| Digital agriculture, one of the most significant approaches to implement the precision management and scientific farming in agriculture, is the frontier technologies, key and core technology for the efficient, ecological, safe and sustainable development in modem agriculture. The basic and key factors of digital precision agriculture are the accurate perception, rapid acquisition and intelligent control of crop-environmental imformation. Digital and informational technologies, which can provide agricultural procedure and management with fast and accurate information acquisition, scientific decision-making and high-efficiency job control, has become a global hotspot in the field of agricultural science and technology. As one of most important natural resources for human being, soil is the foundation of agricultural production. So, the techniques and methods for rapid acquring and detecting the physical-chemical information in soil is keypoint in the field of agricultural environment. Analysis of the soil types can make contribution to build up the system for evaluating soil fertily and quality, which offer a theoretical guidance for soil quality assessment, management, planning and reasonable use. Detection of elemental information in soil can provide a scientific foundation for diagnosising plant nutrition, acquireing real-time farmland information and managing fertilizer. Determination of thr heavy matel content in soil can give a guide for preventing farmland pulltion and guaranteeing high-quality and safe production in agriculture.Initially, the working process and foundation of laser-induced breakdown spectroscopy (LIBS) technology were deeply investigated. And then, the formation mechanism and internal characteristization of laser-induced plasma were expounded. Reling on that theory basic, soil was selected as the research object in this study. First of all, the influence on the soil plasma chacteristics derived from single variable parameters of LIBS system and soil were analyzed, as well as the interaction influence of multivariable parameters of LIBS system. Secondly, discriminant analysis models and methods for different soil types were established based on LIBS technique combined with chemometrics methods, and the selected emission lines and the established models were additionally verified. Thirdly, the quantitative analysis models and methods of elements (like Al, Ca, Mg, Fe, Na, K) content were built using the data from single waveband and multivariable regression models, those models obtained good results. Lastly, the quantitative analysis models of Pb and Cd content were developed based on the LIBS data and corresponding content, which offer a theorical basis for developing the soil physical-chemical information (types, internal element and content information) detector in further work. The main contents and conclusions were showed as below:(1) Discussion of the influence on the soil plasma chacteristics derived from single variable parameters of LIBS system (laser energy, repetition frequency, delay time, sampling manner, and lens to sample distance) and soil (moisture content, particle size, and porosity). After analysis, the optimized parameters were as follows:laser energy:100 mJ; repetition frequency:1 Hz; delay time was depended on element; sampling manner was the average of 20 simpling; and lens to sample distance:98 mm. Meanwhile, the paremeters of soil were investigated using the LIBS intensity of Al ? 309.27 nm, Ca 1422.67 nm, Na ? 588.99 nm, and K I 766.49 nm and the acquirements of soil sample were enumerated as below:moisture content as little as soil contained; the particle size should be less than 0.15 nm; the pressure could kepped in 10 MPa.(2) Based on single variable test, the experiment of three factors (LE, DT, and LTSD) quadiratic regression rotation orthogonal combination was designed using the respond surface method (RSM). According to the emission lines at Si ? 390.55 nm, Al ? 394.40 nm, Fe ? 404.58 nm, Mg ? 518.36 nm, Na ? 588.99 nm, Ca ? 393.37 nm, and K ? 766.49 nm, the comprehensive signal-background-ratio (SBR) were named as the objective function (YSBR.-The interaction influences among three factors on soil plasma chacteristics were explored and the optimized parameters of LIBS were summarized. Results revealed as follows:the factor LE showed a remarkable linear effect to YSBR, and factors of DT and LTSD exhibited an opposite results. The interaction of three factors displayed a non-significant relationship. Meawhile, the factors of LE2, DT2 and LTSD2 had a significant surface relationship. Through the RSM analysis, the optimalized experimental parameters were:LE:103.09 mJ; DT:2.92 us; LTSD:97.69 mm; and a peak value YSBR of 198.602 could be obtained.(3) Establishment of discriminant analysis models and methods for soil types based on LIBS technique coupled with chemometrics methods, and verification of reliability of the selected emission lines and the established models. Based on the results of principal component analysis (PCA) and contents standard soil samples, seven characteristic spectral lines at Si I 390.55 nm, Al ? 394.40 nm, Fe ? 422.74 nm, Mg ? 518.36 nm, Na ? 588.96 nm, Ca? 393.36 nm and K I 766.49 nm (I represented atomic spectral line, II meant ionic spectral lines) were identified. Based on LIBS spectra at 7 characteristic spectral lines, PCA was carried out and an obvious cluster was observed from the score plot of the first 2 principal components (PCs). Meanwhile, partial least squares discriminant analysis (PLS-DA), soft independent modeling of class analogy (SIMCA), and least-squares support vector machine (LS-SVM) models were introduced to establish the discriminant models and the correct rates of discrimination were 98%,90%, and 100%, respectively. Then, the performances of three models were evaluated using receiver operating characteristic (ROC) curve. The results illustrated that the LS-SVM discriminant model was robust. Based on this, eight types of soils from different places were used to conduct the same experiments to to verify the selected seven characteristic spectral lines and discriminant model. Results of PCA displayed an apparent cluster and the LS-SVM model also offered a prediction accuracy of 100%. ROC curve also exhibited an excellent result.(4) Simultaneous quantitative analysis of multi-elements content in soil was conducted by using LIBS technique coupled with calibration curve and chemometrics method. First, the LIBS spectra were normalized eliminated and averaged to reduce the error in experiment process. Then, calibration curves based on the normalized intensity, integrated intensity (peak areas), Si element inernal standard method (intensity ratio) and corresponding contents of elements were fitted. The results indicated that the linear relationship from the calibration curves fitted by normalized intensity and peak areas showed a good linear fitting (except the Fe). Meanwhile, the Si element inernal standard method offered the better results than the methods of normalized intensity and peak areas. In addition, calibration free-LIBS (CF-LIBS) nethod was also employed to compute the content of Al, Ca, Si, Fe, Mg, Na and K, repectively. Lastly, partial least-squares regression (PLSR) was employed to build the quantitative model for predicting the content of Al, Ca, Si, Fe, Mg, Na and K. PLSR methods provided promising results with relatively high correlation coefficient of prediction Rp and showed more advantages than the calibration curve method. The approach revealed that LIBS technology combined with chemometrics methods displayed a bight prospect in the field of spectrochemical analysis.(5) LIBS technology combined with calibration curve and chemometrics methods were employed to quantitative analyze the content of heavy metal Pb and Cd. The emission lines at Pb I 405.78 nm and Cd ? 361.05 nm were identified as characteristic lines, and the models based on LIBS intensity, lorentz fit intensity after normalized, peak areas and corresponding content were established. For Pb element, the linear relationships of coefficient of determination (R2) of three methods were:0.98385,0.97097 and 0.99321. And Cd element failed to provide an effective linear relationship. Meanshile, PLSR model for predicting Pb and Cd were built. Results dsemonstrated that the calibration curve and PLSR model provided similar performance, resulting in correlation coefficient of 0.9485 and root mean square error of 2.044 mg·g-1 in prediction set. And PLSR model for Cd elemrnt prediction revealed the promising results, resulting in correlation coefficient of 0.9949 and root mean square error of 97.05 ?g·g-1 in prediction set. |
PDF Full Text Request