Online monitoring method for beef and mutton processing
By employing nonlinear cross-modal coupling mapping and scoring manifold compression mechanisms, the problems of insufficient real-time performance and accuracy in online monitoring during beef and mutton processing were solved, enabling high-frequency and continuous meat quality testing and improving food safety.
Patent Information
- Application Number
- CN202511796529.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-02
- Publication Date
- 2026-01-02
AI Technical Summary
Existing technologies for online monitoring in beef and mutton processing lack real-time performance, continuity, and accuracy, resulting in delayed meat quality testing, large human errors, and difficulty in achieving full coverage, thus creating potential food safety hazards.
A nonlinear cross-modal coupling mapping transformation and a scoring manifold compression mechanism are adopted, which combine spectral anomaly factor, trend coupling factor, biochemical state factor and temperature fluctuation suppression factor to map the three-modal data to a unified scoring space, and realize online monitoring through a scoring recursive smoothing iteration mechanism.
It improves the comprehensiveness and accuracy of meat quality testing, solves the data silo problem caused by modal differences, and realizes high-frequency, continuous and stable online evaluation of meat quality, which is suitable for high-frequency sampling environments.
Smart Images
Figure CN121256331A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of monitoring and analysis technology, and in particular to an online monitoring method for beef and mutton processing. Background Technology
[0002] With increasingly stringent food safety regulations and higher quality requirements for meat products, beef and mutton processing enterprises are facing greater technological demands for online monitoring of meat quality during production. Traditional meat quality testing relies heavily on manual sensory judgment or laboratory analysis after sampling, such as pH testing, moisture content determination, and microbial detection. While these methods theoretically possess a certain degree of accuracy, they generally suffer from issues like detection lag, slow response speed, large human error, and difficulty in providing real-time coverage of the entire production process. In actual production lines, due to the fast processing pace and high meat turnover rate, offline testing methods struggle to comprehensively evaluate every piece of meat. This often results in substandard meat flowing into downstream processes or mixing with qualified products, leading to inconsistent product quality and posing potential food safety risks.
[0003] In recent years, the rapid development of near-infrared spectroscopy, electronic nose sensing, and micro-electrochemical sensors has made it possible to obtain meat quality information from different modalities (spectral, gas, and electrochemical). However, the data collected by these sensors are often structurally complex, inconsistent in dimensions, and exhibit significant differences in sampling frequency, data units, and variation patterns. For example, spectral data presents a continuous multi-channel curve structure, gas data is represented by discrete response voltages, while pH and temperature are low-frequency scalar signals. These three types of data are difficult to process uniformly in data fusion, trend analysis, and quality assessment. Furthermore, existing scoring systems based on weighted averages, fuzzy rules, or simple regression models generally suffer from sensitivity to abnormal fluctuations, unstable scoring results, and a lack of physical interpretation capabilities, limiting their practical deployment on meat processing production lines.
[0004] In summary, the aforementioned technologies still have technical limitations in online monitoring of beef and mutton processing, including insufficient real-time performance, continuity, and accuracy. Summary of the Invention
[0005] This invention provides an online monitoring method for beef and mutton processing to solve the technical problems of insufficient real-time performance, continuity, and accuracy in online monitoring of beef and mutton processing.
[0006] The present invention provides an online monitoring method for beef and mutton processing, which specifically includes the following technical solutions: A method for online monitoring of beef and mutton processing includes the following steps: S1. Obtain three-modal data of beef and mutton processing stage and preprocess them to obtain preprocessed three-modal data; introduce nonlinear cross-modal coupling mapping transformation, combine spectral anomaly factor, trend coupling factor, biochemical state factor and temperature fluctuation suppression factor, map the preprocessed three-modal data to a unified scoring space to obtain scoring latent variables; S2. A scoring manifold compression mechanism is introduced to compress the latent variables of the scoring and obtain the compressed scoring value; a scoring recursive smoothing iteration mechanism is introduced to iterate the compressed scoring value and obtain the final scoring result, thereby realizing online monitoring.
[0007] Preferably, the three-modal data of the beef and mutton processing stage include: Spectral signal data, gas detection data, and electrochemical parameter data, including real-time pH value of meat and meat contact temperature.
