The Research Of Moisture And Protein Content Of The Detection Method In Rice Based On Near Infrared Spectroscopy

With the continuous improvement of people's living quality and paying more and more attention to physical health,the demand for palatability and nutritional quality in rice of consumer is getting higher and higher.The quality of rice not only enables consumers to enjoy in the senses and is directly related to the body's digestion and absorption of rice,but also in the field of rice trade and related scientific research is also crucial.In the assessment of rice quality,moisture and protein are key factors in evaluating the value of rice.The moisture content of rice not only affects the quality of rice,more importantly,is related to the public food safety issues.Rice and its by-products also provide people with daily energy and protein and the level of protein content plays an important role in the rice taste quality.Therefore taking this as a starting point,this paper uses rice as the research object and combines the near infrared spectroscopy with moisture and protein content to fit the spectrum in order to explore the possibility of rapid detection.The mainly research work carried out in this paper is as follows:(1)A total of 109 different kinds of rice samples were collected from all parts of Heilongjiang and the Antaris II near infrared spectrometer of Thermo Fisher company were used to scanned these samples.Then,the chemical values of moisture and protein which were the main components were measured by the traditional national standard chemical method to Prepare for subsequent data processing and model building.(2)Removing abnormal samples for quantitative analysis model of protein in rice was by 3methods such as Hotelling T2 statistics X-Y residuals and 3D view analysis.The results showed that the RMSECV and R2 of the model were increased from the initial 0.2428 and 0.7626 to the0.2060 and 0.8364 respectively after removing abnormal samples by X-Y residual analysis method.(3)The duplex method was used to divide the samples of moisture and protein into calibration set and prediction set.The result indicated that the two sets were very similar not only in range,but also in the average value and standard deviation of the sample.There,the calibration set and the prediction set which meet the experimental requirements and were uniform distribution were obtained so that the expected effect of this study was achieved.(4)The original spectrum of rice moisture was denoised respectively by three methods including derivative,normalization and smoothing.By comparison,it is found that the smoothingmethod with 15 points is the best model.The original spectrum of rice protein was eliminated by the three denoising methods including first derivative + smoothing,second derivative + smoothing and orthogonal signal correction.The results showed that the second derivative + smoothing had the best denoising effect.The determination coefficient of calibration set of the model R2 was more inclined to 1.At the same time,there was a lower mean squared error root RMSECV of the calibration set.(5)The water spectrum of rice was modeled by MWPLS and IPLS band selection method.It was found that MW-IPLS was an effective method to select the characteristic absorption band of moisture spectrum by comparison.Combining MWPLS and IPLS methods to select the band,the optimal band range of the two cross 4108-4386cm-1 was determined.In this band,PLSR was used to establish the quantitative model of rice moisture.The root mean squared error of prediction set RMSEP was 0.2753,and the determination coefficient R2 was up to 0.8597.(6)The PLSR and PCR models were established for the protein spectrum in rice samples of 80 calibration sets and 24 prediction sets.By comparing the root mean squared error root RMSEP and the determination coefficient R2 of prediction set,the PCR model showed better performance in predicting the protein content of rice.The root mean squared error RMSEP and the determination coefficient R2 of prediction set were respectively up to 0.1288 and 0.8865.In summary,the model established for moisture and protein content in rice by near infrared spectroscopy technology in this paper had a higher accuracy and certain feasibility,providing a new method for rapid detection of rice component in the future.