Study On Freshness Evaluation Of Conditioned Chicken Based On Hyperspectral And Optimized Bp Neural Network

Conditioned meat products are based on livestock and poultry meat,added with appropriate seasonings,and made by a variety of complex processes.Non-instant meat products that require simple processing before consumption.With the improvement of people's living standards and changes in consumption attitudes,conditioned meat products with stable quality,convenient eating,high added value,and balanced nutrition are favored by consumers.Because the production process of prepared meat products is more complicated,it is also extremely susceptible to microorganisms during processing.The action of microorganisms and enzymes will accelerate the spoilage of meat.Corrupted meat products not only lose economic and edible value,but also bring great safety risks to consumers' health.Therefore,real-time,accurate and non-destructive testing of conditioned meat products has important practical significance for the development of the meat industry.The article took conditioned chicken as the research object,using hyperspectral technology,chemometrics,intelligent optimization algorithms,and machine learning to study the non-destructive testing of conditioned chicken freshness parameters,mainly including conditioned chicken data collection and processing,high-level fusion-based algorithms Spectral characteristic wavelength selection,and optimization of BP network parameters based on bird swarm and immune algorithm.The specific methods were as follows:(1)Obtaining the hyperspectral image of each sample by hyperspectral imaging technology,and collecting the sample freshness parameters,including color(L*,a*,b*),TVC,TBA,and pH,in turn using professional instruments and method,the total number of samples collected was 240.Then the SPXY algorithm was used to divide the collected chicken samples into a correction set and a prediction set,the numbers of which were 160 and 80,respectively.Finally,the standard normal variable transformation,the first derivative and the second derivative were used to preprocess the spectral information of the freshness parameter,and the cross-validation was used to establish a PLSR model of the freshness parameter,and the best prediction of the spectral information of each freshness parameter was selected.(2)Samples obtained by resampling Bagging multiple times,the four methods of non-information variable elimination method,continuous projection algorithm,competitive adaptive re-weighting algorithm and random frog leap algorithm was used to select their characteristic wavelengths,and then these characteristics Wavelengths were first locally fused,and then the local fusion results were summarized,and the best characteristic wavelength variable was selected according to the global fusion strategy.The number of characteristic wavelengths obtained by filtering the freshness parameters pH,TBA,TVC,L*,a*,b*through this fusion algorithm was 13,20,25,18,17,16 respectively.By comparing with the full wavelength,the prediction effect of the model established by the fusion algorithm was very close to that of the full wavelength model.Even the prediction effect on the TBA index exceeded the full wavelength model,and its RPD is 3.76.(3)The combination of bird swarm algorithm,immune algorithm and bird swarm immune algorithm was used to optimize the initial weight and threshold of BP,and the prediction models for the freshness parameters of conditioned chicken were established respectively.After using the optimization algorithm to optimize the BP neural network,the model prediction performance had been significantly improved,and the BSA-IA-BP model had the best prediction effect.Among them,the RPD of the BSA-IA-BP prediction model of pH,TBA,and TVC indicators was 3.29,4.54,2.58,and the three parameters L*,a*,and b*of the BSA-IA-BP prediction model were less than 1.5.According to the model performance evaluation standard,it showed that the freshness parameter prediction method proposed in this paper could measure pH,TBA,and TVC well.And color parameters are limited to distinguish high and low chemical values.The verification of the proposed fusion algorithm and based on the BSA-IA-BP model showed that the method was feasible for the detection of freshness parameters of chicken,and provided algorithm support for the research of meat freshness using hyperspectral technology and theoretical basis.