[0008] Preferably, the specific construction methods of the spectral anomaly factor, trend coupling factor, biochemical state factor, and temperature fluctuation suppression factor are as follows: The spectral anomaly factor was constructed based on the preprocessed spectral signal data; the trend coupling factor was obtained by nonlinearly mapping the preprocessed spectral signal data and the preprocessed gas detection data to an angle; the biochemical state factor was constructed based on the real-time pH value of the preprocessed meat and combined with the pH deviation weight control coefficient; and the temperature fluctuation suppression factor was constructed based on the contact temperature of the preprocessed meat and the preprocessed gas detection data, combined with the temperature regulation term.
[0009] Preferably, the implementation process of the scoring manifold compression mechanism includes: By activating the latent rating variables exponentially and combining them with a nonlinear suppression term, the latent rating variables are compressed to obtain compressed rating values.
[0010] Preferably, the nonlinear suppression term is constructed in the following way: The nonlinear suppression term is constructed based on an electrochemical-environment coupled sinusoidal perturbation term and a gas response suppression term. The electrochemical-environment coupled sinusoidal perturbation term is based on the real-time pH value of the pretreated meat and the contact temperature of the pretreated meat, and a periodic perturbation is introduced to generate it. The gas response suppression term is generated based on the gas detection data after pretreatment, combined with an odor burst intensity suppression factor.
[0011] Preferably, the implementation process of the scoring recursive smooth iteration mechanism includes: Using the compressed score value as the initial score value, and based on the real-time pH value and contact temperature of the pretreated meat, periodic biochemical state regulation term and temperature periodic perturbation term are constructed respectively. Combined with the elastic activation coefficient of temperature-induced fluctuation and the local smoothing adjustment factor, the score value is iteratively processed.
[0012] Preferably, after the iteration is completed, the score obtained when the convergence condition is met is taken as the final score result, and the detection result is obtained based on the final score result.
[0013] The beneficial effects of the technical solution of the present invention are: 1. By introducing a nonlinear cross-modal coupling mapping, and constructing physically meaningful factors (spectral anomaly factor, trend coupling factor, biochemical state factor, and temperature fluctuation suppression factor), not only is the information independence of each mode in quality judgment preserved, but the structural coupling relationship between modes is also constructed, thus completing the "interpretable mapping" of originally "incomparable" modes within a unified scoring space. This process overcomes the data silos and information fragmentation problems caused by modal differences in traditional meat quality testing methods, improving the comprehensiveness and accuracy of quality perception.
[0014] 2. A mathematical mechanism was established using a nested, multi-layered nonlinear function structure to compress the state expression of beef and mutton at a certain processing stage into a single latent scoring variable. This latent scoring variable not only integrates the effects of multiple factors such as channel fluctuations, trend consistency, and environmental disturbances, but also embodies a causal chain through its formula structure. For example, increased temperature leads to enhanced gas activity, which in turn affects the intensity of odor response, ultimately causing scoring inhibition.
[0015] 3. The introduced scoring manifold compression mechanism effectively solves the problems of scale expansion and convergence instability of latent scoring variables under extreme conditions through nested high-order functions and perturbation term compensation in structural design. In particular, the pH-T coupling effect of meat products is embedded into the denominator structure in the form of a sine function perturbation, realizing the mathematical simulation of the phenomenon of "periodic increase in microbial activity" and demonstrating the ability to model the real spoilage process of meat products. The gas response inhibition term effectively curbs the sensing allergy problem of electronic nose in high-concentration volatile environment, making the overall scoring structure more stable and significantly enhancing the anti-perturbation ability, which is suitable for high-frequency and continuous sampling environment. Attached Figure Description
[0016] Figure 1 This is a flowchart of an online monitoring method for beef and mutton processing according to the present invention. Detailed Implementation
[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0019] The following description, in conjunction with the accompanying drawings, details the specific scheme of the online monitoring method for beef and mutton processing provided by this invention.
