Rapid And Non-destructive Analysis For Variety Identification And Purity Of Multi-grain Rice Seeds With NIR Spectroscopy

High-yield and high-quality rice seeds are important to solve the problem of food for global populations and the increasing demand for high-quality food.The authenticity and purity of rice seeds affected the breeding,production and circulation of rice.The existed seed authenticity identification and purity analysis methods are based on artificial field planting or molecular instrumental analysis of single seeds.These methods are complexes,time-consuming and inefficient,and cannot meet the needs of large-scale agricultural production.As a fast and non-destructive analytical technique,near-infrared spectroscopy has used for discriminant analysis of single seeds.At present,there is no report on the purity analysis of multi-grain rice seeds,and the identification analysis of multi-grain rice seeds are very rarely.In this paper,near-infrared diffuse reflectance spectroscopy combined with chemometrics as a new method was used to study the authenticity identification and purity analysis of multi-grain rice seeds.The rice variety Y Liang you 900(abbreviated as R1,similar below)was used as the target variety(negative);Four rice varieties,Xiang liang you 900(R2),Nei 5 You 8015(R3),Jing liang you hua zhan(R4),Huang hua zhan(R5),were used as contaminated varieties(positive).Two sets of negative and positive rice seed samples were designed by artificial hybrid method,and the near-infrared spectroscopy experiments were carried out to identify the authenticity(qualification)and purity analysis(quantitative)of multiple rice seeds.The main results are as follows:1.Methodological studies:Standard normal variate(SNV)is used for pretreatment of diffuse reflectance spectra of multi-grain rice seeds;Equidistant combination(EC)method is used for a wide range of characteristic wavelength screening;Wavelength step-by-step phase-out(WSP)method is further proposed to eliminate the interference wavelength of the equally spaced combination model and improve the spectral prediction ability.Based on the partial least squares(PLS)and partial least squares discriminant analysis(PLS-DA)method,the integrated optimization methods such as EC-PLS,WSP-PLS,EC-PLS-DA and WSP-PLS-DA are constructed.These methods will be used for multi-grain rice seeds purity analysis(quantitative)and authenticity analysis(qualitative).2.Authenticity identification of multi-grain rice(qualitative):(1)Sample design for discriminant analysis:Take R1 as the target variety and Prepare 320 samples of negative samples;take R2,R3,R4,R5 as contaminated varieties,and add to R1 according to different concentrations,and prepare 220positive samples(pure positive,Mixed with 4 categories).They were divided into calibration(negative 116,positive 76,total 192),prediction(negative 102,positive 72,total 174),and validation(negative 102,positive 72,total 174)sample set.(2)The global optimal model obtained by EC-PLS-DA method is(located in the short-NIR)I=808nm,N=84,G=1,LV=10;in addition,the optimal model in the long-wave near-infrared region is I=1188 nm,N=78,G=3,and LV=7.Based on the EC-PLS-DA model,the WSP method is used to further optimize the model.The optimal WSP-PLS-DA model is N=54,LV=8(short-NIR)and N=49,LV=7(long-NIR).Both their modeling effects and the complexity of the wavelength model have been significantly improved.The Samples not involved in modeling are used for The Samples not involved in modeling are used for validation.The accuracy of positive recognition,negative recognition and total recognition of the two models reached93.1%,95.1%,and 94.3%(short-NIR model);90.3%,94.1%,and 92.5%(long-NIR model).3.Purity analysis(quantitative)of multi-grain rice seeds:(1)Sample design for quantitative analysis:target species with R1 as purity analysis,R2,R3,R4,and R5 were used as contaminated varieties.And were added to R1 at different concentrations to prepare 164 samples and divided into calibration(60),prediction(52),and inspection(52)sample sets.(2)The global optimal model obtained by the EC-PLS method is(located in the long-wave near-infrared region)I=1164,N=74,LV=9;in addition,the optimal EC-PLS model obtained in the visible-short-wave near-infrared region I=418 nm,N=69,G=4,and LV=8.Based on the EC-PLS model,the WSP method is used to further optimize the model.The optimal WSP-PLS model is N=24,LV=8(long-NIR)and N=19,LV=6(Visible-short NIR).Both their modeling effects and the complexity of the wavelength model have been significantly improved.The Samples not involved in modeling are used for validation.The SEP and R_P between the predicted and measured values of the two models were 0.129,0.894(long-NIR model)and 0.152,0.845(visible-short NIR model).The results showed that it was feasible to identify the authenticity and analyze the purity of multiple rice seeds by near-infrared diffuse reflectance spectroscopy combined with stoichiometry.The proposed wavelength phasing method can effectively eliminate the interference wavelength can improve the spectral prediction effect,and reduce the complexity of the wavelength model.The obtained wavelength model provided a reference for the design of seed-specific near-infrared spectrometers.The authenticity identification and purity analysis of multi-grain rice seeds method can improve work efficiency and also have a great significance for large-scale agricultural production.