Research on meat species and freshness identification method based on spectral characteristics

This paper proposed a method of meat recognition method based on the artificial neural network of wavelet denoising. In this study, visible reflected spectra (from 380 nm to 780 nm) of beef and pork with different freshness were measured with fiber sensor spectrometer. The wavelet multi-resolution analysis was employed and the ideal way of decomposing layers was selected to eliminate the burr noise or abnormal data caused by absorption and scattering spectra signals in optical fiber. Then a kind of with a double-hidden layer was applied to analyze meat spectral reflectance data, and the back propagation-artificial neural network (BP-ANN) was trained by Polak–Ribiere conjugate gradient learning algorithm. The experimental results show that the method can analyse the complex spectrum signals and achieve a good identification on the species and freshness of meat.