[0020] See attached document Figure 1 The diagram illustrates a flowchart of an online monitoring method for beef and mutton processing according to an embodiment of the present invention. The method includes the following steps: S1. Obtain three-modal data of beef and mutton processing stage and preprocess them to obtain preprocessed three-modal data; introduce nonlinear cross-modal coupling mapping transformation, combine spectral anomaly factor, trend coupling factor, biochemical state factor and temperature fluctuation suppression factor, map the preprocessed three-modal data to a unified scoring space to obtain scoring latent variables; After beef and mutton enter the processing production line, the [number]th [period] Each processing stage (such as cutting, deboning, and removing parts) utilizes a near-infrared high-frequency spectrometer, an electronic nose gas-sensitive array, a pH sensor, and a temperature sensor to acquire three-modal data, including spectral signal data, gas detection data, and electrochemical parameter data. The spectral signal data consists of the reflectance of the meat piece at different wavelengths (e.g., 760nm~1100nm), reflecting the meat's chemical composition. The electronic nose gas-sensitive array comprises a combination of various gas-sensitive materials, such as metal oxides and conductive polymers; the gas detection data obtained through this array reflects the type and concentration of volatile organic compounds. The electrochemical parameter data includes the real-time pH value and contact temperature of the meat, reflecting its biochemical state. Furthermore, since the physical meaning, sampling frequency, and unit dimensions of spectral signal data, gas detection data, and electrochemical parameter data are inconsistent, they cannot be directly used for calculation or comparison. Therefore, it is necessary to preprocess the above-mentioned spectral signal data, gas detection data, and electrochemical parameter data to obtain preprocessed three-mode data. The preprocessing process, such as noise reduction, cleaning, synchronization, standardization, and normalization, uses methods that are well known to those skilled in the art and will not be described in detail here. Furthermore, a nonlinear cross-modal coupling mapping transformation is introduced to map the preprocessed three-modal data to a unified scoring space, yielding the latent scoring variables. During this transformation, spectral anomaly factors, trend coupling factors, biochemical state factors, and temperature fluctuation suppression factors are introduced to obtain the state expression within the unified scoring space. The specific implementation formula is as follows: , in, It is the first The state representation of each processing stage in a unified scoring space, i.e., the latent scoring variables; It is the total number of spectral signal data; This represents the total number of gas detection data. It is the first one after preprocessing Spectral signal data for each channel (i.e., band); It is the average value of the preprocessed spectral signal data; It is the variance of the preprocessed spectral signal data; It is a zero-prevention term, a numerically stable constant, and can take values of... ; It is the first one after preprocessing Individual gas detection data; It is the real-time pH value of the pre-treated meat; It is the standardized ideal pH reference value for meat products, used to indicate freshness, and is set to 0.45; This is the pH deviation weighting control coefficient, obtained based on the sensory score regression relationship between meat quality and pH value. The reference value range is... ; It refers to the contact temperature of the pre-treated meat products; This is the standardized ideal cooling temperature setpoint, taken as 0.4667. This is the temperature-based moderating factor for the score, obtained based on the microbial growth rate curve fitting method, with a reference range of values. The microbial growth rate curve fitting method is a well-known technique in the art and will not be described in detail here. It is a spectral anomaly factor used to measure whether the current band deviates from the average state. If a certain channel deviates greatly, the channel has a higher weight in the score. Its goal is to discover "specific response channels" from the spectrum and highlight possible abnormal peaks. It is a trend coupling factor used to perform nonlinear ratio angle mapping between preprocessed spectral signal data and preprocessed gas detection data, with the aim of determining whether the changing trends of spectral signal data and gas detection data are synchronized. It is a biochemical state factor whose value decreases as the real-time pH value of meat deviates from the ideal state. It is used to characterize the possible spoilage or abnormal fermentation process of meat. It is a temperature fluctuation suppression factor used to describe potential risks caused by temperature. Used to emphasize the effect of increased spoilage at high temperatures; Logarithmic compression is performed on the preprocessed gas detection data to avoid unstable behavior such as abnormal amplification of the electronic nose output at high concentrations; This index represents the odor release potential under temperature-driven conditions, used to reflect whether abnormal odor responses are synchronized at high temperatures. This is a temperature regulation item. When the contact temperature of the pretreated meat is higher than the standardized ideal cooling temperature setting, the value of the temperature regulation item approaches 1, reflecting the risk of quality deterioration caused by the increase in contact temperature.
