Pork safety tracing quality nondestructive testing method and system for livestock breeding

Abstract

The invention discloses a pork safety traceability quality nondestructive testing method and system for livestock breeding. The method comprises the following steps: constructing a stable detection environment by utilizing a constant temperature and humidity device, a dynamic environment monitoring unit and an optical shielding material; a Fourier transform infrared spectrometer is adopted to collect spectral data, and a non-contact texture detection device is combined to measure physical parameters; denoising, standardization processing and feature fusion are carried out on the collected data, and key variables are extracted through principal component analysis; grouping the samples by using a K-means algorithm, and verifying sample classification by using linear discriminant analysis; quantitative prediction and classification of pork quality are realized based on partial least squares regression and a support vector machine model; the model is applied to an actual scene, real-time processing and visual display are achieved through online data collection, and a data sample library is continuously expanded. The method has the advantages of high detection speed, high precision, wide application range and the like, and can be widely applied to the fields of meat quality detection and food safety.

Description

Technical Field

[0001] The present invention relates to the technical field of pork detection, and in particular to a nondestructive detection method and system for the safety and traceability quality of pork used in animal husbandry. Background Art

[0002] Pork, a staple meat product in my country, has a quality and safety that is directly linked to public health and food safety. In recent years, consumers have placed higher demands on the freshness, nutritional content, and taste of meat products. At the same time, China is vigorously promoting the development of a meat quality and safety traceability system, emphasizing comprehensive supervision from farm to table. Against this backdrop, accurately and efficiently evaluating the overall quality of fresh pork and linking these results with upstream farming information have become key technical challenges in the livestock and food industry.

[0003] Existing pork quality testing methods mainly include manual sensory evaluation, physical and chemical analysis, and some semi-automatic testing technologies; among them, sensory evaluation relies on experience, is highly subjective and has poor consistency; chemical analysis is destructive, time-consuming, costly, and difficult to apply to large-scale rapid screening of samples; some non-destructive testing methods based on single spectrum or image analysis can improve efficiency, but have problems such as single evaluation indicators, large environmental interference, and inability to comprehensively reflect the overall picture of meat quality. Moreover, most of them have not been effectively integrated with upstream breeding data, making it difficult to form a traceable quality chain.

[0004] In response to the above problems, the present invention provides a non-destructive detection method and system for the safety traceability of pork used in livestock breeding. By optimizing environmental control, introducing advanced spectral and non-contact detection equipment and multimodal data fusion analysis technology, the accuracy and efficiency of detection are significantly improved. It can quickly and non-destructively detect the comprehensive quality of pork, and provide technical support for achieving quality and safety traceability "from source to terminal". Summary of the Invention

[0005] In response to the above problems, the present invention provides a non-destructive quality detection method and system for safety traceability of pork for livestock breeding, so as to solve the problem that the existing technology is easily affected by ambient light and has difficulty in efficiently integrating spectral, texture and textural information to achieve accurate prediction.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: The present invention provides a nondestructive detection method for the safety traceability of pork used in animal husbandry, comprising the following steps: Step S1, controlling the temperature and humidity by a constant temperature and humidity device, equipping with a dynamic environment monitoring unit and a light source intensity detection module, regularly calibrating the Fourier transform infrared spectrometer and texture analyzer, verifying the equipment accuracy using standard samples, and installing an anti-vibration device and optical shielding materials; Wherein, step S1 further includes the following sub-steps: S1-1, uses a constant temperature and humidity device, controls temperature fluctuation within ±1°C, and humidity fluctuation within 5%; is equipped with a dynamic environmental monitoring unit to record environmental parameters in real time; installs a light source intensity detection module to prevent external optical interference from affecting spectral data; uses a multi-layer shielding device to reduce interference from dust, vibration, and ambient light; S1-2. Regularly calibrate the wavelength and intensity of the Fourier transform infrared spectrometer (FTIR), and use standardized spectral reference samples to verify the accuracy of the spectrometer; use texture standard samples to calibrate the pressure and displacement parameters of the texture analyzer to reduce data deviations caused by equipment errors; regularly check the mechanical properties of the texture analyzer to ensure the stability of pressure loading and displacement measurement during the test process; add anti-vibration devices and optical shielding materials around the collection equipment to reduce interference from vibration and ambient light.

