A flexible multi-spectral sensing pork cold chain quality dynamic monitoring system and method

Abstract

The present application belongs to the technical field of food safety nondestructive testing, and relates to a flexible multi-spectral sensing pork cold chain quality dynamic monitoring system and method, which comprises a flexible multi-spectral sensing unit, a data acquisition and wireless transmission module, a data processing and integrated learning module, a battery power supply module, a monitoring terminal and a cloud platform, the flexible multi-spectral sensing unit is adapted to the low-temperature and high-humidity environment of the cold chain and is in close contact with the surface of pork or the inside of pork tissue; the data acquisition and wireless transmission module samples multi-band spectral signals and sends them through WiFi wireless communication; the data processing and integrated learning module pre-processes the received multi-band spectral signals and predicts key indicators of pork quality by using a regression model algorithm; the cloud platform displays and stores the pork quality prediction results; the monitoring terminal communicates with the cloud platform through a network and displays the prediction results in the cloud platform. The system can cover the quality dynamic monitoring of the whole cold chain life cycle.

Description

Technical Field

[0001] This invention belongs to the field of non-destructive testing technology for food safety, and particularly relates to a flexible multispectral sensing system and method for dynamic monitoring of pork cold chain quality. Background Technology

[0002] As a major source of animal protein in the diet of residents, the quality stability and safety of pork are directly related to consumer health and the economic benefits of the meat industry. After slaughter, pork needs to go through a series of cold chain processes, including rapid cooling, aging, refrigerated transportation, and final sales. During this process, a series of complex physiological and biochemical changes occur within the muscle tissue, including pH changes caused by glycogenolysis, decreased tissue water retention capacity, and changes in amino acid content due to protein degradation. These factors collectively determine the tenderness, flavor, and shelf life of pork.

[0003] Currently, pork quality testing mainly relies on methods such as piercing pH measurement, moisture drying determination, and high-performance liquid chromatography (HPLC) analysis of amino acids. While these methods offer high analytical accuracy, they generally suffer from drawbacks such as damaging the integrity of the meat during the testing process, long testing cycles, inability to achieve continuous monitoring, and difficulty in adapting to actual slaughtering, processing, and cold chain logistics environments.

[0004] In recent years, visible / near-infrared spectroscopy has been gradually introduced into the field of meat quality testing due to its non-destructive and rapid characteristics. However, existing spectral detection systems mostly use rigid fiber optic probes or large benchtop devices. The rigidity of these probes makes it difficult to form stable optical coupling with the irregular, soft, and easily deformable surface or internal tissue of pork. In low-temperature and high-humidity environments, optical path drift and signal attenuation easily occur, limiting their practical application in slaughterhouses and cold chain environments. Furthermore, in terms of spectral data modeling, existing technologies mostly use single regression models, such as partial least squares regression or support vector regression. These models have limited prediction accuracy and generalization ability when dealing with highly nonlinear biological data such as pork quality, which exhibits significant individual differences, making it difficult to meet the needs of industrial-grade dynamic monitoring. Although flexible spectral sensors exist, none have been applied to pork cold chain scenarios. Moreover, existing flexible materials are prone to embrittlement at -20°C, leading to electrical connection breakage and preventing long-term stable monitoring. While existing ensemble learning technologies are applied to food testing, they are not combined with flexible multispectral sensors, nor are dynamic weight adjustment mechanisms designed for signal fluctuations in cold chain environments.

[0005] Existing monitoring technologies have significant shortcomings in practical cold chain applications. First, most existing spectral detection devices use rigid probes. However, pork, as a highly heterogeneous and elastic biological tissue, undergoes significant tissue shrinkage and deformation during aging and storage. This makes it difficult for rigid probes to achieve long-term, stable conformal contact with its irregular surface, and the resulting air gaps introduce severe diffuse reflection noise. Second, current quality assessments are often limited to the single aging stage after slaughter, lacking continuous monitoring capabilities across the entire chain from processing and cold chain transportation to final retail, resulting in a serious "monitoring gap." Finally, traditional machine learning models (such as partial least squares regression (PLSR), support vector machine regression (SVR), and extreme gradient boosting (XGBoost) models) are not robust enough to handle signal fluctuations in complex cold chain environments (such as delivery vibrations or temperature fluctuations), making it difficult to achieve high-precision dynamic quality prediction across individuals and stages. Summary of the Invention

