The Study On Detection Of Mutton Freshness Based On Hyperspectral Technology

With the development of economy and the improvement of living standard,consumers’ demand for meat is increasing gradually,and they pay more attention to the safety of meat food.The freshness is an important indicator of meat quality,which puts forward higher requirements for the detection technology and evaluation method of meat freshness.The traditional meat freshness detection technology is tedious and timeconsuming,so it is urgent to develop a fast and stable detection technology of meat freshness.In this thesis,chilled mutton is taken as the research object.At first,hyperspectral technology and stoichiometry are used to obtain spectral information data,physical and chemical index values of freshness(TVB-N,PH),microbial index values of freshness(TVC),and color difference values of freshness(L*,a*,b*).Then,through spectral pretreatment and characteristic wavelength selection,the freshness detection model and grading model are established to quickly and accurately detect the freshness of chilled mutton.The main work of this thesis is as follows:(1)Preprocessing of spectral data.In view of the noise and baseline drift in the original spectral data,the effects of different preprocessing methods on the original spectral data are compared.The experimental results show that the S-G convolution smoothing combined with multivariate scattering correction method is the best method for preprocessing the original spectral data.(2)Based on AW-OPS characteristic wavelength selection algorithm,the detection model of chilled mutton freshness is established.Firstly,the ordered predictive selection algorithm(OPS)is used to filter the characteristic wavelengths,which reduces the dimension of spectral data.Secondly,a novel characteristic wavelength selection algorithm(AW-OPS)based on exponential decay function(EDF)is proposed.AW-OPS uses information vectors to measure the invalid information,interference information and importance of spectral data and uses EDF to remove wavelength variables with lower importance.Then the best characteristic wavelength is obtained according to the minimum RMSECV of the model.AW-OPS and OPS are used to extract the corresponding characteristic wavelengths and the detection model of mutton freshness based on PLSR is established.The experimental results show that the correlation coefficient of AW-OPS-PLSR detection model Rp=0.9781,increased by 1.11%,the minimum root mean square error RMSEP=1.3015,decreased by 0.13,and the operating efficiency increased by 73.97%.Compared with full band,Rp is increased by 2.5%and RMSEP is decreased by 0.6.Compared with other commonly used wavelength selection algorithms,Rp is improved and RMSEP is lower,which proves the effectiveness of AW-OPS feature wavelength extraction algorithm.(3)Based on ADOBA,support vector machine is optimized to establish the freshness grading model of chilled mutton.The chilled mutton samples under different storage periods are divided into fresh,sub fresh and rotten.In order to improve the accuracy of the grading model of mutton freshness,the improved bat algorithm(ADOBA)optimized support vector machine(SVM)cold fresh mutton freshness grading model is established by introducing adaptive inertia weight,disturbance strategy and Cauchy mutation operator mechanism strategy.An improved bat algorithm(ADOBA)is used to optimize the support vector machine to grade the freshness of chilled mutton.The experimental results show that the classification accuracy of ADOBA-SVM model is 97.22%.Compared with the bat algorithm before improvement,the accuracy is increased by 45.82%.Compared with the current commonly used intelligent algorithms,the accuracy has been improved by 34.68%on average,which shows the effectiveness of ADOBA.