Pork freshness detection and identification method

Abstract

The invention discloses a pork freshness detection and identification method. The method comprises the following steps: S1, acquiring electromagnetic scattering parameters of a pork sample as a first feature data set based on a state information collector with wireless signal emission and collection; s2, arranging a hyperspectral imaging module to obtain surface reflection spectrum data of the pork sample, and taking the surface reflection spectrum data as a second feature data set; s3, arranging a gas sensor array to detect the volatile basic nitrogen concentration of the pork sample, and taking the volatile basic nitrogen concentration as a third characteristic data set; s4, arranging a micro impedance spectrometer to measure the dielectric property change of the muscle tissue, and taking the dielectric property change as a fourth characteristic data set; s5, constructing a deep neural network model, and dynamically distributing a weight coefficient based on an attention mechanism; and S6, outputting a freshness grade prediction value. Compared with the prior art, the pork freshness detection and recognition method has the advantages that multi-source data collection and multi-feature weighted evaluation are achieved, and stable detection is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of food testing technology, specifically to a method for detecting and identifying the freshness of pork. Background Technology

[0002] As the world's most consumed meat product, pork's freshness is directly related to food safety and consumer health. Traditional testing methods mainly include sensory evaluation, physicochemical index testing (such as volatile basic nitrogen TVB-N), and microbiological testing, but these methods have limitations such as long testing cycles, complex operations, and sample destruction.

[0003] Currently, single sensors are susceptible to environmental interference and cannot fully reflect the changes in electromagnetic, spectral, and chemical properties during the pork spoilage process. At the same time, traditional methods lack a dynamic weight allocation mechanism for the fusion of multi-source data, which leads to the submergence of key feature information, resulting in poor detection stability and affecting the accuracy of the results. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the above-mentioned technical defects and provide a method for detecting and identifying pork freshness by multi-source data acquisition, weighted evaluation of multiple features, and ensuring stable detection.

[0005] To solve the above-mentioned technical problems, the technical solution provided by the present invention is as follows: a method for detecting and identifying the freshness of pork, comprising the following steps: S1: acquiring electromagnetic scattering parameters of pork samples as a first feature data set based on a state information collector with wireless signal transmission and collection capabilities; S2: deploying a hyperspectral imaging module to acquire surface reflectance spectral data of pork samples as a second feature data set; S3: deploying a gas sensor array to detect the volatile basic nitrogen concentration of pork samples as a third feature data set; S4: deploying a miniature impedance spectrometer to measure changes in the dielectric properties of muscle tissue as a fourth feature data set; S5: constructing a deep neural network model, using the first, second, third, and fourth feature data sets as training data inputs, and dynamically allocating weight coefficients for the second, third, and fourth feature data sets based on an attention mechanism; S6: outputting a predicted freshness level value.

[0006] Preferably, the electromagnetic scattering parameters in S1 include signal strength, rate of change, and multipath effect attenuation coefficient.

[0007] Preferably, step S2 includes extracting the absorbance ratios of the 450nm, 580nm, and 630nm characteristic wavelength bands.

[0008] Preferably, the hyperspectral imaging module uses a tunable liquid crystal filter and a dual-source compensation system, wherein the dual-source compensation system includes a white LED and an ultraviolet LED.

[0009] Preferably, in S3, the gas sensor array consists of functionalized carbon nanotubes, a QCM sensor modified with molecularly imprinted polymers, and a MOS sensor group with temperature and humidity compensation.

[0010] Preferably, the attention mechanism in S5 includes a spatial-spectral attention module and a temporal-chemical attention module;

[0011] The spatial-spectral attention module associates the second feature data group with the temporal-chemical attention module, and the second feature data group associates the third feature data group with the fourth feature data group.

[0012] Preferably, step S1 includes acquiring the CSI phase difference matrix in both 2.4GHz and 5GHz frequency bands, eliminating environmental noise through wavelet transform, and then calculating the ratio of the signal strength standard deviation ΔS to the initial value as the signal strength change rate.

[0013] Preferably, the measurement of the dielectric properties of muscle tissue in S4 includes scanning the impedance spectrum in the frequency range of 10kHz-1MHz, extracting the real conductivity σ' and the imaginary dielectric constant ε", and calculating the relaxation time distribution function as an indicator of the degree of putrefaction.

[0014] Preferably, the deep neural network model includes an electromagnetic feature processing unit, a spectral feature processing unit, and a chemical feature fusion unit.

[0015] The advantages of this invention compared to existing technologies are as follows: In this invention, multiple sources of data, such as electromagnetic scattering parameters, reflectance spectral data, volatile basic nitrogen concentration, and changes in the dielectric properties of muscle tissue, are used as evaluation data. Combined with the attention mechanism, key features in the process of pork spoilage are effectively extracted, thereby improving prediction accuracy. At the same time, no chemical treatment of pork samples is required, thus preserving the integrity of the samples. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a method for detecting and identifying the freshness of pork. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings.

[0018] Combined with appendix Figure 1As shown, a method for detecting and identifying the freshness of pork includes the following steps: S1: Obtaining electromagnetic scattering parameters of pork samples as a first feature data set based on a state information collector with wireless signal transmission and collection capabilities; S2: Deploying a hyperspectral imaging module to obtain surface reflectance spectral data of pork samples as a second feature data set; S3: Deploying a gas sensor array to detect the volatile basic nitrogen concentration of pork samples as a third feature data set; S4: Deploying a miniature impedance spectrometer to measure changes in the dielectric properties of muscle tissue as a fourth feature data set; S5: Constructing a deep neural network model, using the first, second, third, and fourth feature data sets as training data inputs, and dynamically allocating weight coefficients for the second, third, and fourth feature data sets based on an attention mechanism, wherein the deep neural network model includes an electromagnetic feature processing unit, a spectral feature processing unit, and a chemical feature fusion unit; S6: Outputting a predicted freshness level value.

