Intelligent monitoring system and method for whole process of beef steak processing based on Internet of Things

Abstract

The invention discloses a steak processing full-process intelligent monitoring system and method based on the Internet of Things, and relates to the technical field of Internet of Things monitoring, and the method comprises the specific steps: based on an Internet of Things sensing device and a steak processing process flow, collecting steak processing process data in stages, based on the collected multi-stage beef steak processing data, making and executing a multi-stage quality inspection rule, namely predicting beef thawing time through physical parameters of raw beef, performing quality inspection in a raw material treatment stage, performing quality inspection in a rough processing stage according to three-dimensional point cloud data of the cut beef steak, and performing quality inspection in a rough processing stage according to a high-temperature ratio and an RLU value of the beef steak. According to the method, a progressive quality level is formed through unfreezing prediction, segmentation verification and cooking degree and safety joint inspection, and full-process closed-loop quality control of beef steak processing from raw materials to finished products is achieved.

Description

Technical Field

[0001] This invention belongs to the field of Internet of Things (IoT) monitoring technology, specifically relating to an intelligent monitoring system and method for the entire steak processing process based on IoT. Background Technology

[0002] As consumers increasingly demand food safety, consistent quality, and transparency in processing, the traditional steak processing industry is facing the need for a shift from experience-driven to data-driven intelligent transformation. Traditional processing procedures generally rely on manual judgment, resulting in pain points such as large quality fluctuations, low efficiency, and difficulties in traceability.

[0003] In the steak processing, every step, from raw material handling to finished product packaging, can affect the quality of the final product. For example, improper thawing time can lead to uneven internal temperature in the steak, affecting subsequent processing and taste; size deviations and fiber damage during cutting can affect the steak's specifications and eating experience. Therefore, a strict quality inspection system is needed to ensure that every steak meets preset standards and guarantees product consistency. Traditional quality inspection methods often rely on manual labor, which is not only inefficient but also prone to subjective errors. For instance, traditional thawing relies on environmental temperature and humidity control and lacks real-time monitoring of the internal temperature gradient of the beef, leading to under-thawing or over-thawing and wasting raw materials. Traditional cutting precision is insufficient, and manual cutting relies on the experience of technicians, resulting in large deviations in size and weight, affecting product standardization and cost accounting.

[0004] Therefore, there is an urgent need for an intelligent monitoring method for the entire steak processing process based on the Internet of Things. By introducing advanced data acquisition and analysis technologies, automated quality inspection can be achieved, production efficiency can be improved, labor costs can be reduced, and problems can be detected and sorted in a timely manner to prevent unqualified products from entering the next stage and reduce resource waste. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent monitoring system and method for the entire steak processing process based on the Internet of Things, in order to solve the technical problems of difficulty in maintaining product consistency and difficulty in improving production efficiency in the prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] The IoT-based intelligent monitoring method for the entire steak processing process includes:

[0008] S1. Based on IoT sensing devices and steak processing technology, collect steak processing data in stages.

[0009] S2. Based on the collected multi-stage steak processing data, formulate and implement multi-stage quality inspection rules;

[0010] S201. Predict the thawing time of beef by using the physical parameters of raw beef, and conduct quality inspection during the raw material processing stage.

[0011] S202. Based on the three-dimensional point cloud data of the cut steak, conduct quality inspection during the rough processing stage.

[0012] S203. Conduct quality inspections during the finished product processing stage based on the high-temperature ratio and RLU value of the steak.

[0013] Furthermore, based on IoT sensing devices and the steak processing technology, data on the steak processing process is collected in stages. The specific method is as follows:

[0014] According to the steak processing technology, the steak processing process is divided into three stages: raw material processing, preliminary processing, and finished product processing.

[0015] During the raw material processing stage, RFID+temperature sensor tags are embedded in the surface and center of the raw beef. The surface tag is a flexible patch, and the center tag is a puncture probe that penetrates the muscle tissue to a depth of ≥a1 cm. The tags are arranged in a three-dimensional grid with a spacing of ≤a2 cm between adjacent tags. The surface and center temperature data of the raw beef are collected every t1 time.