[0007] Step S2, using a Fourier transform infrared spectrometer to collect spectral data from the pork surface to analyze fatty acid composition, using a non-contact texture detection technology to measure texture parameters, and using a high-definition imaging device to obtain color and texture characteristics; Wherein, in step S2, the following sub-steps are also included: S2-1, using Fourier transform infrared spectrometer (FTIR), collect spectral data of pork surface, and determine the fatty acid composition of fresh pork based on absorbance, as shown in formula (1): Formula (1) in, is the absorbance, is the incident light intensity, is the intensity of transmitted light; S2-2, using non-contact texture detection technology, the hardness, elasticity and chewing texture parameters were measured by laser means, and the changes in mechanical behavior over time were recorded using dynamic mechanical analysis, as shown in formula (2): Formula (2) in, is the storage modulus, is stress, For strain; S2-3, using high-definition imaging equipment, the color and texture of the pork surface are analyzed. The color characteristics are described using parameters in the CIE Lab color space, and the texture characteristics are extracted using the gray-level co-occurrence matrix, as shown in Equations (3) to (5): Formula (3) Formula (4) Formula (5) in, is the contrast of the gray-level co-occurrence matrix, is the joint probability of gray levels i and j in the gray level co-occurrence matrix, is the uniformity of grayscale distribution, is the entropy value of the gray-level co-occurrence matrix, which describes the complexity of the gray-level distribution.

[0008] Step S3, denoising and standardizing the collected data, smoothing the spectral data using a filter, extracting key features through dimensionality reduction by principal component analysis, and generating a fusion feature matrix by integrating spectral, texture, and textural information using feature splicing technology; Wherein, in step S3, the following sub-steps are also included: S3-1, denoise and standardize the collected spectral, texture and color data, and use Savitzky-Golay filter to smooth the data, as shown in formula (6): Formula (6) in, is the smoothed data point, is the filter coefficient, is the window width; S3-2, principal component analysis (PCA) is used to reduce the dimension of variables and screen key features. The key variables are extracted using feature vectors, and the comprehensive analysis of spectral, texture and texture data is achieved through feature splicing, as shown in Equations (7) to (8): Formula (7) concat Formula (8) in, is the feature matrix after dimensionality reduction, is the original data matrix, is the eigenvector matrix, is the fused feature matrix, is the spectral characteristic, is the texture feature, For texture characteristics.

[0009] Step S4, screening spectral features that are highly correlated with texture parameters through correlation analysis, grouping the samples using the K-means algorithm, and verifying the sample classification using linear discriminant analysis to calculate the classification accuracy; Wherein, in step S4, the following sub-steps are also included: S4-1, constructing the spectral feature matrix and texture parameter matrix , calculate the correlation coefficient between the two, and based on the correlation analysis results, screen out the spectral features with the highest correlation with the texture parameters, as shown in formula (9): Formula (9) Among them, Cov is the covariance of spectral and texture parameters, Var ,Var is the variance of spectral and textural parameters; S4-2, classify and group the samples and evaluate the effects of different sources and processing methods on pork quality. The spectral, texture and image features are used as input, and the K-means algorithm is used to classify the samples into Classify different categories of samples using linear discriminant analysis (LDA) and calculate the classification accuracy, as shown in formula (10)-formula (11):

[0010] Formula (11) in, is the objective function, is the number of samples in the i-th class, For the No. samples, For the The centroid of the class, is the discriminant vector, is the bias term, is the sample feature vector.

[0011] Step S5, predicting pork quality based on partial least squares regression and support vector machine models, and evaluating the mean square error and coefficient of determination of the models in combination with external data sets; Wherein, in step S5, the following sub-steps are also included: S5-1, build an accurate prediction model, improve the prediction ability of pork quality through optimization algorithm, based on the partial least squares regression (PLS) model, introduce weight optimization, improve the sensitivity to specific variables, use support vector machine to deal with nonlinear relationships, and select radial basis kernel function (RBF), as shown in formula (12)-formula (13): Formula (12) Formula (13) in, is the prediction matrix, is the weighting matrix, is the latent variable matrix, is the regression loading matrix, is the error matrix, is the kernel function, is the kernel function parameter, is the sample vector; S5-2, calculate the mean square error of the model, introduce an independent external data set, calculate the coefficient of determination, and evaluate the universality of the model, as shown in Equations (14)-(15): Formula (14) Formula (15) in, is the mean square error, is the actual value, is the predicted value, is the sample size, is the coefficient of determination, is the mean of the actual values.