[0006] To address the aforementioned technical problems in existing technologies, this invention provides a flexible multispectral sensing system and method for dynamic quality monitoring of pork in the cold chain. This system is adaptable to the complex biological interface of pork and provides a dynamic quality monitoring solution covering the entire cold chain lifecycle. It achieves conformal fitting between the sensor and the meat sample surface to improve signal acquisition quality; miniaturized integration allows the sensing unit to move with the meat throughout all stages of the cold chain; and an integrated learning model with deep generalization capabilities is constructed to achieve real-time and accurate inversion of physicochemical indicators such as pH, moisture, and amino acids under dynamic environmental interference.

[0007] To achieve the above objectives, the present invention employs the following technical solution: A flexible multispectral sensing system for dynamic monitoring of pork cold chain quality includes a flexible multispectral sensing unit, a data acquisition and wireless transmission module, a data processing and integrated learning module, a battery-powered module, a monitoring terminal, and a cloud platform. The data processing and integrated learning module is electrically connected to the flexible multispectral sensing unit, the data acquisition and wireless transmission module, the battery-powered module, and the cloud platform, respectively. The flexible multispectral sensing unit is adapted to the low temperature and high humidity environment of the cold chain and conformally fits to the surface of pork or the interior of pork tissue. The data acquisition and wireless transmission module samples the multi-band spectral signal and transmits it wirelessly via WiFi; the data processing and integrated learning module preprocesses the received multi-band spectral signal and uses a regression model algorithm to predict key indicators of pork quality; the key indicators of pork quality are at least one of pH value, moisture content, and amino acid content. The cloud platform displays and stores the pork quality prediction results; The monitoring terminal communicates with the cloud platform via the network and displays the prediction results from the cloud platform.

[0008] This invention also provides a method for dynamic monitoring of pork cold chain quality using flexible multispectral sensing, comprising: Multi-band spectral data and corresponding key quality indicators of pork were collected during the cold chain process at preset sampling intervals. The sampling intervals were adjusted according to the cold chain stage (high-frequency sampling during rapid cooling / acid removal stage, and low-frequency sampling during distribution / sales stage). The multi-band spectral data is sent to the data processing and integrated learning module; The model algorithm preprocesses the multi-band spectral data and establishes a pork quality prediction model based on the ensemble learning model algorithm. The preprocessing includes at least one of the following: standardization of the multi-band spectral data, construction of first-order / second-order difference features, and calculation of statistical features. The ensemble learning model is a stacked ensemble learning structure, which includes multiple regression model base learners and a fusion weight allocation module. Output the prediction results of key indicators of pork quality; The prediction results are sent to the monitoring terminal and cloud platform for storage and display, so as to realize dynamic monitoring of the trend of pork quality changes.

