Research On Portable Detection System Of Soybean Crude Fat Content Based On Near Infrared Spectroscopy

As the main economic crop in China,soybean has the characteristics of wide distribution,high nutritional value and market demand.Its fat content is one of the important indicators to evaluate the quality of soybean.Traditional chemical detection method has high precision,but its process is cumbersome,time-consuming and high cost.Near infrared spectroscopy has been widely used in the field of grain and crop composition detection because of its fast speed and simple operation.In this paper,based on the nearinfrared spectroscopy technology,the spectral modeling analysis of soybean particle samples and powder samples was carried out,on this basis,the feasibility of using the particle sample spectrum to establish the detection model of soybean crude fat content was verified.The characteristic wavelength of the particle sample spectrum was extracted,and according to the extracted characteristic wavelength,a portable soybean crude fat content detection system was designed to realize the detection of soybean crude fat content.The rapid non-destructive detection of crude fat content has great significance for guiding soybean procurement and evaluating soybean quality.The research contents and conclusions are as follows :(1)Near infrared spectroscopy(NIR)was used to establish the detection model of crude fat content in soybean granules and powders based on full spectrum.Based on near infrared spectrometer,the spectral information of samples was collected,and the crude fat content of soybean was determined by physical and chemical analysis method.The original spectra were pretreated with multiple scattering correction,normalization and moving window smoothing,and the partial least squares regression models were established respectively.The comparison results showed that the prediction effect of the original spectra after normalization was the best: For the detection model of crude fat content in soybean granules,the calibration set value of Rc and RMSECV are 0.8873 and 0.7146,respectively,the prediction value of set Rp and RMSEP were 0.8795 and 0.7621,respectively.For the detection model of crude fat content in soybean powder,the calibration set value of Rc and RMSECV were 0.9084 and 0.6897,respectively,the prediction set value of Rp and RMSEP were 0.9295 and 0.6462,respectively.The results showed that the accuracy of soybean crude fat content detection model will be improved after proper pretreated.At the same time,although the accuracy of soybean granules crude fat content detection model is slightly lower than that of soybean powder crude fat content detection model,its Rc and Rp are still in a higher range,which can achieve the prediction of crude fat content.Therefore,the detection model of crude fat content in soybean granules samples can be used for rapid non-destructive detection of crude fat content in soybean.(2)The characteristic wavelengths are extracted to simplify the model and used in the design of detection system.The competitive adaptive reweighting method,continuous projection algorithm and variable combination cluster analysis method were used to extract the characteristic wavelengths of soybean granules samples,Partial Least Squares Regression models were established respectively,and the results showed that the five wavelength points extracted by variable combination cluster analysis had the best model effect,with Rc value of 0.8795,RMSECV value of 0.7883,Rp value of 0.8398 and RMSEP value of 1.0734.Through the selection of characteristic wavelength,the calculation time is greatly shortened,the detection model is simplified,and It provides a basis for the design of detection system.(3)A portable system for detection soybean crude fat content based on spectral characteristic wavelength was designed,The modular design was adopted and the prototype of detection system was made.The system is used to collect data from 48 samples,and establish the model,the Rc value is 0.8093,the RMSECV value is 1.2518,the Rp value is0.7986,RMSEP value is 1.3013,the model can detect the crude fat content of soybean,the model is imported into the detection system,on this basis,12 samples are used for the actual component detection,the detection results show that the portable detection system can be applied to the rapid non-destructive detection of soybean crude fat content.