[0012] Step S6: Apply the model to actual scenarios, achieve real-time processing and visualization through online data collection, generate heat maps and radar maps to display pork quality information, and continuously expand the data sample library.

[0013] Wherein, in step S6, the following sub-steps are also included: S6-1, the model is applied to actual scenarios, using sensors and spectrometers to achieve online data collection and real-time processing, and provide intuitive result display, as shown in formula (16):

[0014] in, is the two-dimensional heat map value, is the kernel function, is the sample point, is the weight coefficient; outputs multi-dimensional data display forms of heat map and radar map to intuitively present the test results. The heat map shows the fatty acid distribution, and the radar map shows the comparison of texture parameters; S6-2, continue to collect actual test data, expand the data sample library to cover more varieties and environments, build a data storage matrix, optimize model parameters based on the loss function, regularly evaluate the robustness and adaptability of the model, and calculate a new determination coefficient, as shown in formula (17): Formula (17) in, is the loss value, is the actual value, is the predicted value, N is the number of samples, and the model is continuously improved based on the evaluation results.

[0015] The present invention provides a nondestructive detection system for the safety and traceability of pork used in livestock breeding, comprising: Environmental control module, data acquisition module, data processing module, model building and analysis module, data visualization and result display module, and data storage and optimization module; The environmental control module includes: a constant temperature and humidity device, a dynamic environment monitoring unit, a light source intensity detection unit, a multi-layer shielding device and an anti-vibration device; the environmental control module provides stable environmental conditions, controls temperature, humidity and light intensity fluctuations, monitors environmental parameters in real time, and shields against vibration and optical interference; The data acquisition module includes: a Fourier transform infrared spectrometer (FTIR), a non-contact texture detection device and a high-definition imaging device; the data acquisition module is used to collect multimodal data of pork, and the multimodal data includes: spectral data, texture parameters and image data; The data processing module includes: a data denoising and standardization unit and a feature extraction unit; the data processing module removes data noise, standardizes multimodal data, and extracts core features; The model building and analysis module includes: a spectrum and texture correlation analysis unit, a sample classification unit and a quality prediction unit; the model building and analysis module is used to analyze the correlation between spectrum, texture and texture, perform sample classification and quantitative prediction of pork quality; The data visualization and result display module includes: a heat map generation unit and a radar map generation unit; the data visualization and result display module intuitively displays the test results and supports real-time presentation of multi-dimensional quality information, which is convenient for users to quickly interpret; The data storage and optimization module mainly includes a data storage matrix; the data storage and optimization module is used to continuously collect detection data and expand the sample library, optimize model parameters based on new data, and improve the robustness and adaptability of the system.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention uses a Fourier transform infrared spectrometer, non-contact texture detection equipment and high-definition imaging equipment to achieve non-destructive testing of pork quality, and greatly improves detection efficiency through multimodal data fusion technology and automated analysis process.

[0017] Through comprehensive analysis of spectral, texture and textural characteristics, the present invention constructs a model based on partial least squares regression and support vector machine, which can accurately predict pork quality parameters. Combined with dynamic environmental monitoring and data calibration, it significantly reduces environmental interference and equipment errors, and improves the reliability of detection results.

[0018] The present invention can detect the chemical composition and physical properties of pork, and can also analyze its appearance characteristics. This comprehensive quality assessment method is applicable to pork samples of different types, sources and processing methods, and has strong versatility and practicality.

[0019] The present invention has data storage and model optimization functions, and can continuously improve the robustness and adaptability of the system by continuously expanding the data sample library and optimizing algorithm parameters to meet the needs of diverse application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. It is understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 is a flow chart of the method of the present invention; Figure 2 It is a system structure diagram of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the invention for which protection is claimed, but is merely for selected embodiments of the present invention.

[0023] Please refer to Figure 1-Figure 2 The present invention provides a method and system for nondestructive quality testing of pork for livestock farming, including the following steps: Step S1, controlling the temperature and humidity by a constant temperature and humidity device, equipping with a dynamic environment monitoring unit and a light source intensity detection module, regularly calibrating the Fourier transform infrared spectrometer and texture analyzer, verifying the equipment accuracy using standard samples, and installing an anti-vibration device and optical shielding materials; Wherein, step S1 further includes the following sub-steps: S1-1, uses a constant temperature and humidity device, controls temperature fluctuation within ±1°C, and humidity fluctuation within 5%; is equipped with a dynamic environmental monitoring unit to record environmental parameters in real time; installs a light source intensity detection module to prevent external optical interference from affecting spectral data; uses a multi-layer shielding device to reduce interference from dust, vibration, and ambient light; S1-2. Regularly calibrate the wavelength and intensity of the Fourier transform infrared spectrometer (FTIR), and use standardized spectral reference samples to verify the accuracy of the spectrometer; use texture standard samples to calibrate the pressure and displacement parameters of the texture analyzer to reduce data deviations caused by equipment errors; regularly check the mechanical properties of the texture analyzer to ensure the stability of pressure loading and displacement measurement during the test process; add anti-vibration devices and optical shielding materials around the collection equipment to reduce interference from vibration and ambient light.

