Detection Of Quality Attributes Of Purple-Fleshed Sweet Potato During Ultrasound Assisted Hot Air Drying By Hyperspectral Imaging Technique

Drying is one of the most important preservation methods in food industry.Hot air drying is a traditional drying technology.However,the traditional hot air drying has some deficiencies including lengthy drying time and poor product quality.Therefore,some effective physical processing techniques including ultrasonic reinforcement could be adopted to promote the mass transfer rate of drying process,so as to achieve rapid and high quality drying.On the other hand,quality monitoring is very important for drying process control.However,the conventional methods of quality detection can not satisfy the requirements of rapid,non-destructive and real-time detection in modern drying process.The spectral technique can detect the internal quality characteristics of food,but it can not acquire the distribution of quality information and determinate external attributes.Machine vision technology can extract the external attributes of the spatial information of food,but it is difficult to detect and analyze the internal quality of the object.Hyperspectral imaging technology is a new nondestructive determination method which combines the advantages of spectroscopy technology and machine vision technology.It can obtain the spectral information and spatial information of food simultaneously,so as to achieve rapid prediction of physical properties,chemical attributes and obtain the internal spatial information in samples.The application of hyperspectral technology in food quality and safety detection has attracted more and more attention and plays an important role in the field of food processing and quality detection.Therefore,the application of hyperspectral imaging technology in drying process can theoretically achieve rapid,accurate and non-destructive detection of product quality.The fresh purple-fleshed sweet potato(PFSP)was treated as the research object.Firstly the drying characteristics and quality characteristics of the ultrasound assisted hot air drying was studied,secondly the moisture content and freezable water content of PFSP slices during drying process were detected rapidly using hyperspectral imaging technique,thirdly the potential of hyperspectral imaging for rapid prediction of anthocyanin content of PFSP slices during drying process was investigated and the spatial distribution of anthocyanin content during drying process was visualized,and forthly the effect of different spectral information for accuracy of total flavonoids prediction models during ultrasound drying process was explored.The research results would provide theoretical support for rapid and nondestructive prediction ofquality in PSFP slices by hyperspectral imaging system during drying process.The main research contents are as follows:1.A contact ultrasound assisted hot air dryer was used to investigate the drying characteristics and quality characteristics of PFSP drying drying at different drying temperatures and ultrasound powers.The results showed that the increases of drying temperature and ultrasound power significantly reduced drying time and increased drying rate.Weibull distribution model could fit the drying curve accurately under different drying conditions.The moisture diffusion coefficient values increased with the increases in drying temperature and ultrasound power.The activation energy decreased as ultrasound power increased.The scanning electron micrographs of dried PFSP slices showed that the increase of ultrasound power could produce more microchannels and dilate intercellur space.Differential scanning calorimetry(DSC)results showed that the application of ultrasound during HAD process of PFSP could improve the mobility of water and hence decrease glass transition temperature.Under the same ultrasound power,the amount of total phenolic and flavonoid increased at low drying temperature and then decreased with the increase of drying temperature.Both phenolic content and total flavonoid content increased with the increase of ultrasound power at low drying temperatures,and decreased with the increase of ultrasound power at higher temperatures.The energy consumption of drying process was significantly reduced with the increase of ultrasound power.Therefore,the contact ultrasound technology is an effective method to accelerate drying rate and achieve better quality.2.The rapid and accurate detection of moisture content and freezable water content during drying process was achieved by using hyperspectral imaging.The corresponding mean reflectance spectra from regions of interests in hyperspectral images were extracted.Moving average,multiplicative scatter correction(MSC)and Savitzky-Golay smoothing filter were investigated to preprocess the raw spectra and partial least square regression(PLSR)calibration model was established to analyze the relationship between the extracted spectral data and measured quality attributes.Instead of choosing full range spectra data,optimal wavelengths were identified based on the regression coefficients(RC)method.Then PLSR,multiple linear regression(MLR),and backpropagation(BP)neural network were used to establish models to predict quality attributes of samples simultaneously.The results showed that comparing the performance of models based on different preprocessing methods,thePLSR model with MSC pre-treatment presented best results.Ten wavelengths and nine wavelengths were selected by RC method for moisture content and freezable water content,respectively.The RC-MLR model was considered as the best for determining moisture content and freezable water content of PFSP slices with coefficients of determination for calibration of 0.9417 and 0.8708,respectively,and coefficients of determination for prediction of 0.9359 and 0.8592,respectively.3.The hyperspectral imaging technique in tandem with chemometrics analysis was applied to determine anthocyanin content within PFSP during drying process rapidly and nondestructively.PLSR and least-squares support vector machine(LS-SVM)were applied to establish the calibration models based on raw extracted spectrum and the spectrum preprocessed by four different methods.In order to simplify the calibration model,three algorithms including PLSR,LS-SVM and multiple linear regression(MLR)were used to build models based on optimal wavelengths selected by RC method.The results showed that the MSC pretreatments could improve the accuracy of PLSR and LS-SVM.RC methods obtained tem optimal wavelengths.And the RC-MLR model yielded the best results.Finally,distribution maps were developed based on an image processing algorithm to visualize anthocyanin content of PFSP at different drying periods which cannot be achieved with conventional methods.4.The effect of different spectral brands for the detection of total flavonoids content in PFSP during drying process at different ultrasonic powers were investigated.Different spectral preprocessing methods were applied in original full band spectrum(371~1023 nm),original visible band spectrum(400~760 nm)and original near infrared band spectrum(760~1023 nm).And the PLSR and LS-SVM models were established between the total flavonoids content with the preprocessed spectrum.The results showed that the prediction results of MSC pretreatment model were the best.After MSC pretreatment,the prediction accuracy of PLSR model based on visible band spectrum was higher than the prediction accuracy of LS-SVM model based on the full band spectrum and the prediction accuracy of PLSR model based on near infrared spectrum,and the coefficient of determination for training set was0.8941 and the coefficient of determination for prediction set was 0.7280.