Study On The Degree Of Aging Of Rice Seeds Based On Near-infrared Spectroscopy

Aging seed vigor declined naturally,affecting the whole process of budding rate and plant growth,and then affected the yield of crops.As the main grain crop in China,rice seed aging test is worthy of further study.At present,most of the studies on the method of rapid detection of rice seed aging are mainly chemical experiments.The experimental process is complex and the experimental period is long,which is not conducive to rapid detection.In this paper,the identification of different aging degrees of rice seeds was carried out on the basis of near infrared spectroscopy.The aim was to distinguish rice seeds with different aging degrees through a rapid and accurate method.This study takes the cultivation rice No.9 planted by the Sanjiang administration of Heilongjiang Reclamation Bureau as the object of study.The seeds with different aging degree were obtained by high temperature and humidity method.Then the spectra were collected on the aging degree of rice seeds is,in the data smoothing and correction based on pretreatment,using principal component analysis(PCA),partial least squares(PLS)prediction method of data analysis and modeling.The experimental results show that the method proposed in this paper can effectively distinguish rice seeds with different aging time on 0 days(d),4 days(d)and 8 days(D).The main contents of this paper are as follows:(1)By means of high temperature and humidity,artificial seeds with different degrees of aging were obtained.The rice seeds were divided into 5 parts,of which 4 were placed in a drying oven with a temperature of 50 DEG C and humidity of 99%,respectively,aged 2 days(d),4 days(d),6 days(d)and 8 days(d),so that they were naturally dried.The seed samples of 5 different degrees of aging were obtained with the aging degree of 0d,2D,4D,6D and 8D,and the near infrared spectra of various samples were obtained.(2)The spectral pretreatment of rice seeds with different aging degrees was carried out.Firstly,the appropriate wavelength range,i.e.,1771nm~2499nm with high signal-to-noise ratio,was selected.Choose 5 different pretreatment methods were Savitzky-Golay(S-G)first derivative + multiplicative scatter correction(MSC),S-G smoothing first derivative + average smoothing,standard normal transformation(SNV)+ Ddetrending,S-G first derivative to remove the trend of first order derivative PLS model the same evaluation,final selection of the S-G smoothing + MSC method for optimal smoothing for next analysis modeling.(3)On the basis of S-G smoothing +MSC pretreatment method,3 different models are set up: PLS-DA model,PCA+BP neural network classification model and PCA+SVM classification model.PLS modeling and prediction sets were 0.8162 R2 and 0.7726 RMSE,respectively 1.3931 and 1.5577;PCA+BP neural network model,the processed data were analyzed by principal component analysis to reduce the dimension,the first seven principal components as input layer BP neural network,training network,the accuracy of the classification model the 99%;the first seven principal components using PCA to establish the SVM classification model,obtained a validation set accuracy of 99.17% models.The experimental results show that the method used in this paper has high accuracy.This study verified the feasibility of identification of rice seed aging based on near infrared spectroscopy.It can be applied to the detection of rice seed aging.It can effectively improve the aging rate of seeds,research efficiency,and provide an effective method for seed selection and seed management.It lays a foundation for further rapid detection of rice activity.