[0024] It should be noted that slight fluctuations in temperature may change the physical and chemical properties of pork samples, such as water volatilization and changes in fat distribution, which in turn affect the spectral absorption characteristics and texture parameters. Therefore, controlling the temperature within the range of ±1°C can minimize changes in the sample state and ensure the stability of the test results.

[0025] Changes in humidity can affect the moisture content of the pork surface, thereby affecting the absorption characteristics of spectral data (especially in the infrared region). At the same time, high humidity may cause condensation on the optical components of the instrument, and low humidity may cause the sample surface to dry out. Limiting humidity fluctuations to within 5% can effectively avoid the above problems and ensure the consistency of the experimental environment.

[0026] Step S2, using a Fourier transform infrared spectrometer to collect spectral data from the pork surface to analyze fatty acid composition, using a non-contact texture detection technology to measure texture parameters, and using a high-definition imaging device to obtain color and texture characteristics; Wherein, in step S2, the following sub-steps are also included: S2-1, using Fourier transform infrared spectrometer (FTIR), collect spectral data of pork surface, and determine the fatty acid composition of fresh pork based on absorbance, as shown in formula (1): Formula (1) in, is the absorbance, is the incident light intensity, is the intensity of transmitted light; S2-2, using non-contact texture detection technology, the hardness, elasticity and chewing texture parameters were measured by laser means, and the changes in mechanical behavior over time were recorded using dynamic mechanical analysis, as shown in formula (2): Formula (2) in, is the storage modulus, is stress, For strain; S2-3, using high-definition imaging equipment, the color and texture of the pork surface are analyzed. The color characteristics are described using parameters in the CIE Lab color space, and the texture characteristics are extracted using the gray-level co-occurrence matrix, as shown in Equations (3) to (5): Formula (3) Formula (4) Formula (5) in, is the contrast of the gray-level co-occurrence matrix, is the joint probability of gray levels i and j in the gray level co-occurrence matrix, is the uniformity of grayscale distribution, is the entropy value of the gray-level co-occurrence matrix, which describes the complexity of the gray-level distribution.

[0027] It should be noted that by using spectral data to analyze fatty acid composition, by analyzing the characteristic absorption peaks in the spectrum and combining it with a standardized spectral database, the fatty acid type, such as saturated fatty acids and unsaturated fatty acids, can be identified; the fatty acid content can be quantitatively analyzed by calibrating the linear relationship between the absorption peak intensity and concentration to determine the fatty acid content; to determine the quality of pork, the fatty acid composition is closely related to the taste, flavor and freshness of the pork, and different spectral characteristics reflect different qualities.

[0028] Step S3, denoising and standardizing the collected data, smoothing the spectral data using a filter, extracting key features through dimensionality reduction by principal component analysis, and generating a fusion feature matrix by integrating spectral, texture, and textural information using feature splicing technology; Wherein, in step S3, the following sub-steps are also included: S3-1, denoise and standardize the collected spectral, texture and color data, and use Savitzky-Golay filter to smooth the data, as shown in formula (6): Formula (6) in, is the smoothed data point, is the filter coefficient, is the window width; S3-2, principal component analysis (PCA) is used to reduce the dimension of variables and screen key features. The key variables are extracted using feature vectors, and the comprehensive analysis of spectral, texture and texture data is achieved through feature splicing, as shown in Equations (7) to (8): Formula (7) concat Formula (8) in, is the feature matrix after dimensionality reduction, is the original data matrix, is the eigenvector matrix, is the fused feature matrix, is the spectral characteristic, is the texture feature, For texture characteristics.