[0016] In the rough processing stage, a 3D scanner is deployed before and after the cutting station to capture the three-dimensional point cloud data of the raw material before cutting and the steak after cutting, respectively. A dynamic weighing module is installed under the conveyor belt to record the weight of each steak synchronously. PCL is used to filter out outliers in the original point cloud, and noise points are removed by statistical filtering. The surface is smoothed by moving least squares surface fitting to smooth the surface unevenness and noise. The minimum bounding box of the steak is fitted based on the RANSAC algorithm to calculate the major axis, minor axis and thickness. Based on the YOLOv8 segmentation model, the size bounding boxes of steaks of different specifications are labeled for model training. The input data includes RGB images, depth maps and weight data. The output size deviation of the trained model is within the preset deviation range.

[0017] During the finished product processing stage, infrared thermal imaging devices are deployed on the grilling line to capture the temperature field distribution and high-temperature area ratio on the surface of the steak in real time. Before packaging, the surface components of the steak are sampled with swabs, and the RLU value is output by the ATP biofluorescence detector.

[0018] Furthermore, the thawing time of the raw beef is predicted by its physical parameters, and quality inspection is conducted during the raw material processing stage. The specific method is as follows:

[0019] During the raw material processing stage, each piece of raw beef is assigned a unique number, and its density, specific heat capacity, and geometric thickness are linked to it. The raw beef pieces are sequentially numbered 1, 2, ..., n. Starting from the moment the raw beef enters the thawing chamber, the temperature gradient of the raw beef is collected every t1 time interval. N temperature gradients are collected for each piece of raw beef. By combining the temperature gradient, beef density, specific heat capacity, and thickness of the raw beef, the thawing time is predicted. A preset time deviation threshold p is set, and the difference between the actual and predicted thawing times for all steaks is determined and denoted as the thawing time difference. An absolute value for the thawing time difference is set. p represents the condition for the steak to meet the thawing quality standards. Conversely, if the conditions are not met, the robotic arm is triggered to sort the raw material to the quality inspection line for inspection, while the remaining raw beef enters the preliminary processing stage for processing.

[0020] Furthermore, the specific method for predicting the thawing time of beef is as follows:

[0021] Using formula Predict the thawing time of beef, where i represents the raw beef with the number i. This indicates the predicted thawing time of raw beef i. This represents the density of raw beef i. This indicates the specific heat capacity of the raw beef i. Indicates the core temperature of the target. This represents the initial core temperature of raw beef i. This represents the thickness of raw beef i, specifically the geometric thickness of the beef chunk in the thawing direction. Let i represent the temperature gradient of the raw beef, and k be the thermal conductivity.

[0022] Furthermore, based on the three-dimensional point cloud data of the cut steak, a preliminary quality inspection is conducted. The specific method is as follows:

[0023] The bounding box dimensions output by the YOLOv8 segmentation model are jointly verified with the weighing data. In the 3D point cloud, the hemisphere is divided into 10×10 grids, with each grid corresponding to the intervals of polar angle θ and azimuth angle φ. The distribution of all normal vectors in each grid is counted, and the proportion of normal vectors in each grid to the total number of normal vectors is calculated, denoted as the grid probability p(x), where x represents the x-th grid. This is achieved using the formula... Let x represent the entropy value, where x represents the x-th grid and j represents the j-th steak after cutting. This represents the grid probability of the x-th grid of steak j after cutting. Steaks with an entropy value ≤ 1.2 are considered to have minor fiber damage and can proceed to the finished product processing stage. Steaks with an entropy value > 1.2 are considered to have excessive fiber breakage and trigger sorting. Based on the entropy value judgment result, the robotic arm transfers steaks with H > 1.2 to the manual re-inspection station or rework line. If 10 consecutive steaks have H > 1.2, the system pushes a tool replacement prompt.

