Nondestructive Testing Of Chilled Mutton Based On Hyperspectral Imaging Technology

The health and safety quality of chilled mutton is of great significance for food safety control and consumer safety.The quality parameters of mutton can be used to measure its health and safety.However,the efficiency of traditional detection methods is low and the samples will be affected in the detection process.Therefore,spectral analysis technology has been developed,which can be based on the spectrum of different substances for quality detection.Paper for cold fresh lamb study,wherein the detected basic nitrogen(TVB-N)based on spectroscopy,p H,coliform(MPN)and Total number of colonies(TBC)relevant indicators.In the wavelength range of 400 nm ? 1000 nm,the reflection spectrum is extracted from the hyperspectral image,combined with deep learning and hyperspectral imaging technology to study the classification and identification of cold fresh lamb.The following is the main content of the experiment:(1)Three mutton samples were processed and stored for 14 days,during which hyperspectral imaging data were collected.3 × 4 samples were selected and sent to the biological laboratory to determine the physicochemical and microbiological indexes.1 × 4 samples were left in the laboratory to collect spectral data.(2)In this thesis,vector normalization and standard normal variable(SNV)are used to smooth and denoise the spectral data.(3)Firstly,the data is preprocessed,then the CNN model is used for experiments,and then repeated experiments and repeated debugging of the model are performed,and the final accuracy rate is 87.12%.(4)The data is preprocessed first,and then the RNN model is used for experiments,and then repeated experiments and repeated debugging of the model are performed.The final accuracy rate obtained is 75.51%,and the RNN model is slightly lower.(5)Firstly,the data is preprocessed,and then the DNN model is used for experiments,and then repeated experiments and repeated debugging of the model are performed.The final accuracy rate is 90.09%,and the result is better than CNN and RNN.