[0029] It should be noted that the goal of feature stitching is to fuse the following three types of data: spectral features, which are spectral data from a Fourier transform infrared spectrometer (FTIR) that reflect the chemical composition of the sample; texture features, which are obtained by non-contact texture detection equipment and include physical parameters such as hardness, elasticity, and chewiness; and texture features, which are color and texture data extracted by high-definition imaging equipment, such as grayscale co-occurrence matrix features (contrast, uniformity, entropy) and CIE Lab color space parameters.

[0030] The numerical ranges of different data sources may vary greatly. For example, spectral data are usually decimals, while texture parameters may be integers. To avoid unbalanced feature weights, each feature needs to be standardized to ensure that they have the same contribution during the splicing process.

[0031] Step S4, screening spectral features that are highly correlated with texture parameters through correlation analysis, grouping the samples using the K-means algorithm, and verifying the sample classification using linear discriminant analysis to calculate the classification accuracy; Wherein, in step S4, the following sub-steps are also included: S4-1, constructing the spectral feature matrix and texture parameter matrix , calculate the correlation coefficient between the two, and based on the correlation analysis results, screen out the spectral features with the highest correlation with the texture parameters, as shown in formula (9): Formula (9) Among them, Cov is the covariance of spectral and texture parameters, Var ,Var is the variance of spectral and textural parameters; S4-2, classify and group the samples and evaluate the effects of different sources and processing methods on pork quality. The spectral, texture and image features are used as input, and the K-means algorithm is used to classify the samples into Classify different categories of samples using linear discriminant analysis (LDA) and calculate the classification accuracy, as shown in formula (10)-formula (11):

[0032] Formula (11) in, is the objective function, is the number of samples in the i-th class, For the No. samples, For the The centroid of the class, is the discriminant vector, is the bias term, is the sample feature vector.

[0033] It should be noted that screening the spectral features with the highest correlation with texture parameters, such as hardness, elasticity, and chewiness, is a key step in data analysis. Its purpose is to extract the most explanatory features for texture parameters from high-dimensional spectral data, thereby improving the performance and computational efficiency of subsequent models; by analyzing the correlation between the characteristics of each band in the spectral data and the texture parameters, the band features with the highest correlation are selected for subsequent analysis. Correlation analysis can quantify the linear relationship between spectral features and texture parameters.

[0034] Step S5, predicting pork quality based on partial least squares regression and support vector machine models, and evaluating the mean square error and coefficient of determination of the models in combination with external data sets; Wherein, in step S5, the following sub-steps are also included: S5-1, build an accurate prediction model, improve the prediction ability of pork quality through optimization algorithm, based on the partial least squares regression (PLS) model, introduce weight optimization, improve the sensitivity to specific variables, use support vector machine to deal with nonlinear relationships, and select radial basis kernel function (RBF), as shown in formula (12)-formula (13): Formula (12) Formula (13) in, is the prediction matrix, is the weighting matrix, is the latent variable matrix, is the regression loading matrix, is the error matrix, is the kernel function, is the kernel function parameter, is the sample vector; S5-2, calculate the mean square error of the model, introduce an independent external data set, calculate the coefficient of determination, and evaluate the universality of the model, as shown in Equations (14)-(15): Formula (14) Formula (15) in, is the mean square error, is the actual value, is the predicted value, is the sample size, is the coefficient of determination, is the mean of the actual values.

[0035] It should be noted that the traditional partial least squares regression model is mainly used for regression analysis of a single data type, while the present invention extends the PLS model to the analysis of comprehensive data through the fusion and dimensionality reduction of multimodal features (spectral, texture, and texture). Through the comprehensive feature matrix generated by feature splicing, PLS can not only process high-dimensional spectral data, but also capture the complex relationship between spectral features and texture and texture features, thereby comprehensively reflecting the quality of pork.

[0036] In addition, based on the traditional PLS model, a weight optimization strategy was introduced: higher weights were assigned to features that were highly correlated with pork quality, thereby enhancing the impact of these variables on the model's prediction results. The weight optimization algorithm combined with the correlation analysis of texture parameters and spectral data made the model more sensitive to key features, which helped to improve prediction accuracy.

[0037] Spectral data are usually high-dimensional and have serious collinearity problems. The PLS model effectively solves the multicollinearity problem by constructing latent variables and projecting the original high-dimensional features into a low-dimensional space.

[0038] Step S6: Apply the model to actual scenarios, achieve real-time processing and visualization through online data collection, generate heat maps and radar maps to display pork quality information, and continuously expand the data sample library.