[0009] Compared with the prior art, the present invention has the following beneficial effects: This invention utilizes flexible multispectral sensors combined with deep ensemble learning algorithms to provide a system and method for in-situ, continuous, and dynamic monitoring of the evolution of physicochemical indicators in pork throughout the entire cold chain process (including rapid cooling, aging, cutting and packaging, refrigerated distribution, and terminal sales). The system synchronizes all chain data to the cloud via the MQTT asynchronous communication protocol. By constructing a pork quality prediction model based on ensemble learning, it establishes a mapping relationship between multi-band spectral data and pork quality indicators, comprehensively depicting the physicochemical evolution trajectory of pork from slaughter to retail. Compared to traditional single regression models, this model improves the accuracy and stability of pork quality indicator prediction, enabling real-time prediction of key indicators such as pH value, moisture content, and amino acid content, and achieving continuous dynamic monitoring of pork throughout the cold chain process. This method can acquire information on pork quality changes without damaging the meat structure, improving the real-time nature and reliability of quality monitoring during cold chain logistics, and providing technical support for meat quality safety supervision and cold chain logistics management. Using a PI / PET composite flexible substrate, the flexibility retention rate is ≥95% and the electrical connection stability is ≥99% within the cold chain temperature range of -20℃ to 4℃, solving the technical problem of embrittlement of existing flexible sensors at low temperatures. The stacked integrated learning structure, combined with dynamic weight adjustment based on the cold chain environment, improves the robustness to environmental interference by 40% compared to the traditional single model, achieving high-precision continuous monitoring across the cold chain stage, which is impossible with existing technologies. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the entire process of pork cold chain business in this invention; Figure 2 This is a schematic diagram of the flexible multispectral sensing system for dynamic monitoring of pork cold chain quality according to the present invention. Figure 3 This is a schematic diagram of the spectral acquisition principle of the flexible multispectral sensor in the dynamic monitoring system for pork cold chain quality of the present invention. Figure 4 This is a schematic diagram of the flexible multispectral sensor implantation in the flexible multispectral sensing dynamic monitoring system for pork cold chain quality of the present invention. Figure 5 This is a schematic diagram of the cloud platform monitoring interface of the flexible multispectral sensing dynamic monitoring system for pork cold chain quality of the present invention. Figure 6 This is a hardware circuit diagram of the flexible multispectral sensor in the dynamic monitoring system for pork cold chain quality of the present invention. Detailed Implementation

[0011] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0012] Example

[0013] like Figure 2 As shown, the present invention discloses a flexible multispectral sensing system for dynamic monitoring of pork cold chain quality, comprising a flexible multispectral sensing unit, a data acquisition and wireless transmission module, a data processing and integrated learning module, a battery power supply module, a monitoring terminal, and a cloud platform. The data processing and integrated learning module is electrically connected to the flexible multispectral sensing unit, the data acquisition and wireless transmission module, the battery power supply module, and the cloud platform, respectively. The flexible multispectral sensing unit is adapted to the low-temperature and high-humidity environment of the cold chain and conformally fits to the surface of pork or the interior of pork tissue. The data acquisition and wireless transmission module samples the multi-band spectral signals and transmits them wirelessly via WiFi to achieve stable data transmission under low-temperature and high-humidity conditions. The data processing and integrated learning module preprocesses the received multi-band spectral signals and uses a regression model algorithm to predict key indicators of pork quality. The key indicators of pork quality are at least one of pH value, moisture content, and amino acid content. A cloud platform is used to display and store the pork quality prediction results; The monitoring terminal communicates with the cloud platform via the network and displays the prediction results from the cloud platform.

[0014] This cold chain quality dynamic monitoring system is used to achieve dynamic quality monitoring of pork throughout the entire process from slaughter to final sale. The multi-band spectral signals of pork collected by the flexible multispectral sensing unit are used to construct a pork quality prediction model, and the mapping relationship between multispectral data and pork quality indicators is modeled and analyzed.

[0015] Optionally, the flexible multispectral sensing unit includes a flexible substrate with low-temperature resistance, a multi-band light source module, a spectral sensing module, and a flexible printed circuit. The multi-band spectral signals acquired by the flexible multispectral sensing unit are transmitted to the data acquisition and wireless transmission module via the flexible printed circuit.

[0016] Optionally, the flexible substrate is made of one or a composite polymer material of polyimide (PI) and polyethylene terephthalate (PET), and the flexible substrate has low-temperature resistance and high flexibility. The PI / PET polymer material of the flexible substrate not only achieves conformal bonding, but also maintains stable electrical connections in a low-temperature environment of -20℃ to 4℃.

[0017] Optionally, one or more flexible multispectral sensing units are placed on the pork chunks after slaughter, and are either attached to the surface of the pork or implanted inside the pork muscle tissue during the rapid cooling and aging stages. Figure 4 (This is used to collect multi-band spectral signals of pork during the cold chain process in real time.)

[0018] Optionally, the data processing and integrated learning module uses an embedded processing unit, specifically an ESP32 control module.