[0024] Furthermore, based on the high-temperature ratio and RLU value of the steak, quality inspection is carried out during the finished product processing stage. The specific method is as follows:

[0025] An infrared thermal imager deployed along the grilling line captures the surface temperature distribution of the steak at a rate of 10 frames per second, re-dividing the steak surface into grid regions, divided into U... A V-shaped grid is used to determine the average temperature of each grid cell, and the percentage of high-temperature steaks is monitored in real time. when When 85% of the surface charring of the steak is determined to be up to standard, the robotic arm sorts the steaks with up to standard to the automatic packaging area. In the automatic packaging area, a sampling swab is used to smear the surface area S of the steak, which is then dissolved in buffer solution to measure the RLU value. A dynamic safety threshold is set according to the type of steak product. If the actual measured RLU value is greater than the dynamic safety threshold, automatic packaging is triggered, and the UV-C sterilization device is started. After a 30-second delay, the test is repeated. Otherwise, automatic packaging is performed.

[0026] Furthermore, the high-temperature percentage of the steak is monitored in real time, using the following method:

[0027] Using formula This represents the percentage of high-temperature samples, where y represents the y-th grid cell and j represents the steak numbered j. This represents the average temperature of the y-th grid cell. This indicates the critical temperature at which the Maillard reaction occurs completely. This represents the area of ​​the y-th grid cell. This represents the surface area of ​​the steak numbered j. This indicates the uniformity of the surface temperature gradient of the steak numbered j. This is an indicator function that returns 1 if the condition * is met, and 0 otherwise. To adjust the weights.

[0028] Furthermore, the uniformity of the temperature gradient on the surface of the steak specifically includes:

[0029] Using formula This represents the uniformity of the temperature gradient on the surface of the steak, where y represents the y-th grid and j represents the steak numbered j. This represents the average temperature of the y-th grid cell. This represents the overall average temperature of steak j.

[0030] This invention also provides an IoT-based intelligent monitoring system for the entire steak processing process, applicable to an IoT-based intelligent monitoring method for the entire steak processing process, comprising:

[0031] The phased data acquisition module collects data on the steak processing process in stages, based on IoT sensing devices and the steak processing technology.

[0032] The raw material processing stage quality inspection module predicts the thawing time of beef based on the physical parameters of the raw beef and conducts quality inspection during the raw material processing stage.

[0033] The quality inspection module for the rough processing stage performs quality inspection based on the three-dimensional point cloud data of the cut steak.

[0034] The finished product processing stage quality inspection module performs quality inspections based on the high-temperature ratio and RLU value of the steak.

[0035] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0036] 1. This invention achieves real-time monitoring of temperature gradient by embedding different forms of RFID temperature sensing tags on the surface and center of raw beef. Based on the collected multi-dimensional data such as temperature gradient, beef density, specific heat capacity and thickness of raw beef, the thawing time of beef is predicted by a specific formula, and the absolute value of the thawing time difference is set as a quality standard condition. This enables quantitative evaluation of thawing quality, making the thawing process more scientific and controllable, and effectively avoiding raw material quality problems caused by improper thawing.

[0037] 2. This invention verifies the bounding box dimensions output by the segmentation model with the weighing data, and rigorously inspects the steak specifications from multiple dimensions such as major axis, minor axis, thickness, and weight. This ensures that the steaks entering the next processing stage meet the standard requirements, improves the quality control level in the rough processing stage, and determines whether there is fiber damage in the beef after cutting by calculating the entropy value of the normal vector in the three-dimensional point cloud. This provides a new quantitative evaluation index for steak cutting quality and effectively ensures the consistency and stability of steak cutting quality.

[0038] 3. This invention improves the accuracy and timeliness of charring determination by real-time inspection of the high-temperature ratio of steak and considering the uniformity of the temperature gradient on the steak surface. It uses a specific formula to determine whether the charring of the steak surface meets the standard, ensuring that the steak grilling quality meets the requirements. This solution binds the data collected at each stage to the steak number, supports blockchain traceability, and ensures that the entire process of steak processing from raw material processing to finished product processing has detailed and traceable records. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 A flowchart illustrating the steps of an IoT-based intelligent monitoring method for the entire steak processing process is provided.

[0041] Figure 2 A flowchart illustrating the intelligent monitoring steps for the entire steak processing process based on the Internet of Things is shown.

[0042] Figure 3 A block diagram of an IoT-based intelligent monitoring system for the entire steak processing process is shown. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0044] Example 1, such as Figure 1 , Figure 2 As shown, the intelligent monitoring method for the entire steak processing process based on the Internet of Things includes the following steps:

[0045] S1. Based on IoT sensing devices and steak processing technology, collect steak processing data in stages.