[0039] Wherein, in step S6, the following sub-steps are also included: S6-1, the model is applied to actual scenarios, using sensors and spectrometers to achieve online data collection and real-time processing, and provide intuitive result display, as shown in formula (16):

[0040] in, is the two-dimensional heat map value, is the kernel function, is the sample point, is the weight coefficient; outputs multi-dimensional data display forms of heat map and radar map to intuitively present the test results. The heat map shows the fatty acid distribution, and the radar map shows the comparison of texture parameters; S6-2, continue to collect actual test data, expand the data sample library to cover more varieties and environments, build a data storage matrix, optimize model parameters based on the loss function, regularly evaluate the robustness and adaptability of the model, and calculate a new determination coefficient, as shown in formula (17): Formula (17) in, is the loss value, is the actual value, is the predicted value, N is the number of samples, and the model is continuously improved based on the evaluation results.

[0041] It should be noted that the data storage matrix is ​​the core part used to save and manage detection data in the system. Its purpose is to achieve efficient storage, classification and call of multimodal detection data; the matrix contains the following levels of data: spectral data layer, which records the absorbance value of each sample at different wavelengths; texture data layer, which records the physical properties of the sample such as hardness, elasticity, and chewiness; texture data layer, including grayscale co-occurrence matrix features and color features.

[0042] Integrating spectral, textural, and texture features into a storage matrix enables unified management of multimodal data, facilitating subsequent comprehensive analysis. By calling historical data from the storage matrix, the sample library coverage is expanded, model parameters are optimized, and the robustness and adaptability of the model are improved. Support for real-time data acquisition and dynamic expansion ensures that the model can quickly adapt to new data, improving flexibility in practical applications.

[0043] A nondestructive testing system for the safety and traceability of pork used in livestock breeding, comprising: Environmental control module, data acquisition module, data processing module, model building and analysis module, data visualization and result display module, and data storage and optimization module; The environmental control module includes: a constant temperature and humidity device, a dynamic environment monitoring unit, a light source intensity detection unit, a multi-layer shielding device and an anti-vibration device; the environmental control module provides stable environmental conditions, controls temperature, humidity and light intensity fluctuations, monitors environmental parameters in real time, and shields against vibration and optical interference; The data acquisition module includes: a Fourier transform infrared spectrometer (FTIR), a non-contact texture detection device and a high-definition imaging device; the data acquisition module is used to collect multimodal data of pork, and the multimodal data includes: spectral data, texture parameters and image data; The data processing module includes: a data denoising and standardization unit and a feature extraction unit; the data processing module removes data noise, standardizes multimodal data, and extracts core features; The model building and analysis module includes: a spectrum and texture correlation analysis unit, a sample classification unit and a quality prediction unit; the model building and analysis module is used to analyze the correlation between spectrum, texture and texture, perform sample classification and quantitative prediction of pork quality; The data visualization and result display module includes: a heat map generation unit and a radar map generation unit; the data visualization and result display module intuitively displays the test results and supports real-time presentation of multi-dimensional quality information, which is convenient for users to quickly interpret; The data storage and optimization module mainly includes a data storage matrix; the data storage and optimization module is used to continuously collect detection data and expand the sample library, optimize model parameters based on new data, and improve the robustness and adaptability of the system.

[0044] It should be noted that this system is not only suitable for quality assessment in the slaughtering process, but can also serve as an important component of the quality and safety traceability system in the terminal link. When establishing the sample data storage matrix, the system introduced the "breeding batch label" field to identify the source information of individual pigs or batches, including: farm number, feed formula, breeding cycle, vaccination records and environmental parameters and other basic data, to achieve the integration of upstream and downstream data of pork quality and breeding process.

[0045] This label field can be automatically entered by connecting to the existing breeding management platform and IoT identification equipment, and stored in the database together with texture characteristics, spectral data and image features during the detection stage. By analyzing the correlation between quality parameters and breeding labels during model training, the impact of breeding factors on the quality of terminal meat products can be further explored, providing feedback and adjustment basis for pork quality control.

[0046] In addition, the system provides a standardized interface and can serve as the terminal quality node of the pork quality and safety traceability system, enabling the ability to trace back from terminal detection to the breeding source, thereby improving the information closed loop and transparency of the entire traceability chain.

