Study On Real Time Nondestructive Detection Of Harmful Additives Content Of Flour Based On Raman Imaging Technology

Flour is the main product of wheat,which accounts for a large proportion of people’s diet structure,and the quality of flour not only can directly affect the quality of the pasta that use the product,but also is related to the body and psychology health and economic interests of consumers.In this paper,real-time qualitative identification and quantitative prediction model were developed of the mainly flour additives(benzoyl peroxide,VC)based on Raman imaging technology.Real-time Raman imaging detection system devices were built then debug and correct the hardware systems.The real-time control and analysis software was developed using Lab VIEW.The model of harmful additives content in flour was established for quantitative prediction,and two additives prediction models were verified on real-time detection analysis.The main conclusions are as follows:1.The scanning Raman imaging detection system is constructed,and the Raman spectrum and Raman image of the powder sample can be acquired.According to the actual needs of the detection and accuracy requirements of the detection system,Raman imaging spectrometer,CCD camera,lens,laser light source,mobile translation table,optical filter,and other major components were chosen.Then,the system was installed and its performance was tested.The corrected spectral CCD camera can detect Raman spectral range of-679.3～2885.7cm-1.The system ensured the actual spatial resolution was 0.22mm/pixel via space adjustion and stability of the test performances well,Raman images acquisition can be achieved.2.The real-time detection and analysis software of Raman imaging detection system was developed using LabVIEW programming.The detection software interface was designed applying modular and integrated design ideas in the LabVIEW-based software development environment.By using the software,the CCD camera,laser light source and the mobile station translational,parameter settings and hardware-detection state monitoring can be controlled and real-time collection and display of synthetic Raman image of test sample and Raman spectral curve dynamic display can also be completed.Using Matlab algorithms in LabVIEW to complete data were acquired and analysis and calculation results were saved.The scan line image synthesis and real-time display were completed using ENVI programming in LabVIEW.Software used one-button operation design and had the advantages of simple interface,easy operation and software portability.By combining with the software,the hardware detection system can totally meet the requirement of real-time analysis and application.3.The qualitative and quantitative model was established to identify and predict the major flour additive(benzoyl peroxide and VC).The respective Raman peaks and peaks of bond ownership of benzoyl peroxide and VC were determined by testing standard product using Raman imaging system.Benzoyl peroxide has obvious characteristic Raman peaks at 619 cm-1,848 cm-1,890 cm-1,1001 cm-1,1234 cm-1,1603 cm-1,1777cm-1 and the obvious characteristic peaks of VC at 449 cm-1,563 cm-1,630 cm-1,707 cm-1,822 cm-1,874 cm-1,1027 cm-1,1132 cm-1,1258 cm-1,1321 cm-1,1498 cm-1,1656 cm-1.The quantitative analysis models of benzoyl peroxide and VC were established with PLSR and MLR method.The correlation coefficient R2 of prediction set is 0.9891 of PLSR model of peroxide benzoyl.the prediction correlation coefficient R2 of MLR model is 0.9916 based on Raman characteristic peak at 1001 cm"1.The correlation coefficient R2 of MLR model based on 1001cm-1 and 1777cm-1 two characteristic peaks is 0.9932.The PLSR prediction model of VC was established with correlation coefficient R2 of 0.9886 and the correlation coefficient of prediction set of MLR model using the Raman peak at 630 cm-1 is 0.9887.MLR model based on 630cm-1 and 1656cm-1 two characteristic peaks was established with the correlation coefficient R2 in prediction set of 0.9892.The results indicated that MLR prediction model based on plurality of Raman peaks is better than PLSR model with higher accuracy.4.The implanted MLR quantitative analysis model of two additives in flour was verified and visualed in real-time detection.Adding flour benzoyl peroxide alone,real-time predictive correlation coefficient R2 is 0.9924.Real-time predictive correlation coefficient R2 when adding VC separately is 0.9888.In further research,flour mixed with the two additives was tested for real-time forecast verification with the prediction correlation coefficient R2 of 0.9912 for benzoyl peroxide,and real-time predictive correlation coefficient R2 of VC is 0.9878.Real-time validation of analytical results shows that the establishment of benzoyl peroxide and VC quantitative analysis model can meet the requirements of real-time test.