Study On Nondestructive Detection Method Of Chilled Mutton Quality Based On Hyperspectral Technology

The health safety quality of chilled mutton can be reflected by its quality parameters. The traditional testing method has the problems of low efficiency, long cycle and can destroy the samples. With the development of high spectrum technology, it has become a hot research topic to reveal the quality of the method by using different spectra of different substances.The research object of this paper is cold and fresh mutton, in the range of different wavelengths, using hyperspectral imaging technology, the quality of cold and fresh mutton parameters to carry out rapid nondestructive testing research. The prediction model of the whole wave band and the characteristic band of cold fresh mutton quality parameters (L*, a*, b*,the value of PH, thevalue of TVB-N) were established. First determine the best pre processing method, followed by comparison comparison of different wavelength range and different bands, different modeling methods to determine the best method of nondestructive testing for cold fresh meat quality parameters, cold fresh mutton online rapid nondestructive detection system laid the foundation. The research results are as follows:(1) Standard value of detecting chilled mutton samples quality parameters and processed get color parameters (L in the 400-1100nm range and the optimum pre processing method is FD+MC. In the range:900-1700 best pretreatment method is MC. Color parameter a* in the 400-1100nm range and the optimum pre processing method is FD+MSC. In the range:900-1700 best pretreatment method is MC. Color parameter a* in the 400-1100nm range and the optimum pre processing method is SNV+MSC. In the range:900-1700 original spectra have the best prediction result. Parameter pH value in the 400-1100nm range of the best pretreatment method is FD, in the 900-1700nm range of the best pretreatment method is MC. Parameter TVB-N value in the 400-1100nm range of the best pretreatment method is S-G convolution smoothing, in the 900-1700nm range of the best pretreatment method is FD.(2) The prediction model established in the 400-1100nm wavelength range is better than the prediction model established in the 900-1700nm wavelength range.(3) Based on the full band using PLSR and BP neural network method of chilled mutton sample quality parameters, a prediction model is established parameters a*, b*, pH value of the PLSR prediction model, the parameter L*, TVB-N values of BP neural network prediction model is better.(4) Based on the characteristic band L*, a*, b*, pH value, TVB-N value is the PLSR prediction model is good, and is better than the whole band prediction model. The correlation coefficient of the parameter L* calibration set is Rc=0.958, the root mean square error is RMSEC=0.650, the correlation coefficient of the prediction set is Rp=0.924, and the root mean square error is RMSEP=0.890. The correlation coefficient of calibration set parameter a* is Rc=0.932, the RMS error is RMSEC=1.352, the correlation coefficient of the prediction set was Rp=0.909, the RMS error is RMSEP=1.374; the correlation coefficient of calibration set parameter b* is Rc=0.952, the RMS error is RMSEC=0.658, the correlation coefficient of prediction set was Rp=0.892, the RMS error is RMSEP=0.669; the correlation coefficient of parameters the pH value of calibration set is Rc=0.913, the RMS error is RMSEC=0.085, the correlation coefficient of the prediction set was Rp=0.906, the RMS error is RMSEP=0.082; the parameter TVB-N value of the correlation coefficient of calibration set is Rc=0.923, the RMS error is RMSEC=3.677, the correlation coefficient of the prediction set was Rp=0.835, the RMS error is RMSEP=5.759.