[0047] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A nondestructive testing method for the safety and traceability of pork used in livestock breeding, characterized in that: The following steps are involved: Step S1, controlling the temperature and humidity by a constant temperature and humidity device, equipping with a dynamic environment monitoring unit and a light source intensity detection module, regularly calibrating the Fourier transform infrared spectrometer and texture analyzer, verifying the equipment accuracy using standard samples, and installing an anti-vibration device and optical shielding materials; Step S2, using a Fourier transform infrared spectrometer to collect spectral data from the pork surface to analyze fatty acid composition, using a non-contact texture detection technology to measure texture parameters, and using a high-definition imaging device to obtain color and texture characteristics; Step S3, denoising and standardizing the collected data, smoothing the spectral data using a filter, extracting key features through dimensionality reduction by principal component analysis, and generating a fusion feature matrix by integrating spectral, texture, and textural information using feature splicing technology; Step S4, screening spectral features that are highly correlated with texture parameters through correlation analysis, grouping the samples using the K-means algorithm, and verifying the sample classification using linear discriminant analysis to calculate the classification accuracy; Step S5, predicting pork quality based on partial least squares regression and support vector machine models, and evaluating the mean square error and coefficient of determination of the models in combination with external data sets; Step S6: Apply the model to actual scenarios, achieve real-time processing and visualization through online data collection, generate heat maps and radar maps to display pork quality information, and continuously expand the data sample library.

2. The nondestructive testing method for safety and traceability of pork for livestock breeding according to claim 1, characterized in that: Wherein, step S1 further includes the following sub-steps: S1-1, uses a constant temperature and humidity device, controls temperature fluctuation within ±1°C, and humidity fluctuation within 5%; is equipped with a dynamic environmental monitoring unit to record environmental parameters in real time; installs a light source intensity detection module to prevent external optical interference from affecting spectral data; uses a multi-layer shielding device to reduce interference from dust, vibration, and ambient light; S1-2. Regularly calibrate the wavelength and intensity of the Fourier transform infrared spectrometer (FTIR), and use standardized spectral reference samples to verify the accuracy of the spectrometer; use texture standard samples to calibrate the pressure and displacement parameters of the texture analyzer to reduce data deviations caused by equipment errors; regularly check the mechanical properties of the texture analyzer to ensure the stability of pressure loading and displacement measurement during the test process; add anti-vibration devices and optical shielding materials around the collection equipment to reduce interference from vibration and ambient light.

3. The nondestructive testing method for safety and traceability of pork for livestock breeding according to claim 1, characterized in that: Wherein, in step S2, the following sub-steps are also included: S2-1, using Fourier transform infrared spectrometer (FTIR), collect spectral data of pork surface, and determine the fatty acid composition of fresh pork based on absorbance, as shown in formula (1): Formula (1) in, is the absorbance, is the incident light intensity, is the intensity of transmitted light; S2-2, using non-contact texture detection technology, the hardness, elasticity and chewing texture parameters were measured by laser means, and the changes in mechanical behavior over time were recorded using dynamic mechanical analysis, as shown in formula (2): Formula (2) in, is the storage modulus, is stress, For strain; S2-3, using high-definition imaging equipment, the color and texture of the pork surface are analyzed. The color characteristics are described using parameters in the CIE Lab color space, and the texture characteristics are extracted using the gray-level co-occurrence matrix, as shown in Equations (3) to (5): Formula (3) Formula (4) Formula (5) in, is the contrast of the gray-level co-occurrence matrix, is the joint probability of gray levels i and j in the gray level co-occurrence matrix, is the uniformity of grayscale distribution, is the entropy value of the gray-level co-occurrence matrix, which describes the complexity of the gray-level distribution.

4. The nondestructive testing method for safety and traceability of pork for livestock breeding according to claim 1, characterized in that: Wherein, in step S3, the following sub-steps are also included: S3-1, denoise and standardize the collected spectral, texture and color data, and use Savitzky-Golay filter to smooth the data, as shown in formula (6): Formula (6) in, is the smoothed data point, is the filter coefficient, is the window width; S3-2, principal component analysis (PCA) is used to reduce the dimension of variables and screen key features. The key variables are extracted using feature vectors, and the comprehensive analysis of spectral, texture and texture data is achieved through feature splicing, as shown in Equations (7) to (8): Formula (7) concat Formula (8) in, is the feature matrix after dimensionality reduction, is the original data matrix, is the eigenvector matrix, is the fused feature matrix, is the spectral characteristic, is the texture feature, For texture characteristics.

