Identification Of Rice Based On Analysis Of Raman Spectrum And Organic Ingredients

In the process of implementing the food safety strategy,the traceability of food sources is one of the important ways to ensure food safety.Rice is the top one of the five cereals,and its safety has always been the focus of attention.Rice planting area accounts for a large proportion of China's grain and there are many varieties of rice.Traditional rice identification methods mainly stay in sensory detection and chemical detection,which are not only time-consuming and laborious,but also not easy to adopt.Some illegal traders substitute high-quality rice with non-high-quality rice to serve costumers.At the same time,the advancement of food source traceability technology has further improved the requirements of controling rice quality carefully.The research of the origin of single rice,the rapid and non-destructive testing technology of rice is of great significance to protect the income of rice farmers,to maintain the order of trade and to guarantee the safety of the tongue.The study regards northeast rice as the research object.The study concentrates on solving the problem of similar rice producing area and difficult classification of non-destructive testing with Raman spectroscopy to combine stoichiometric methods and computer programming techniques,and exploring single grain rice producing area and variety detection method.The main research contents and results are as follows:(1)The characteristics of rice spectrum were studied.In order to identify the rice varieties and the origin of rice,the rice spectrum of different laser intensity,different time of integration and different scanning position were firstly collected with Raman spectrometer.Secondly the spectrum was preprocessed through MATLAB2010 using mean,normalization,standardization and centralized Savitzky-Golay convolution transformation,smoothing,standard normal variables to trend,polynomial fitting,adaptive iterative weighted punishment least-square method and wavelet transform method,finally features were extracted with the wavelet transform.The results show that the optimal acquisition parameter is h,the integral time is 4s,the scanning frequency is 4,and the sampling location is rice root,middle and tip.After feature extraction,we got characteristic wavelength of 20,269,480,865,941,1129,1260,1339,1461,1339 cm-1 in Raman characteristic peak that showed strong vibration model,with its main nutritional components of starch,protein and fat correspond.(2)A model of rice identification based on Raman spectrum is established.In order to study classification identification of rice of the similar region,different varieties and similar variety,the same area,we obtained five varieties,8 kinds of rice spectrum of five origins in heilongjiang province.8 kinds of rice spectrum according to the varieties and regional is divided into three kinds of recognition,each contains three kinds of rice.We comply with the path of spectroscopy preprocessing-wavelength filter-classification algorithm for classification of spectrum analysis technique,and we compress and filter the spectrum data with principal component analysis and successive projection algorithm,and compare classification results of different combination with linear discriminant analysis and least squares support vector machine(SVM),partial least squares method to identify different groups of rice.The results show that the recognition effect of partial least square method is optimal,and the classification accuracy of different recognition groups is 100%.The linear discriminant analysis model of characteristic bands was established through the data of mean centralization and continuous projection algorithm,and the recognition accuracy of different recognition groups was 100%.Through principal component analysis and continuous projection algorithm,the dimension of rice spectral data can be effectively reduced,which can improve the speed of model operation recognition.(3)Application and optimization of rice identification model.In order to study the adaptive ability and generalization ability based on the Raman spectra of rice classification model of partial least squares,6 kinds of northeast rice were increased on the basis of the above eight kinds of samples,which covers the nine origins and seven varieties.We can compare the partial least squares method with the feature variables,compare full spectrum and 400~1700-1 band data of the continuous projection algorithm under different pretreatments,compare the effects of standardization and Kennard-Stone sample sets on the model.The results show that with the increase of rice variety,the prediction accuracy of different combination data shows a decreasing trend.After the Kennard-Stone sample set was divided,the accuracy of corresponding recognition increased.We use different spectral data and the classification of the partial least squares method based on the results of the different pretreatment and spectral characteristics,based on band Savitzky-Golay convolution smoothing-normalized spectral data processing-polynomial fitting-standardization modeling,and 14 kinds of rice classification accuracy is 97.14%,the result is optimal.(4)The identification mechanism of rice chemical composition and Raman spectra was carried out,and the rice identification method based on the characteristic peak area of the standard material Raman was proposed.For three varieties of 4 kinds of rice in two regions,we further collects organic chemical composition information,which contains protein,amylose and amylopectin,17 kinds of amino acid hydrolysate,35 species of 55 kinds of fatty acids.first of all,we study classification method based on chemical composition data.secondly,We determine the chemical composition which plays a main role in classification through the principal component load matrix and collect corresponding standard material for chemical composition of Raman spectra collection.finally,we study rice origin,variety classification method based on the organic combination of Raman spectra.The results showed that four kinds rice of the organic components is 32 through removing zero information,we can get rice identity recognition based on the data of 32 kinds of chemical composition with principal component analysis(pca)combined with cosine similarity,the accuracy is 100%;We can determine the content of aspartic acid and other 19 species and more than 70% of pullulan a total of 20 species that plays a main role in the chemical composition of classification,and put forward calculation method with standard material the Raman spectral characteristic peak area normalization into for the input of the partial least squares-cosine similarity,and calculated the 20 standard substance of Raman spectra characteristics of spectral peak interval and rice contains standard material characteristic spectrum peak area with partial least squares combining with cosine similarity analysis,the correct recognition rate can reach 96.875%.(5)Rice identification software system was developed.Firstly,rice raman spectrum and component information database based on cloud platform was created.Secondly,we develop the smooth,normalization preprocessing methods,such as developing the method of principal component analysis,partial least squares and support vector machine(SVM)mathematical model,developing the cosine similarity and adjusted cosine similarity and similarity comparison method;Again,we study and formulate the rice identification process based on the Raman spectra,which determine the process according to the sample set import and cloud database storage,modeling training set selection and pretreatment process Settings,algorithm Settings and modeling and forecasting data import,export order model prediction and result;In the end,the actual test were carried out on the data of Tailai rice raman spectrum and the rice composition of wuyou 2 rice in Dumeng jiangwan respectively,the result is accurate,which can be used as an identification tool for rice.Above research results provide an important scientific basis for rice quality monitoring and quick nondestructive testing to the quality and safety.Developing real-time online detection equipment has great guiding significance and application value.