Cu stress.(2)The detection methods on parameter optimization under three different wavelengths for detecting Cd and Cu content in rice leaves based on LIBS were established.The Cd and Cu signal strength,stability(relative standard deviation)and sensitivity(signal-to-back ratio and signal-to-noise ratio)change rules and influence rules towards the temperature and density of plasma at different delay times and laser energy using three laser wavelengths(1064,532 and 266 nm)were systematically compared.Above researches determined the optimal system parameters of LIBS,establishing univariate and multivariable models for rapid quantitative detection of heavy metals in rice leaves.The results showed that the optimal delay time of the three wavelengths(1064,532and 266 nm)for detecting Cd was 2μs,with the energy 80,80 and 21 m J respectively.The PLS model with highest prediction accuracy was when the wavelength is 532 nm,and the R_p is 0.9883.The optimal delay times under the three wavelengths(1064,532,and 266 nm)for detecting Cu were 3μs,and the energy was 80,80,and 21 m J,respectively.When the wavelength was 266 nm,the prediction accuracy of the PLS model was the highest,and the R_p was 0.9813.(3)A rapid quantitative detection method for Cd and Cu content in rice leaves based on enhanced argon enviro nment regulation was proposed.An argon enviro nment control device was developed,and the variation rule of Cd and Cu signals changing with delay time and laser energy under argon and air enviro nment was proved.It was found that the argon enviro nment could significantly increase the signal strength,signal-to-back ratio and signal-to-noise ratio of Cd and Cu,and the signal was more than 3 times stronger than that of air.The optimal delay time for Cd and Cu detection under argon enviro nment was 3μs,the optimal integration time was 9and 16μs,and the optimal laser energy was 80 and 21 m J,respectively.The quantitative detection models of Cd and Cu in rice leaves under argon atmosphere were established.The best detection effect for univariate analysis of Cd was obtained by Cd I 228.80 nm with R_p of 0.9809,and the best multivariate analysis was obtained by LS-SVM model with R_p of 0.9913.The best detection effect for univariate analysis of Cu was obtained by Cu I 327.39 with R_p of 0.9485,and the best multivariate analysis was obtained by PLS model with R_p of 0.9655.(4)A rapid quantitative detection method for Cd and Cu in rice leaves based on pyrolysis treatment was proposed.The effect of different pyrolysis heating rate and the final temperature on the signal strength,stability and sensitivity of Cd and Cu was studied.It was found that pyrolysis treatment can significantly reduce RSD of rice leaf detection,and improve stability,signal-to-background ratio and signal noise ratio.The optimal heating rate of Cd and Cu is10℃/min,and the final temperature is 300℃.A mass loss rate conversion model and the quantitative detection models for heavy metals content were established.For univariate analysis,Cd I 228.80 nm obtained the best prediction with R_p of 0.9801;for multivariate analysis of Cd,LS-SVM model obtained the best prediction with R_p of 0.9933.For univariate analysis,Cu I327.39 obtained the best prediction with R_p of 0.9598.For multivariate analysis of Cu,PLS model obtained the best prediction with R_p of 0.9853.(5)LIBS characteristic indexes system for Cd and Cu detection in rice leaves was established.First,quantitative detection models based on LIBS full spectrum were analyzed.The best model for Cd quantitation was LS-SVM model based on raw spectrum of 211.48-232.46 nm and the best R_p was 0.8949.The best model for Cu quantitation was PLS model based on raw spectrum of 319.48-339.95 nm and the best R_p was 0.9557.Then interval partial least squares(i PLS),backward interval partial least squares(bi PLS),GA,and Bw algorithms were used to filter the characteristic variables,and rapid quantitative detection models based on the characteristic variables were established.The results showed that for the detection of Cd in rice leaves,the 85 variables selected by i PLS of the monochromator spectrum obtained the best R_p of0.9425.For the detection of Cu in rice leaves,the 1835 variables selected by bi PLS of the Echelle spectrograph spectrum obtained the best R_p of 0.9383.In this study,the characteristic indexes models for rapid quantitative detection of Cd and Cu were constructed for the first time.The highest R_p for Cd detection was 0.9259,obtained by index Cd1 based on i PLS-GA.The highest R_p for Cu detection was 0.9540,obtained by index Cu3 based on i PLS-Bw.In this study,through full-spectrum analysis,feature variables selection,sensitive spectral lines positioning,and element feature indexes establishment,a rapid detection method system for heavy metals quantitation using LIBS was constructed,which provided the technical support for the rapid quantitative detection,prevention and repair of heavy metal pollution in crops. |
PDF Full Text Request