[0019] In this invention, the coefficients in S1 include signal strength, rate of change, and multipath effect attenuation coefficient. After collecting the CSI phase difference matrix in the 2.4GHz and 5GHz dual-band and eliminating environmental noise through wavelet transform, the ratio of the standard deviation of signal strength ΔS to the initial value is calculated as the rate of change of signal strength.

[0020] S2 includes extracting the absorbance ratio of the characteristic wavelength bands of 450nm, 580nm, and 630nm. The hyperspectral imaging module uses a tunable liquid crystal filter and a dual-source compensation system, which includes a white LED and an ultraviolet LED.

[0021] In S3, concentration analysis was performed using a gas sensor array with functionalized carbon nanotubes, a QCM sensor modified with molecularly imprinted polymers, and a MOS sensor group with temperature and humidity compensation.

[0022] The measurement of the dielectric properties of muscle tissue in S4 includes scanning the impedance spectrum in the frequency range of 10kHz-1MHz, extracting the real part conductivity σ' and the imaginary part dielectric constant ε", and calculating the relaxation time distribution function as an indicator of the degree of putrefaction. The attention mechanism in S5 includes a spatial-spectral attention module and a temporal-chemical attention module.

[0023] The attention mechanism in S5 includes a spatial-spectral attention module and a temporal-chemical attention module. The spatial-spectral attention module is used for the second feature data group, and the temporal-chemical attention module is used for the third and fourth feature data groups.

[0024] In practice

[0025] The CSI phase difference matrix of pork samples was collected in the 2.4 GHz band at a sampling frequency of 100 Hz. The CSI data was decomposed into 5 levels by wavelet transform, and the signal was reconstructed after removing high-frequency noise. The signal strength, rate of change, and multipath effect attenuation coefficient were obtained.

[0026] A hyperspectral camera was used to acquire the surface reflectance spectrum of pork, and the absorbance values ​​in the 450nm, 580nm, and 630nm bands were extracted and the ratios were calculated. A gas sensor array was used to detect the TVB-N concentration, and dielectric characteristics were collected at the same time.

[0027] Electromagnetic features are normalized, while spectral features are analyzed using principal component analysis. Gas and dielectric features are directly used as inputs. The results are then tested and predicted using a deep neural network consisting of three fully connected layers and two attention modules.

[0028] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

[0029] The scope of protection of this invention is defined by the appended claims and their equivalents. Therefore, the following detailed description of embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention.

[0030] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0031] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A method for detecting and identifying the freshness of pork, characterized in that: Includes the following steps: S1: The electromagnetic scattering parameters of the pork sample are obtained as the first feature data group based on the state information collector with wireless signal transmission and collection. S2: Deploy a hyperspectral imaging module to acquire surface reflectance spectral data of pork samples and use it as the second feature data group; S3: Deploy a gas sensor array to detect the concentration of volatile basic nitrogen in pork samples and use it as the third feature data group; S4: Deploy a miniature impedance spectrometer to measure changes in the dielectric properties of muscle tissue and use it as the fourth characteristic data set; S5: Construct a deep neural network model, using the first feature data group, the second feature data group, the third feature data group, and the fourth feature data group as training data inputs, and dynamically assign weight coefficients to the second feature data group, the third feature data group, and the fourth feature data group based on the attention mechanism; S6: Output the predicted freshness level.

2. The pork freshness detection and identification method according to claim 1, characterized in that: The electromagnetic scattering parameters in S1 include signal strength, rate of change, and multipath effect attenuation coefficient.

3. The pork freshness detection and identification method according to claim 1, characterized in that: S2 includes extracting the absorbance ratios of the characteristic wavelength bands of 450nm, 580nm, and 630nm.

4. The pork freshness detection and identification method according to claim 3, characterized in that: The hyperspectral imaging module uses a tunable liquid crystal filter and a dual-source compensation system, which includes a white LED and an ultraviolet LED.

5. The pork freshness detection and identification method according to claim 1, characterized in that: The S3 includes a gas sensor array with functionalized carbon nanotubes, a QCM sensor modified with molecularly imprinted polymers, and a MOS sensor group with temperature and humidity compensation.

6. The pork freshness detection and identification method according to claim 1, characterized in that: The attention mechanism in S5 includes a spatial-spectral attention module and a temporal-chemical attention module; The spatial-spectral attention module associates the second feature data group with the temporal-chemical attention module, and the second feature data group associates the third feature data group with the fourth feature data group.

7. The pork freshness detection and identification method according to claim 2, characterized in that: S1 includes acquiring the CSI phase difference matrix in the 2.4GHz and 5GHz dual-band, eliminating environmental noise through wavelet transform, and then calculating the ratio of the signal strength standard deviation ΔS to the initial value as the signal strength change rate.

8. The pork freshness detection and identification method according to claim 1, characterized in that: The measurement of the dielectric properties of muscle tissue in S4 includes scanning the impedance spectrum in the frequency range of 10kHz-1MHz, extracting the real part conductivity σ' and the imaginary part dielectric constant ε", and calculating the relaxation time distribution function as an indicator of the degree of putrefaction.

9. The pork freshness detection and identification method according to claim 6, characterized in that: The deep neural network model includes an electromagnetic feature processing unit, a spectral feature processing unit, and a chemical feature fusion unit.

Images