[0046] According to the steak processing technology, the steak processing process is divided into three stages: raw material processing, preliminary processing, and finished product processing.

[0047] During the raw material processing stage, medical-grade RFID temperature sensing tags (hereinafter referred to as tags) are embedded in the quality-inspected raw beef to achieve real-time monitoring of temperature gradient. The RFID + temperature sensor tags are embedded in the surface and center of the raw beef. The surface tag is a flexible patch (0.5mm thick), and the center tag is a puncture probe (2mm in diameter) that penetrates ≥a1 cm into the muscle tissue. The tags are arranged in a three-dimensional grid, and the spacing between adjacent tags is ≤a2 cm (for large pieces of raw meat ≥20kg). The surface and center temperature data of the raw beef are collected every t1 time.

[0048] In the rough processing stage, a 3D scanner is deployed before and after the cutting station to capture the 3D point cloud data of the raw material before cutting and the cut steak after cutting, respectively. A dynamic weighing module is installed under the conveyor belt with an accuracy of ±m grams to record the weight of each steak simultaneously. PCL (Point Cloud Library) is used to filter out outliers in the original point cloud. Noise points are removed by statistical filtering (50 neighborhood points, standard deviation multiple of 1.0). Surface noise is smoothed by moving least squares (MLS) surface fitting. The minimum bounding box of the steak is fitted based on the RANSAC algorithm to calculate the major axis, minor axis, and thickness. Based on the YOLOv8 segmentation model, the size bounding boxes of different steak sizes (such as tenderloin 200g±5g, sirloin 300g±10g) are labeled for model training. The input data includes RGB images, depth maps, and weight data. The output size deviation of the trained model is within the preset deviation range.

[0049] During the finished product processing stage, infrared thermal imaging devices are deployed on the grilling line to capture the temperature field distribution and high-temperature area ratio on the surface of the steak in real time. Before packaging, the surface components of the steak are sampled with swabs, and the RLU value is output by the ATP biofluorescence detector.

[0050] S2. Based on the collected multi-stage steak processing data, formulate and implement multi-stage quality inspection rules.

[0051] S201. Predict the thawing time of beef based on its physical parameters and conduct quality inspection during the raw material processing stage:

[0052] During the raw material processing stage, each piece of raw beef is assigned a unique number and its density (i.e., the mass of beef per unit volume), specific heat capacity (derived from a regression model based on historical data to determine the average amount of heat required to warm a unit volume of frozen beef), and geometric thickness (extracted from 3D scanning data) are assigned numbers 1, 2, ..., n according to the order of the raw beef. Starting from the moment the raw beef enters the thawing chamber, the temperature gradient (i.e., the temperature difference between the average surface temperature and the core temperature of the beef) of the raw beef is collected every t1 time interval. N temperature gradients are collected for each piece of raw beef. By combining the temperature gradient, beef density, specific heat capacity, and thickness of the raw beef, the thawing time of the beef is predicted. The specific formula is shown below:

[0053] ;

[0054] Where i represents the raw beef with the number i. This indicates the predicted thawing time of raw beef i. This represents the density of raw beef i. This indicates the specific heat capacity of the raw beef i. Indicates the core temperature of the target. This represents the initial core temperature of raw beef i. This represents the thickness of raw beef i, specifically the geometric thickness of the beef chunk in the thawing direction. The temperature gradient of raw beef i is represented by k, which is the thermal conductivity, a parameter reflecting the thermal conductivity of different types of beef, and is related to fat content and fiber structure.

[0055] A preset time deviation threshold p is set to determine the difference between the actual and predicted thawing times for all steaks, denoted as the thawing time difference. An absolute value for the thawing time difference is also set. p represents the condition for the steak to meet the thawing quality standards. Conversely, if the conditions are not met, the robotic arm is triggered to sort the raw material to the quality inspection line for inspection, while the remaining raw beef enters the preliminary processing stage for processing.

[0056] S202. Based on the three-dimensional point cloud data of the cut steak, conduct quality inspection during the rough processing stage:

[0057] Based on a high-precision line laser 3D scanner deployed before and after the cutting station, RGB images, depth maps and point cloud data of the raw materials before cutting and the cut steaks after cutting are collected simultaneously. The cut steaks are renumbered in sequence. Based on a pressure sensor array installed under the conveyor belt, the weight of each steak as it passes through is recorded in real time and aligned with the timestamp of the 3D scanning data.