[0019] Optionally, the multi-band light source module covers the visible light band (450nm~760nm) to the near-infrared band (780nm~880nm); specifically... Figure 5 The 450nm and 730nm wavelengths are the visible spectrum bands.

[0020] Optionally, the flexible multispectral sensing unit is externally provided with a high-transmittance elastic encapsulation layer. Preferably, the elastic encapsulation layer is made of PDMS Sylgard with a transmittance ≥90% and a wavelength range of 450nm~880nm.

[0021] Optionally, the data acquisition and wireless transmission module samples and digitizes the multi-band spectral signals, then transmits them to the data processing and integrated learning module via a wireless communication module. The wireless communication module uses a WiFi wireless communication network and transmits data via the MQTT communication protocol. It can operate stably in low-temperature and high-humidity environments.

[0022] like Figure 5As shown, the monitoring terminal and cloud platform are used to uniformly store, display, and manage the quality prediction results at different stages of the cold chain, thereby realizing continuous dynamic monitoring and traceability of pork quality throughout the entire cold chain process. The monitoring terminal communicates with the cloud platform via a wireless communication network to view and visualize the quality prediction results on the cloud platform, specifically via a mobile phone, tablet, or other data terminal. The cloud platform is used for data storage, model computation, and data management. The cloud platform includes a data storage module, a data processing and integrated learning module, and a visualization management module. Specifically, the data processing and integrated learning module runs an integrated learning model to predict and analyze pork quality indicators; the data storage module stores the collected multi-band spectral data and quality prediction results; and the visualization management module graphically displays and manages the quality change trends of pork at each stage of the cold chain.

[0023] (1) Fabrication of flexible multispectral sensing unit like Figure 3 As shown, the flexible substrate is made of a polymer material with good flexibility and low-temperature resistance. On the flexible substrate, the aforementioned flexible printed circuit is fabricated into multi-channel conductive lines through laser etching or printing to form a flexible printed circuit used to realize the electrical connection between the multi-band light source module and the spectral sensing module; specifically, the flexible printed circuit is etched using a CO2 laser with a power of 10~20W and an etching depth of 5~10μm. Figure 6 As shown, the flexible substrate supports the multi-band light source module and the spectral sensing module, allowing the overall structure to maintain stable electrical connections even when bent or attached. The multi-band light source module is located on one side of the flexible substrate and is used to emit multi-band light signals to the pork tissue, covering the visible and near-infrared bands. The spectral sensing module is located close to the multi-band light source module and is used to receive the light signals scattered or reflected by the pork tissue and convert the light signals into electrical signals.

[0024] After the optoelectronic components are installed, a highly transparent elastic encapsulation layer is applied to the outside of the flexible multispectral sensing unit to improve the stability of the sensing unit in low temperature and high humidity environments and reduce the impact of external moisture or pollutants on the optical signal.

[0025] (2) Attachment of flexible multispectral sensing unit After the pork is slaughtered and initially processed, the prepared flexible multispectral sensing unit is attached to a designated location on the pork surface. The flexible substrate has good flexibility, allowing the flexible multispectral sensing unit to adhere to the pork surface, thereby reducing interference from air gaps on the optical signal.

[0026] During the attachment process, the multi-band light source module and the spectral sensing module are positioned facing the pork tissue to ensure that the light signal can enter the pork tissue and be effectively collected.

[0027] (3) Acquisition and transportation of multi-band spectral data After pork enters the cold chain storage or transportation stage, the flexible multispectral sensing unit operates according to a preset time interval.

[0028] The flexible multispectral sensing unit's multi-band light source module sequentially emits light signals of different bands, while the spectral sensing module simultaneously acquires the light signals reflected by the pork tissue. The spectral sensing module receives the spectral response signals of multiple bands after reflection or scattering by the pork tissue and converts these light signals into electrical signals of corresponding bands, forming a multi-band spectral signal. The multi-band spectral signal acquired by the flexible multispectral sensing unit is transmitted to the data acquisition and wireless transmission module via a flexible printed circuit. The data acquisition and wireless transmission module samples and digitizes the multi-band spectral signal, then sends it to the data processing and integrated learning module via a wireless communication module. The wireless communication module uses WiFi communication and transmits data via the MQTT communication protocol. It can operate stably in low-temperature and high-humidity environments.

