Research On Nondestructive Detection Of Varieties, Moisture And Starch In Rice Based On Hyperspectral Imaging Technology

With the continuous development of hyperspectral image technology,more and more domestic and foreign scholars began to pay attention to its application in the nondestructive detection of agricultural products.Hyperspectral image technology combines the advantages of spectrum technology and image technology,which can be used to detect the intrinsic quality of agricultural products by the spectrum technology and be detected the external quality by the image technology.Rice is the staple food of more than half of the world's population.With the improvement of living standards,consumers pay more and more attention to the quality of rice.The traditional rice quality detection methods are mainly based on the human sensory judgment with low precision,poor efficiency,but the hyperspectral technology can non-destructive detection of rice quality rapidly and accurately.In this paper,the rice was taken as the research object.Research on nondestructive detection of varieties,moisture and starch in rice based on hyperspectral imaging technology.In this study,a new method for rapid detection of rice variety,water content and starch content based on hyperspectral image technology is proposed.The main research contents of this paper are as follows:(1)First of all,the hyperspectral image technology was used to identify four varieties of rice from four different areas.Hyperspectral images of rice samples were obtained by hyperspectral imaging system,and the region of interest of rice was determined by ENVI software,the texture and morphological characteristics of rice were extracted by using MATLAB V7.8 software.Then the principal component analysis(PCA)was used to extract the characteristic wavelengths of the spectrum,texture and morphology,nine characteristic wavelengths(499.8,543.6,606.8,669.3,706.6,722.1,815.7,877.6,950.6nm).Seven SVM models based on full wavelength and characteristic wavelength are established respectively.Comparing the two kinds of models,the optimal model is based on the combination of spectrum,morphological and texture features of the characteristic wavelength.The precision reached to 91.67%.Therefore,it is feasible to detect rice varieties by using hyperspectral image technology.(2)Secondly,the moisture of rice was quantitatively detected by hyperspectral image technology in this paper.In this study,a method for quantitative determination of rice moisture based on hyperspectral image technology was proposed.First,hyperspectral images of rice samples in the range of 780-1800 nm were obtained by hyperspectral imaging system.The ENVI software was used to determine the region of interest of rice samples to extract the spectrum data.After the spectrum data extract,the moisture content of rice samples was determined by the method of constant temperature of 105?,according to the “GB5497-85 determination of moisture content in grain and oilseeds”.The original moisture content CMW0=13.5%.To simplify the calibration model,successive projections algorithm(SPA)was used for feature selection and the number of characteristic wavelengths was determined as 25.Principal component analysis(PCA)was used for feature extraction and the cumulative contribution rate of the first six principal components reached 99%,which could reflect most of the information of the full spectrum data.Then,SVR,LS-SVR and BCC-LS-SVR algorithm are used to establish the full wavelength spectrum and characteristic wavelength spectrum detection model.Compared with the old model,the accuracy of the BCC-LSSVR model based on radial basis function is the highest,with Rp2 of 0.980,RMSEP of 0.967%,Rc2 of 0.985 and RMSEC of 0.591%.The overall results from this study demonstrated that hyperspectral image technology is feasible to detect rice moisture.(3)Finally,the starch content of rice was detected by hyperspectral image technology.In this study,a method for quantitative determination of rice starch based on hyperspectral imaging technology was proposed.Firstly,the hyperspectral imaging system in the spectrum range of 871-1766 nm was used to collect the hyperspectral images of 100 rice samples of 10 starch grades.According to the national standard “GB/T 5009.9-2008 determination of starch in food” by acid hydrolysis method as the standard chemical analysis method detection of starch content of rice sample.Support vector regression(SVR)model was established to determine the starch content by using full wavelengths spectrum data.The principal component analysis(PCA)and successive projections algorithm(SPA)were used to extract the feature extraction of rice samples,and the SVR regression model was established by using the full wavelength spectrum and the characteristic wavelength spectrum.The accuracy of PCA-SVR model based on RBF kernel function is the highest,Rc2 is 0.989,RMSEC is 0.445%,Rp2 is,RMSEP is 0.669%.The results show that it is feasible to detect rice starch content based on hyperspectral image technology.