[0058] The bounding box dimensions (RGB + depth map) output by the YOLOv8 segmentation model are jointly verified with the weighing data, using the following formula:

[0059] ;

[0060] Where j represents the steak numbered j, and Ca represents the length of the major axis of the steak after dicing. This indicates the length of the major axis of a standard-sized steak. This indicates the length of the minor axis of the cut steak. Ls represents the minor axis length of a standard-sized steak (in this example, Ls equals 15cm and Ws equals 8cm), and La represents the geometric thickness of the steak after cutting. This indicates the geometric thickness of a standard-sized steak. This indicates the weight of the steak after it has been cut. This indicates the weight of a standard-sized steak. This is the area error threshold. This is the thickness error threshold value. The threshold is the quality error threshold. If any condition is not met, the robotic arm will automatically sort the samples to the rework line.

[0061] In a 3D point cloud, the normal vector of each point is perpendicular to the surface it lies on. Muscle fibers are typically arranged in a certain direction. If the cut is done properly, the fibers are neatly arranged, resulting in a smoother cut surface with a more consistent normal vector direction. If the cut causes fiber breakage, the surface becomes rougher, and the normal vector directions become more dispersed. When cutting a steak, the direction of the blade's movement is fixed (e.g., cutting from top to bottom). The normal vector of the cut surface is mainly distributed on one side of the blade's movement direction (i.e., the "upper hemisphere" or "lower hemisphere"). For example, when the blade cuts downwards, the normal vector of the cut surface points upwards, which is the opposite direction of the blade's entry. This results in a hemisphere (the normal vector of the cut surface) pointing upwards. The vector direction range (0°~180°) is divided into 10×10 grids. Each grid corresponds to the polar angle θ (the angle between the normal vector and the main direction of tool movement, such as the vertical downward direction, ranging from 0° to 90°, with a step size of 9°) and the azimuth angle φ (the projection direction of the normal vector onto the horizontal plane, ranging from 0° to 360°, with a step size of 36°). The distribution of all normal vectors in each grid is counted, and the proportion of the number of normal vectors in each grid to the total number of normal vectors is calculated, denoted as the grid probability p(x), where x represents the x-th grid. Entropy value is used to determine whether there is fiber damage in the beef after cutting. The specific entropy value calculation formula is as follows:

[0062] ;

[0063] Where x represents the x-th grid, and j represents the steak numbered j after cutting. Let x represent the grid probability of the x-th grid after the steak j is cut.

[0064] Steaks with an entropy value ≤ 1.2 are considered to have minor fiber damage and can proceed to the finished product processing stage. Steaks with an entropy value > 1.2 are considered to have excessive fiber breakage and trigger sorting. Based on the entropy value determination result, the robotic arm transfers steaks with H > 1.2 to the manual re-inspection station or rework line. If 10 consecutive steaks have H > 1.2, the system pushes a tool replacement prompt. The entropy value data is bound to the steak number, supporting blockchain traceability.

[0065] S203. Conduct quality inspections during the finished product processing stage based on the high-temperature ratio and RLU value of the steak.

[0066] An infrared thermal imager deployed along the grilling line captures the surface temperature distribution of the steak at a rate of 10 frames per second, re-dividing the steak surface into grid regions, divided into U... A grid of V is used to determine the average temperature of each grid cell, and the percentage of high-temperature steak is monitored in real time. The specific formula for calculating the percentage of high-temperature steak is shown below:

[0067] ;

[0068] Where y represents the y-th grid, and j represents the steak numbered j. This represents the average temperature of the y-th grid cell. The critical temperature at which the Maillard reaction occurs fully (equal to 75°C) is indicated. This represents the area of ​​the y-th grid cell. This represents the surface area of ​​the steak numbered j. This indicates the uniformity of the surface temperature gradient of the steak numbered j. This is an indicator function (it takes the value 1 if the condition is met, otherwise it takes the value 0). To adjust the weights.