[0029] (4) Prediction and display of quality indicators After receiving multi-band spectral data, the data processing and integrated learning module preprocesses the multi-band spectral data and then inputs the processed multi-band spectral data into the model for analysis to obtain the final quality prediction result. The model predicts key indicators of pork quality based on base learners, and obtains the final quality prediction result by fusing the output results of each base learner. The prediction results are sent to the monitoring terminal and cloud platform for storage and display, so as to realize dynamic monitoring of the trend of pork quality changes.

[0030] Example 2

[0031] (1) Fabrication of flexible multispectral sensing unit The structure of the flexible multispectral sensing unit in this embodiment is basically the same as that in Embodiment 1, including a flexible substrate, a multi-band light source module, a spectral sensing module, a flexible printed circuit, and an elastic encapsulation layer. The difference is that the overall size of the flexible multispectral sensing unit is designed to be embedded in pork muscle tissue to ensure its stable existence within the tissue and to not affect the normal cold chain processing procedure.

[0032] (2) Embedding of flexible multispectral sensing units like Figure 4 As shown, during the rapid cooling or acid removal stage after pork slaughter, a containment space is formed in the pork muscle tissue, and a flexible multispectral sensing unit is placed in the containment space.

[0033] Because of its excellent flexibility, the flexible multispectral sensing unit can bend with the contraction or deformation of muscle tissue, thereby maintaining stable contact with the surrounding tissue and ensuring the continuity of the optical coupling state.

[0034] (3) Spectral acquisition during the entire cold chain process Throughout the entire cold chain process of pork, including rapid cooling, aging, refrigeration, transportation, and terminal sales, the flexible multispectral sensing unit continuously collects multi-band spectral data of the internal tissues of pork at preset time intervals.

[0035] The acquired multi-band spectral data is processed by the data acquisition and wireless transmission module and then transmitted to the data processing and integrated learning module via wireless communication.

[0036] (4) Continuous quality monitoring and data management The data processing and ensemble learning module processes continuously acquired multi-band spectral data, and the model predicts key indicators of pork quality.

[0037] This invention also provides a method for dynamic monitoring of pork cold chain quality using flexible multispectral sensing, comprising: 1) Collect multi-band spectral data and corresponding key quality indicators of pork during the cold chain process according to the preset sampling interval. The sampling interval is adjusted according to the cold chain stage. 2) Send the multi-band spectral data to the data processing and integrated learning module; 3) The model algorithm preprocesses the multi-band spectral data and establishes a pork quality prediction model based on the ensemble learning model algorithm. The preprocessing includes at least one of the following: standardization of the multi-band spectral data, construction of first-order / second-order difference features, and calculation of statistical features. The ensemble learning model is a stacked ensemble learning structure, which includes multiple regression model base learners and a fusion weight allocation module. The preprocessing is used to eliminate the interference caused by environmental factors and individual differences during the acquisition of multi-band spectral data, improve the stability and consistency of spectral data, and enhance the accuracy and reliability of subsequent model predictions of pork quality indicators. 4) Output the prediction results of key pork quality indicators; 5) The prediction results are sent to the monitoring terminal and cloud platform for storage and display, so as to realize dynamic monitoring of the trend of pork quality changes.

[0038] High-frequency sampling is used during the rapid cooling / acid removal stage, and low-frequency sampling is used during the delivery / sales stage; specifically, the sampling interval is set to 5~10 minutes / time during the rapid cooling / acid removal stage and 30~60 minutes / time during the delivery / sales stage.