[0021] The above process will be the first The preprocessed trimodal data from each processing stage is mapped to a state representation in a unified scoring space. Latent variables, or scoring variables, are used to describe the implicit state of the overall quality indicators of meat products at present, and are the initial activation variables of the scoring system.
[0022] S2. A scoring manifold compression mechanism is introduced to compress the latent variables of the scoring and obtain the compressed scoring value; a scoring recursive smoothing iteration mechanism is introduced to iterate the compressed scoring value and obtain the final scoring result, thereby realizing online monitoring.
[0023] To avoid the problem of nonlinear scale expansion in latent rating variables, a rating manifold compression mechanism is introduced. This mechanism, based on multivariate modal cooperative perturbation modeling theory and information squeezing coding model, projects latent rating variables onto a nonlinear nested rating space. Through multi-nested compression of the rating scale, extreme values converge, the median region is enhanced, and the compressed rating value is finally obtained, achieving stability of the rating structure. The specific formula is as follows: , in, No. The compressed score value for each processing stage; It is the scoring sensitivity amplification factor, determined by the Platt calibration method, with a reference value range of [value missing]. The Platt calibration method is a well-known technique in the art and will not be described in detail here. This is an electrochemical interference modulator used to control the downregulation weight of pH and meat contact temperature on the score compression value. It is extracted based on the pH-T sensitivity coefficient of the Gompertz growth model, with a reference value range of [missing value]. The method for extracting the pH-T sensitivity coefficient of the Gompertz growth model is a well-known technique in the art and will not be described in detail here. This is an odor burst intensity suppression factor, used to measure the ability of odor response intensity to suppress score shift. It is obtained using existing GC-MS+E-nose joint detection technology, and the reference value range is [value missing]. ; It is an exponential activation term that can enhance the discriminative power of latent rating variables by exponential activation. It is an electrochemical-environment coupled sinusoidal perturbation term, which simulates the potential interference of microbial activity on the scoring system by introducing periodic perturbations; It is a gas response inhibition term that describes the saturation effect on persistent odor stimulation; It is a non-linear suppression term used to avoid score explosion and to build a stable control base.
[0024] To avoid rating compression To address the instability caused by the failure of trimodal data, a recursive smoothing iteration mechanism is introduced to construct a dynamic convergence process for the score values. This mechanism enables self-correction of outlier data disturbances, yielding the final score result. An initial score value is defined. Then the following recursion is executed: , in, It is the first The score value of the next iteration; It is the first The score value of the next iteration; This is a local smoothing adjustment factor used to control the trend of score changes. It is determined based on a fuzzy scoring mechanism, and its reference value range is [range to be filled in]. The fuzzy scoring mechanism is a technical means well known to those skilled in the art, and will not be described in detail here; It is the elastic activation coefficient for temperature-induced fluctuations, obtained by least squares fitting based on preprocessed historical meat contact temperatures and historical ratings extracted from existing databases. The reference value range is [value missing]. The least squares method is a well-known technique in the art and will not be elaborated here. This is a temperature disturbance frequency control factor, obtained through a spectrum extraction method based on the contact temperature of the pretreated meat. The reference value range is... The spectrum extraction method is a well-known technique in the art and will not be described in detail here. It is a recovery enhancement item for low to medium scores, used to provide an upward recovery force when the score is at a low or middle value; It is a temperature periodic perturbation term, which can reflect periodic micro-perturbations or changes in biological activity caused by ambient temperature; It is a low-score suppression regulation term. When the score is extremely low, this term increases rapidly, thereby suppressing score growth in the denominator. It is a periodic biochemical state regulation term used to introduce a periodic correlation between the score and pH; It can simulate the evolutionary momentum of meat quality without external interference, and construct a rating term with self-driven (high quality maintenance), self-inhibition (prevention of extreme values), and external adaptability (temperature interference compensation). It is a nonlinear damping term used to prevent the score from oscillating violently or rising meaninglessly under abnormal conditions; The above formula is iterated repeatedly until the convergence condition is met: when The calculation stops when the time is right, where The convergence threshold is set to 0.05. After the iteration is complete, the score obtained when the convergence condition is met will be used as the score of the first iteration. Final scoring results for each processing stage ; will the first The final score of each processing stage and the threshold preset based on expert experience. , , , By comparing the results, the detection outcome is obtained: when At that time, the testing grade was Grade 1 (high quality), indicating excellent ingredients, high freshness, and stable pH; when At that time, the inspection level was Grade 2 (Good), indicating slight fluctuations but overall meeting the process requirements; when When the detection level is Level 3 and usable, it indicates that the indicator deviates slightly and the potential risk is low; when At that time, the detection level was a level four warning, indicating a significant deviation in indicators and a suspected downward trend in quality; when When the test result is level five (unqualified), it indicates that multiple indicators are abnormal and there is a risk of corruption or deterioration.