[0069] Furthermore, the specific formula for calculating the uniformity of the temperature gradient on the surface of the steak is as follows:

[0070] ;

[0071] Where y represents the y-th grid, and j represents the steak numbered j. This represents the average temperature of the y-th grid cell. This represents the overall average temperature of steak j.

[0072] when When 85% of the steaks are deemed to have reached the standard for surface charring, the robotic arm sorts the steaks with the qualified surface charring to the automatic packaging area based on the judgment result.

[0073] In the automated packaging area, a sampling swab is used to smear the surface area S of the steak. After dissolving in a buffer solution, the RLU value (Relative Light Unit) is measured. A dynamic safety threshold is set according to the type of steak product (e.g., the dynamic safety threshold is 200 RLU for chilled steak and 500 RLU for frozen steak). If the actual measured RLU value is greater than the dynamic safety threshold, automated packaging is triggered, and the UV-C sterilization device (dose ≥ 50 mJ / cm²) is activated. After a 30-second delay, the test is repeated. Otherwise, automated packaging proceeds.

[0074] Example 2, as follows Figure 3 As shown, the IoT-based intelligent monitoring system for the entire steak processing process includes the following:

[0075] The phased data acquisition module divides the steak processing process into three stages according to the steak processing technology: raw material processing stage, preliminary processing stage and finished product processing stage.

[0076] During the raw material processing stage, medical-grade RFID temperature sensing tags (hereinafter referred to as tags) are embedded in the quality-inspected raw beef to achieve real-time monitoring of temperature gradients. The RFID + temperature sensor tags are embedded in the surface and center of the raw beef. The surface tag is a flexible patch (0.5mm thick), and the center tag is a puncture probe (2mm in diameter) that penetrates ≥a1 cm into the muscle tissue. The tags are arranged in a three-dimensional grid, with a spacing of ≤a2 cm between adjacent tags (for large pieces of raw meat ≥20kg). In the thawing chamber, one temperature and humidity sensor is deployed every a3 square meters. Airflow velocity sensors (range 0-15m / s) are distributed at equal intervals along the airflow direction. The surface and center temperature data of the raw beef and the environmental parameters (temperature, humidity, and wind speed) of the thawing chamber are collected every t1 time interval.

[0077] In the rough processing stage, a 3D scanner is deployed before and after the cutting station to capture the 3D point cloud data of the raw material before cutting and the cut steak after cutting, respectively. A dynamic weighing module is installed under the conveyor belt with an accuracy of ±m grams to record the weight of each steak simultaneously. PCL (Point Cloud Library) is used to filter out outliers in the original point cloud. Noise points are removed by statistical filtering (50 neighborhood points, standard deviation multiple of 1.0). Surface noise is smoothed by moving least squares (MLS) surface fitting. The minimum bounding box of the steak is fitted based on the RANSAC algorithm to calculate the major axis, minor axis, and thickness. Based on the YOLOv8 segmentation model, the size bounding boxes of different steak sizes (such as tenderloin 200g±5g, sirloin 300g±10g) are labeled for model training. The input data includes RGB images, depth maps, and weight data. The output size deviation of the trained model is within the preset deviation range.

[0078] During the finished product processing stage, infrared thermal imaging devices are deployed on the grilling line to capture the temperature field distribution and high-temperature area ratio on the surface of the steak in real time. Before packaging, the surface components of the steak are sampled with swabs, and the RLU value is output by the ATP biofluorescence detector.

[0079] The quality inspection module in the raw material processing stage assigns a unique number to each piece of raw beef, binding it with density (i.e., the mass of beef per unit volume), specific heat capacity (derived from a regression model based on historical data to determine the average heat required to warm a unit volume of frozen beef), and geometric thickness (extracted from 3D scanning data). The raw beef pieces are sequentially numbered 1, 2, ..., n. Starting from the moment the raw beef enters the thawing chamber, the temperature gradient (i.e., the temperature difference between the average surface temperature and the core temperature of the beef) is collected every t1 time interval. N temperature gradients are collected for each piece of raw beef. By combining the temperature gradient, beef density, specific heat capacity, and thickness of the raw beef, the thawing time is predicted. The specific formula is shown below:

[0080] ;

[0081] Where i represents the raw beef with the number i. This indicates the predicted thawing time of raw beef i. This represents the density of raw beef i. This indicates the specific heat capacity of raw beef i. Indicates the core temperature of the target. This represents the initial core temperature of raw beef i. This represents the thickness of raw beef i, specifically the geometric thickness of the beef chunk in the thawing direction. The temperature gradient of raw beef i is represented by k, which is the thermal conductivity, a parameter reflecting the thermal conductivity of different types of beef, and is related to fat content and fiber structure.

