Research On Quantitative Detection Model Of Plum Deer Antler Cap Powder Adulterated With Cow Bone Powder

The Antler cap is a kind of deer medicinal herbs with high medicinal value,with the treatment of osteoporosis,antioxidant,antibacterial,inhibit breast cancer and other functions,which plum blossom antler cap in the antler cap varieties of medicinal value is the highest.Because the plum antler cap is the antler base after shedding the antlers,hard texture,can not be taken directly,so need to be beaten into powder to take,which provides a favorable condition for unscrupulous businessmen to adulterate the plum antler cap.After market research learned that the adulterated counterfeit powder is mainly cow bone powder.Since the two powders are extremely similar in odor,color and other surface characteristics,it is difficult to distinguish whether it is counterfeit product by odor and naked eye after mixing.At present,most of the existing detection methods are chemical detection methods,which have the disadvantages of long detection time and high cost.Another method is mid-infrared spectroscopy detection,which has the advantages of high efficiency,non-destructive and low cost.However,the existing model is an identification model,mainly to detect whether adulteration,can not detect the specific proportion of adulteration,and the identification is whether the adulteration proportion of 50%and below of the sample is adulterated.In this paper,we will conduct an in-depth study on this aspect to establish a prediction model to extend the adulteration proportion to 100%,and the established prediction model can predict the specific proportion of adulterated cow bone powder in the plum deer antler cap.The specific work is as follows:(1)Relevant materials were collected.Samples of plum antler caps were obtained from five places:Xingkai Lake in Heilongjiang,Shuangyang District in Changchun,Jilin,Dongfeng County in Jilin,Changbai Mountain in Jilin,and Xifeng County in Liaoning Province,the farmers’trading market in Changchun City,the farmers’market in Nanguan District in Changchun,and the drug potassium bromide crystals required for the spectroscopic experiments were from Tianjin Optical Factory.(2)In response to the adulteration of plum deer antler cap powder with fake cow bone powder,this paper used mid-infrared spectroscopy to scan 30 samples each of plum deer antler cap adulterated with 0%,5%,10%,20%,40%,60%,80%,and 100%cow bone powder,for a total of 240 samples.Since the mid-infrared spectra are susceptible to the influence of temperature,humidity,carbon dioxide content,sample state,light source state,etc.to generate a lot of noise as well as baseline drift and other errors that lead to the reduction of modeling accuracy.After reviewing the literature and combining with the actual situation,this paper adopts eight pre-processing methods combined with normalization(Norm)and principal component analysis dimensionality reduction(PCA)to optimize the spectral data,eliminate the interference factors in the original spectral data,and improve the modeling accuracy.(3)Due to the large number of bands in the mid-infrared spectrum,the feature dimensionality is extremely high,which will increase the model complexity and affect the model accuracy to some extent when building the model.In order to reduce the dimensionality and make the modeling more accurate,this paper chose to use principal component analysis(PCA)for dimensionality reduction,and combined with support vector regression(SVR),random forest(RF),and partial least squares regression(PLSR)algorithms to build PCA-SVR,PCA-RF,and PCA-PLSR models.The results showed that the coefficient of determination(Rp~2)of the best prediction set for PCA-SVR and PCA-PLSR models were 0.9874 and 0.9770,respectively,and the model fit was high.the value of the coefficient of determination(Rp~2)of PCA-RF model was 0.9495,and the coefficient of determination was poor compared with the above three algorithms.(4)Subsequently,the model was built by full spectrum(FS),and the optimal prediction set coefficient(Rp~2)value of the FS-RF model was 0.9909.Comparing the optimal results of the three algorithms,it is clear that the FD-RF model is best modeled by normalizing(Norm)the full spectrum data(FS)through first-order derivative(FD)and multiple scattering correction(MSC).The final coefficient of determination(Rp~2),mean absolute percentage error(MAPEP),root mean square error(RMSEP),and relative analysis error(RPD)are 0.9909,0.2142,0.0333,and 10.4748,respectively.the above evaluation indexes indicate that the model has a high degree of fit and good prediction ability.The research in this paper has some practicality and can be used to predict the proportion of adulterated(cow bone powder)of plum deer antler cap products in the market,which is of practical application to realize the quality monitoring of plum deer antler cap.