[0039] Optionally, the flexible multispectral sensing unit's multi-band light source module sequentially emits light signals of different bands, and the spectral sensing module synchronously collects the light signals reflected by the pork tissue. The spectral sensing module receives spectral response signals of multiple bands after reflection or scattering by pork tissue, and converts the optical signals into electrical signals of corresponding bands to form multi-band spectral signals. The multi-band spectral signals are transmitted to the data acquisition and wireless transmission module via a flexible printed circuit. After sampling and digitizing the multi-band spectral signals, the data acquisition and wireless transmission module sends them to the data processing and integrated learning module via a wireless communication module. After receiving multi-band spectral data, the data processing and ensemble learning module preprocesses the data during the pork quality prediction model's operation. The preprocessed data is then input into the model for analysis to obtain the final prediction results for key pork quality indicators. Specifically, the wireless communication module uses WiFi communication and transmits data via the MQTT communication protocol.

[0040] Optionally, the multi-band spectral data is acquired through a flexible multispectral sensing unit. Spectral data enhancement processing: when signal fluctuations are caused by vibrations during cold chain transportation, Gaussian perturbation is used for data enhancement; when spectral baseline drift is caused by temperature fluctuations in the cold chain, random scaling is used for data enhancement.

[0041] Optionally, multi-band spectral data can be collected at preset intervals throughout the entire pork cold chain process. After the pork enters the cold chain storage or transportation stage, the flexible multi-spectral sensing unit operates at preset time intervals.

[0042] Optionally, the ensemble learning model algorithm includes multiple base learners and fuses the outputs of each base learner. It predicts key indicators of pork quality based on the base learners and fuses the outputs of each base learner through a stacked ensemble learning structure to obtain the predicted results of the key indicators of pork quality. The model algorithm is used to train and learn from multi-band spectral data to construct a pork quality prediction model. In actual operation, the processed multi-band spectral data is input into the prediction model for analysis to obtain the predicted results of the key indicators of pork quality.

[0043] Optionally, the multi-band spectral signals acquired by the flexible multispectral sensing unit are used to construct a pork quality prediction model, and to model and analyze the mapping relationship between multi-band spectral data and pork quality indicators. The original multi-band spectral data acquired by the flexible multispectral sensing unit is as follows:

[0044] Where N represents the number of samples, M represents the number of spectral bands or feature dimensions, and x i,jThis represents the spectral response value of the i-th sample in the j-th band.

[0045] To eliminate the impact of dimensional differences between different bands on model training, the original multi-band spectral data is standardized, and its mathematical expression is as follows:

[0046] Where, μ j and σ j denoted as the mean and standard deviation of the spectral data for the j-th band in the training samples, respectively.

[0047] To improve the model's adaptability to different pork samples and changes in the cold chain environment, multi-band spectral data were augmented during the model training phase. The mathematical expression is as follows:

[0048] Where, ε i,j This represents a random perturbation term that follows a Gaussian distribution.

[0049] In another embodiment, the spectral data is randomly scaled, and its mathematical expression is:

[0050] Wherein, the random scaling factor α i ∈[0.9,1.1].

[0051] Based on the standardized multi-band spectral data, spectral features were further constructed to enhance the model's ability to characterize changes in pork quality.

[0052] In one embodiment, the first-order difference feature is calculated for adjacent spectral bands, and its mathematical expression is:

[0053] In another embodiment, the second-order difference characteristics of adjacent spectral bands are calculated, and their mathematical expression is as follows:

[0054] Furthermore, statistical characteristics, including mean, standard deviation, and range, can be calculated based on the spectral vector of a single sample, with the following mathematical expressions:

[0055]

[0056]

[0057] Based on the constructed multispectral features, multiple regression models are established as base learners.

[0058] The mathematical expression for base learner 1 is as follows: Symbols are not defined, such as E and f in base learner 1, and l and Ω in base learner 3.

[0059] The prediction results are obtained through coefficient matrix B:

[0060] Where: T is the latent variable matrix, B is the regression coefficient matrix, E is the residual matrix, and f is the model fitting error.

[0061] The prediction function for base learner 2 is:

[0062] The objective function to be optimized is:

[0063] The objective function for base learner 3 is:

[0064] The regular expression term is:

[0065] Where l is the loss function; Ω is the regularization term used to prevent the model from overfitting.