5. The nondestructive testing method for safety and traceability of pork for livestock breeding according to claim 1, characterized in that: Wherein, in step S4, the following sub-steps are also included: S4-1, constructing the spectral feature matrix and texture parameter matrix , calculate the correlation coefficient between the two, and based on the correlation analysis results, screen out the spectral features with the highest correlation with the texture parameters, as shown in formula (9): Formula (9) Among them, Cov is the covariance of spectral and texture parameters, Var ,Var is the variance of spectral and textural parameters; S4-2, classify and group the samples and evaluate the effects of different sources and processing methods on pork quality. The spectral, texture and image features are used as input, and the K-means algorithm is used to classify the samples into Classify different categories of samples using linear discriminant analysis (LDA) and calculate the classification accuracy, as shown in formula (10)-formula (11): ; Formula (11); in, is the objective function, is the number of samples in the i-th class, For the No. samples, For the The centroid of the class, is the discriminant vector, is the bias term, is the sample feature vector.

6. The nondestructive testing method for safety and traceability of pork for livestock breeding according to claim 1, characterized in that: Wherein, in step S5, the following sub-steps are also included: S5-1, build an accurate prediction model, improve the prediction ability of pork quality through optimization algorithm, based on the partial least squares regression (PLS) model, introduce weight optimization, improve the sensitivity to specific variables, use support vector machine to deal with nonlinear relationships, and select radial basis kernel function (RBF), as shown in formula (12)-formula (13): Formula (12); Formula (13); in, is the prediction matrix, is the weighting matrix, is the latent variable matrix, is the regression loading matrix, is the error matrix, is the kernel function, is the kernel function parameter, is the sample vector; S5-2, calculate the mean square error of the model, introduce an independent external data set, calculate the coefficient of determination, and evaluate the universality of the model, as shown in Equations (14)-(15): Formula (14); Formula (15); in, is the mean square error, is the actual value, is the predicted value, is the sample size, is the coefficient of determination, is the mean of the actual values.

7. The nondestructive testing method for safety and traceability of pork for livestock breeding according to claim 1, characterized in that: Wherein, in step S6, the following sub-steps are also included: S6-1, the model is applied to actual scenarios, using sensors and spectrometers to achieve online data collection and real-time processing, and provide intuitive result display, as shown in formula (16): ; in, is the two-dimensional heat map value, is the kernel function, is the sample point, is the weight coefficient; outputs multi-dimensional data display forms of heat map and radar map to intuitively present the test results. The heat map shows the fatty acid distribution, and the radar map shows the comparison of texture parameters; S6-2, continue to collect actual test data, expand the data sample library to cover more varieties and environments, build a data storage matrix, optimize model parameters based on the loss function, regularly evaluate the robustness and adaptability of the model, and calculate a new determination coefficient, as shown in formula (17): Formula (17) in, is the loss value, is the actual value, is the predicted value, N is the number of samples, and the model is continuously improved based on the evaluation results.

8. A nondestructive testing system for the safety and quality traceability of pork for livestock breeding, applied to the testing method according to any one of claims 1 to 7, characterized in that: include: Environmental control module, data acquisition module, data processing module, model building and analysis module, data visualization and result display module, and data storage and optimization module; The environmental control module includes: a constant temperature and humidity device, a dynamic environment monitoring unit, a light source intensity detection unit, a multi-layer shielding device and an anti-vibration device; the environmental control module provides stable environmental conditions, controls temperature, humidity and light intensity fluctuations, monitors environmental parameters in real time, and shields against vibration and optical interference; The data acquisition module includes: a Fourier transform infrared spectrometer (FTIR), a non-contact texture detection device and a high-definition imaging device; the data acquisition module is used to collect multimodal data of pork, and the multimodal data includes: spectral data, texture parameters and image data; The data processing module includes: a data denoising and standardization unit and a feature extraction unit; the data processing module removes data noise, standardizes multimodal data, and extracts core features; The model building and analysis module includes: a spectrum and texture correlation analysis unit, a sample classification unit and a quality prediction unit; the model building and analysis module is used to analyze the correlation between spectrum, texture and texture, perform sample classification and quantitative prediction of pork quality; The data visualization and result display module includes: a heat map generation unit and a radar map generation unit; the data visualization and result display module intuitively displays the test results and supports real-time presentation of multi-dimensional quality information; The data storage and optimization module mainly includes a data storage matrix; the data storage and optimization module is used to continuously collect detection data and expand the sample library, and optimize model parameters based on new data.

Images