Research On Non-destructive Detection Of Seed Vigor For Single Rice Seed By Near Infrared Spectroscopy

Rice is an important food crop in China.The quality of seeds directly affects the growth of crops and grain yield.The traditional seed vigor detection methods mainly include biological and physicochemical categories.These traditional methods are long-time consuming,expensive,technically difficult in the detection process,and there are high technical requirements for the operator.Moreover,these methods are destructive to the seeds,and the detected seeds cannot germinate again.In this paper,NIR spectroscopy is used to study the non-destructive testing of rice seed vigor.The main research work of this article is as follows:(1)To establish a quantitative evaluation model for the germination rate on near infrared spectroscopy of single rice seedA gradient artificial aging test was designed.The rice seeds of Shunliangyou 1158 were processed in the gradient artificial aging test to obtain rice seeds with different vigor levels.The germination rate of each aging gradient was measured by the standard germination test and the result shows that the germination rate decreased with the aging degree deepening.Three quantitative models of PLSR,SVM and PCR were established with the germination rate as response variable Y.For the 340-1010 nm wave band,SVM model has the best effect,followed by the PLSR model,and the 930-1655 nm wave band PLSR model has the best results.Using UVE and SPA for feature extraction can reduce the 1642 variables in the 340-1010 nm spectral matrix to 7% and 18%,respectively.(2)To establish a quantitative evaluation model for the vigor level based on near infrared spectroscopy of single rice seedThe germination process of some single seeds after gradient aging treatment was tracked and recorded,and the change of seedling length was measured every day.The process of growth and change of seedlings shows clustering,so the 6 artificial aging gradients are further divided into 3 vitality levels.According to the qualitative analysis of the level of vitality,spectral data was qualitatively analyzed,and four different classification models of SIMCA,SVM,PLS-DA and PCR-DA were established.The performance of the SVM classification model is better,with the highest precision and recall rate.The four models have a high accuracy rate for identifying high vigor seeds,but are not sensitive to low vigor seeds.(3)Detection research for storage time of rice seed with near infrared spectroscopy of single seedThe single seed spectra of three kinds of rice seeds with different storage time in Guangdong Province were collected.The principal component analysis is carried out on the spectra and the ANOVA of the first PC revealed that the repeatability between different storage times is large,and it is difficult to distinguish samples merely based on the data we have.Spectral data from Wuyou rice in 930-1655 nm wave band has a less inter-class repeatability phenomenon and spectral preprocessing and PLSR model establishment is done with these data.The RMSEC is 5.1404 and the square of determination coefficient of calibration set is 0.8051.It indicates that it is feasible to use near infrared spectroscopy to detect the storage time of rice seeds,but its accuracy is not so high.In this paper,based on near infrared spectroscopy analysis technology,detection of seed vigor of rice samples in South China was studied.Combined with chemometric methods and mathematical models such as spectral pretreatment and abnormal sample rejection,it can realize non-destructive and rapid analysis and identification of rice seed vigor.Extracting the characteristic bands can greatly reduces the amount of data required for the model.It provides support for the future practical application of seed vigor nondestructive testing technology.