[0066] To synthesize the prediction results of each base learner, a stacked ensemble learning structure is used to fuse the outputs of multiple base learners. Let the prediction results of the i-th sample obtained by the K base learners be as follows:

[0067] The prediction results are used to construct a fusion feature vector:

[0068] The prediction results of the fusion model are expressed as follows:

[0069] Among them, w k denoted as , where b represents the fusion weights corresponding to each base learner, and b is the bias term.

[0070] To improve the model's stability and generalization ability, regularization constraints are introduced during model training, and the optimization objective function is:

[0071] Wherein, the regularization coefficient λ∈[0.001,0.01].

[0072] The above-mentioned ensemble learning model can be used to predict key indicators of pork quality, enabling dynamic monitoring of pork quality during the cold chain process.

[0073] This invention also provides a specific implementation process for a flexible multispectral sensing method for dynamic monitoring of pork cold chain quality: like Figure 1 As shown, rapid cooling and acid removal dynamic tracking: A flexible multispectral sensing unit is directly attached to or inserted into a longitudinal fiber-direction incision made in the pork sample after slaughter. The PDMS-encapsulated flexible multispectral sensing unit is then implanted, and a high-frequency sampling mode is initiated under low-temperature air-cooling conditions of -20℃ and 0–4℃. This stage monitors the biochemical dynamics of the tissue transition from a living state to a meat-like state. Due to the interruption of blood oxygen supply caused by slaughter, similar anaerobic glycogenolysis occurs within the cells. The model algorithm analyzes the coupling rate in the visible light band to map the pH value in real time as it decreases from approximately 7.0 to a final value of 5.5–5.7. Simultaneously, myofibrillar protein denaturation induced by tissue acidification leads to a decrease in water-holding capacity (WHC), and the sensor captures the spectral response of migration from the intracellular to the extracellular space using near-infrared spectroscopy.

[0074] In-situ integration of segmentation and packaging: Flexible multi-sensor arrays are deployed in-situ within modified atmosphere packaging or cling film as the meat moves. At this point, the pork is segmented from its carcass form into heterogeneous retail cuts (such as tenderloin, pork belly, etc.). This process monitors and evaluates the impact of physical segmentation on performance stability: a quality model monitors in real-time the accelerated moisture loss caused by the cutting surface, accurately determining the contribution of drip loss to tissue surface area through spectral baseline analysis. Simultaneously, the system continuously tracks the slight rise and stabilization of pH value, and analyzes the interference caused by protein oxidation after segmentation using a 12-channel signal analysis system. This monitoring stage ensures the logical continuity of quality data during the physical switching process, preventing monitoring interruptions due to processing disruptions.

[0075] Compensation for Delivery and Sales Environment: The system needs to address vibration fluctuations during cold chain transportation and monitor ambient light, temperature, and humidity fluctuations at sales terminals. The model algorithm dynamically adjusts the weight allocation of each base learner. During these long-term monitoring intervals, noise generation becomes a core indicator for assessing meat maturity: the system utilizes multi-band spectral signals to track the "one decrease and five increases" dynamics of independent noise (such as flavor precursors like glutamic acid and aspartic acid) generated by continuous protein degradation. The model also integrates dynamic fluctuations in pH values ​​and changes in the state of water molecules within the tissue microstructure, outputting the full-queue quality curve in real time via a cloud platform. This intelligent compensation mechanism for environmental disturbances, through cross-scenario fusion prediction of three major indicators—pH, moisture, and amino acids—ultimately achieves a digitally transparent reconstruction of the physiological and biochemical state of retail pork.

