Detection Methods Of Chilled Pork Freshness With Terahertz Spectroscopy

Freshness is the most important aspect of pork quality.To overcome defects of traditional freshness detection methods,such as subjective sensory evaluation,long physical and chemical analysis period and destruction,and to provide technical basis for rapidly determining meat shelf life and supervising food safety,a new method of pork freshness detection based on THz spectroscopy was designed by utilizing THz spectrometer as experimental platform.Spectra prediction model of pork freshness was established through comparison and optimization based on chemometrics principle.Mechanism of pork freshness detection by THz spectra and key technology in experiments and model establishment were studied.These researches lay a theoretical foundation for the development of meat freshness detection equipment based on THz spectroscopy.Through summarizing existing evaluation indices of pork freshness,close relationship between freshness indices and meat quality change,and measurement convenience,accuracy and repeatability were compared.K value was selected as reference quantitative index of freshness in THz spectral prediction model.K values of 80 pork samples with different freshness were measured by high performance liquid chromatography(HPLC).The basic law that K value increases with freshness decrease was studied.Based on inspection principles of THz spectra analysis technology,several THz spectral acquisition methods for biological samples were compared and studied.In order to effectively overcome the interference of water on THz wave and sample preparation,attenuated total reflection(ATR)model was selected as THz detection method in this paper.Meanings and correlations of several optical parameters of THz-ATR were studied.Attenuated total reflectance R(?)was used to express ATR absorption characteristics of samples.According to standard variance spectrum of repeat spectra(SVSRS),effective frequency band was determined as0.2~2THz.Accuracy of THz experiments was verified by THz spectrum of pure water.Operation steps and other related parameters of THz experiments were also determined.Absorption characteristics of muscle,fat and skin tissues in THz band were compared.It was found that THz spectra characteristics are obviously different among these tissues,and absorption characteristics are determined by water content and chemical compositions of tissues.On the basis of comparing chemical compositions and degeneration mechanism of muscle and adipose tissues,K value of muscle tissues was used to express sample freshness.THz spectra of 80 pork samples were obtained,and prediction models of pork freshness were established.Although THz spectra of pork samples had no obvious absorption peak,mathematical models based on spectral pretreatments and chemometrics were used to study THz characteristic expression of biomolecules in freshness quality and to analyze the relationship between spectra and K value.Principal component analysis(PCA)method was used for dimensionality reduction and compression of original spectral data.Three K prediction models,such as principal component regression(PCR),partial least squares regression(PLSR)and back-propagation neural network(BP-ANN),were built after spectra were pretreated respectively by multiple scatter correction(MSC),standard normal variate transformation(SNVT),first derivative(FD)and second derivative(SD).Combination of different pretreatment methods and model parameters in THz spectra were compared.The study showed that FD pretreatment method can effectively eliminate baseline drift of original spectra.BP-ANN prediction model can better analyze complicated regression relationship between THz spectra data and K value and is more suitable for K value prediction.Correlation coefficient of prediction set(R_P)of BP-ANN model after FD pretreatment was 0.75,and root mean square error of prediction set(RMSEP)was 14.36%.BP-AdaBoost algorithm was used to optimize BP-ANN prediction model of pork freshness.Calibration set samples were trained in BP-ANN model,and 10 weak predictors with different prediction errors were obtained.Integration weights were calculated according to prediction error.Integration weight was large with small prediction error,and integration weight was small with large prediction error.Prediction precision of BP-ANN model was further improved after 10weak predictors were integrated into one strong predictor by weighting strategy.R_P was increased to 0.84 and RMSEP was reduced to 9.89%in the prediction set.The freshness detection software was designed based on BP-AdaBoost model,which could inspect K value of pork conveniently.It lays the foundation for the practical application of the inspection method.This study showed that THz spectroscopy analysis combined with suitable pretreatment method,scientific prediction model,ATR detection mode and detection frequency band in0.2~2THz can be used to predict pork freshness non-destructively.Spectral prediction model is the key of detection.Compared with traditional method of HPLC,sample preparation process is omitted.The whole process from THz spectra measurement to K value output could be completed in 3 min,so as to meet the needs of rapid detection of production and circulation links.Convenient THz detection equipments without radiation can be further designed and developed.This study extends the application of THz technology in food safety inspection field and thus has important practical significance.