[0025] In summary, an online monitoring method for beef and mutton processing has been developed.
[0026] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.
[0027] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.
[0028] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A method for online monitoring of beef and mutton processing, characterized in that, Includes the following steps: S1. Obtain three-modal data of beef and mutton processing stage and preprocess them to obtain preprocessed three-modal data; introduce nonlinear cross-modal coupling mapping transformation, combine spectral anomaly factor, trend coupling factor, biochemical state factor and temperature fluctuation suppression factor, map the preprocessed three-modal data to a unified scoring space to obtain scoring latent variables; S2. A scoring manifold compression mechanism is introduced to compress the latent variables of the scoring and obtain the compressed scoring value; a scoring recursive smoothing iteration mechanism is introduced to iterate the compressed scoring value and obtain the final scoring result, thereby realizing online monitoring.
2. The online monitoring method for beef and mutton processing according to claim 1, characterized in that, The three-modal data for the beef and mutton processing stage include: Spectral signal data, gas detection data, and electrochemical parameter data, including real-time pH value of meat and meat contact temperature.
3. The online monitoring method for beef and mutton processing according to claim 2, characterized in that, The specific construction methods for the spectral anomaly factor, trend coupling factor, biochemical state factor, and temperature fluctuation suppression factor are as follows: The spectral anomaly factor is constructed based on the preprocessed spectral signal data; The trend coupling factor is obtained by nonlinearly mapping the preprocessed spectral signal data and the preprocessed gas detection data by angle. Biochemical state factors were constructed based on the real-time pH value of the pretreated meat and combined with pH deviation weighting control coefficients. The temperature fluctuation suppression factor is constructed based on the contact temperature of the pretreated meat and the gas detection data after pretreatment, combined with the temperature regulation term.
4. The online monitoring method for beef and mutton processing according to claim 3, characterized in that, The implementation process of the scoring manifold compression mechanism includes: By activating the latent rating variables exponentially and combining them with a nonlinear suppression term, the latent rating variables are compressed to obtain compressed rating values.
5. The online monitoring method for beef and mutton processing according to claim 4, characterized in that, The specific construction method of the nonlinear suppression term is as follows: The nonlinear suppression term is constructed based on an electrochemical-environment coupled sinusoidal perturbation term and a gas response suppression term. The electrochemical-environment coupled sinusoidal perturbation term is based on the real-time pH value of the pretreated meat and the contact temperature of the pretreated meat, and a periodic perturbation is introduced to generate it. The gas response suppression term is generated based on the gas detection data after pretreatment, combined with an odor burst intensity suppression factor.
6. The online monitoring method for beef and mutton processing according to claim 5, characterized in that, The implementation process of the scoring recursive smooth iteration mechanism includes: Using the compressed score value as the initial score value, and based on the real-time pH value and contact temperature of the pretreated meat, periodic biochemical state regulation term and temperature periodic perturbation term are constructed respectively. Combined with the elastic activation coefficient of temperature-induced fluctuation and the local smoothing adjustment factor, the score value is iteratively processed.
7. The online monitoring method for beef and mutton processing according to claim 6, characterized in that, After the iteration is completed, the score obtained when the convergence condition is met is taken as the final score result, and the detection result is obtained based on the final score result.