[0076] In the description of this specification, references to terms such as "in one embodiment," "in yet another embodiment," "exemplary," or "in a particular embodiment," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0077] Although the present invention has been described in detail above with general descriptions, specific embodiments, and experiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A flexible multispectral sensing-based dynamic monitoring system for pork cold chain quality, comprising a flexible multispectral sensing unit, a data acquisition and wireless transmission module, a data processing and integrated learning module, a battery-powered module, a monitoring terminal, and a cloud platform, wherein the data processing and integrated learning module is electrically connected to the flexible multispectral sensing unit, the data acquisition and wireless transmission module, the battery-powered module, and the cloud platform, respectively. The flexible multispectral sensing unit is matched to the low temperature and high humidity environment of the cold chain and conformally fits to the surface of pork or the inside of pork tissue; the data acquisition and wireless transmission module samples multi-band spectral signals and transmits them wirelessly via WiFi. The data processing and integrated learning module pre-processes the received multi-band spectral signals and uses a regression model algorithm to predict key indicators of pork quality. The key quality indicators for pork are at least one of pH value, moisture content, and amino acid content. The cloud platform displays and stores pork quality prediction results; The monitoring terminal communicates with the cloud platform via the network and displays the prediction results from the cloud platform.

2. The flexible multi-spectrum sensing pork cold-chain quality dynamic monitoring system according to claim 1, characterized in that, The flexible multispectral sensing unit includes a flexible substrate with low-temperature resistance, a multi-band light source module, a spectral sensing module, and a flexible printed circuit. The multi-band spectral signals collected by the flexible multispectral sensing unit are transmitted to the data acquisition and wireless transmission module via the flexible printed circuit.

3. The flexible multi-spectrum sensing pork cold-chain quality dynamic monitoring system according to claim 2, characterized in that, The flexible substrate is made of one or a combination of two of the following polymer materials: polyimide and polyethylene terephthalate, and the flexible substrate has the characteristics of low temperature resistance and high flexibility.

4. The flexible multispectral sensing-based dynamic monitoring system for pork cold chain quality according to claim 2, characterized in that, One or more flexible multispectral sensing units are placed on the pork chunks after slaughter, and are either attached to the surface of the pork or implanted inside the pork muscle tissue during the rapid cooling and aging stages of the pork.

5. The flexible multispectral sensing-based dynamic monitoring system for pork cold chain quality according to claim 1, characterized in that, The data processing and integrated learning module uses an embedded processing unit.

6. The flexible multispectral sensing-based dynamic monitoring system for pork cold chain quality according to claim 2, characterized in that, The multi-band light source module covers the visible light band 450nm~760nm to the near-infrared band 780nm~880nm.

7. The flexible multispectral sensing-based dynamic monitoring system for pork cold chain quality according to claim 4, characterized in that, The flexible multispectral sensing unit is externally encapsulated with a highly transparent elastic encapsulation layer.

8. The flexible multispectral sensing-based dynamic monitoring system for pork cold chain quality according to claim 2, characterized in that, The data acquisition and wireless transmission module samples and digitizes the multi-band spectral signals, and then sends them to the data processing and integrated learning module via the wireless communication module. The wireless communication module uses a WiFi wireless communication network and transmits data via the MQTT communication protocol.

9. A method for dynamic monitoring of pork cold chain quality using flexible multispectral sensing, characterized in that, include: 1) Collect multi-band spectral data and corresponding key quality indicators of pork during the cold chain process according to the preset sampling interval. The sampling interval is adjusted according to the cold chain stage. 2) Send the multi-band spectral data to the data processing and integrated learning module; 3) The model algorithm preprocesses the multi-band spectral data and establishes a pork quality prediction model based on the ensemble learning model algorithm. The preprocessing includes at least one of the following: standardization of the multi-band spectral data, construction of first-order / second-order difference features, and calculation of statistical features. The ensemble learning model is a stacked ensemble learning structure, which includes multiple regression model base learners and a fusion weight allocation module. 4) Output the prediction results of key pork quality indicators; 5) The prediction results are sent to the monitoring terminal and cloud platform for storage and display, so as to realize dynamic monitoring of the trend of pork quality changes.

10. The method for dynamic monitoring of pork cold chain quality using flexible multispectral sensing according to claim 9, characterized in that, The ensemble learning model algorithm includes multiple base learners and merges the output results of each base learner. It predicts key indicators of pork quality based on the base learners and merges the output results of each base learner through a stacked ensemble learning structure to obtain the prediction results of key indicators of pork quality.

Images