[0082] A preset time deviation threshold p is set to determine the difference between the actual and predicted thawing times for all steaks, denoted as the thawing time difference. An absolute value for the thawing time difference is also set. p represents the condition for the steak to meet the thawing quality standards. Conversely, if the conditions are not met, the robotic arm is triggered to sort the raw material to the quality inspection line for inspection, while the remaining raw beef enters the preliminary processing stage for processing.

[0083] The quality inspection module in the rough processing stage is based on a high-precision line laser 3D scanner deployed before and after the cutting station. It simultaneously collects RGB images, depth maps and point cloud data of the raw materials before cutting and the cut steaks after cutting. The cut steaks are renumbered in sequence. Based on the pressure sensor array installed under the conveyor belt, the weight of each steak as it passes through is recorded in real time and aligned with the timestamp of the 3D scan data.

[0084] The bounding box dimensions (RGB + depth map) output by the YOLOv8 segmentation model are jointly verified with the weighing data, using the following formula:

[0085] ;

[0086] Where j represents the steak numbered j, and Ca represents the length of the major axis of the steak after cutting. This indicates the length of the major axis of a standard-sized steak. This indicates the length of the minor axis of the cut steak. Ls represents the minor axis length of a standard-sized steak (in this example, Ls equals 15cm and Ws equals 8cm), and La represents the geometric thickness of the steak after cutting. This indicates the geometric thickness of a standard-sized steak. This indicates the weight of the steak after it has been cut. This indicates the weight of a standard-sized steak. This is the area error threshold. This is the thickness error threshold value. The threshold is the quality error threshold. If any condition is not met, the robotic arm will automatically sort the samples to the rework line.

[0087] In a 3D point cloud, the normal vector of each point is perpendicular to the surface it lies on. Muscle fibers are typically arranged in a certain direction. If the cut is done properly, the fibers are neatly arranged, resulting in a smoother cut surface with a more consistent normal vector direction. If the cut causes fiber breakage, the surface becomes rougher, and the normal vector directions become more dispersed. When cutting a steak, the direction of the blade's movement is fixed (e.g., cutting from top to bottom). The normal vector of the cut surface is mainly distributed on one side of the blade's movement direction (i.e., the "upper hemisphere" or "lower hemisphere"). For example, when the blade cuts downwards, the normal vector of the cut surface points upwards, which is the opposite direction of the blade's entry. This results in a hemisphere (the normal vector of the cut surface) pointing upwards. The vector direction range (0°~180°) is divided into 10×10 grids. Each grid corresponds to the polar angle θ (the angle between the normal vector and the main direction of tool movement, such as the vertical downward direction, ranging from 0° to 90°, with a step size of 9°) and the azimuth angle φ (the projection direction of the normal vector onto the horizontal plane, ranging from 0° to 360°, with a step size of 36°). The distribution of all normal vectors in each grid is counted, and the proportion of the number of normal vectors in each grid to the total number of normal vectors is calculated, denoted as the grid probability p(x), where x represents the x-th grid. Entropy value is used to determine whether there is fiber damage in the beef after cutting. The specific entropy value calculation formula is as follows:

[0088] ;

[0089] Where x represents the x-th grid, and j represents the steak numbered j after cutting. Let x represent the grid probability of the x-th grid after the steak j is cut.

[0090] Steaks with an entropy value ≤ 1.2 are considered to have minor fiber damage and can proceed to the finished product processing stage. Steaks with an entropy value > 1.2 are considered to have excessive fiber breakage and trigger sorting. Based on the entropy value determination result, the robotic arm transfers steaks with H > 1.2 to the manual re-inspection station or rework line. If 10 consecutive steaks have H > 1.2, the system pushes a tool replacement prompt. The entropy value data is bound to the steak number, supporting blockchain traceability.

[0091] The quality inspection module in the finished product processing stage uses an infrared thermal imager deployed on the grilling line to capture the surface temperature distribution of the steak at a rate of 10 frames per second, and then re-divides the grid area of ​​the steak surface into U-shaped regions. A grid of V is used to determine the average temperature of each grid cell, and the percentage of high-temperature steak is monitored in real time. The specific formula for calculating the percentage of high-temperature steak is shown below:

[0092] ;

[0093] Where y represents the y-th grid, and j represents the steak numbered j. This represents the average temperature of the y-th grid cell. The critical temperature at which the Maillard reaction occurs fully (equal to 75°C) is indicated. This represents the area of ​​the y-th grid cell. This represents the surface area of ​​the steak numbered j. This indicates the uniformity of the surface temperature gradient of the steak numbered j. This is an indicator function (it takes the value 1 if the condition is met, otherwise it takes the value 0). To adjust the weights.

[0094] Furthermore, the specific formula for calculating the uniformity of the temperature gradient on the surface of the steak is as follows:

[0095] ;

[0096] Where y represents the y-th grid, and j represents the steak numbered j. This represents the average temperature of the y-th grid cell. This represents the overall average temperature of steak j.

[0097] when When 85% of the steaks are deemed to have reached the standard for surface charring, the robotic arm sorts the steaks with the qualified surface charring to the automatic packaging area based on the judgment result.

[0098] In the automated packaging area, a sampling swab is used to smear the surface area S of the steak. After dissolving in a buffer solution, the RLU value (Relative Light Unit) is measured. A dynamic safety threshold is set according to the type of steak product (e.g., the dynamic safety threshold is 200 RLU for chilled steak and 500 RLU for frozen steak). If the actual measured RLU value is greater than the dynamic safety threshold, automated packaging is triggered, and the UV-C sterilization device (dose ≥ 50 mJ / cm²) is activated. After a 30-second delay, the test is repeated. Otherwise, automated packaging proceeds.

[0099] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

[0100] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to specific implementations. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for intelligent monitoring of the entire steak processing process based on the Internet of Things, characterized in that, include: S1. Based on IoT sensing devices and steak processing technology, collect steak processing data in stages. S2. Based on the collected multi-stage steak processing data, formulate and implement multi-stage quality inspection rules; S201. Predict the thawing time of beef by using the physical parameters of raw beef, and conduct quality inspection during the raw material processing stage. S202. Based on the three-dimensional point cloud data of the cut steak, conduct quality inspection during the rough processing stage. S203. Conduct quality inspections during the finished product processing stage based on the high-temperature ratio and RLU value of the steak.

2. The intelligent monitoring method for the entire steak processing process based on the Internet of Things as described in claim 1, characterized in that, Based on IoT sensing devices and the steak processing technology, data on the steak processing process is collected in stages. The specific method is as follows: According to the steak processing technology, the steak processing process is divided into three stages: raw material processing, preliminary processing, and finished product processing. During the raw material processing stage, RFID+temperature sensor tags are embedded in the surface and center of the raw beef. The surface tag is a flexible patch, and the center tag is a puncture probe that penetrates the muscle tissue to a depth of ≥a1 cm. The tags are arranged in a three-dimensional grid with a spacing of ≤a2 cm between adjacent tags. The surface and center temperature data of the raw beef are collected every t1 time. In the rough processing stage, a 3D scanner is deployed before and after the cutting station to capture the three-dimensional point cloud data of the raw material before cutting and the steak after cutting, respectively. A dynamic weighing module is installed under the conveyor belt to record the weight of each steak synchronously. PCL is used to filter out outliers in the original point cloud, and noise points are removed by statistical filtering. The surface is smoothed by moving least squares surface fitting to smooth the surface unevenness and noise. The minimum bounding box of the steak is fitted based on the RANSAC algorithm to calculate the major axis, minor axis and thickness. Based on the YOLOv8 segmentation model, the size bounding boxes of steaks of different specifications are labeled for model training. The input data includes RGB images, depth maps and weight data. The output size deviation of the trained model is within the preset deviation range. During the finished product processing stage, infrared thermal imaging devices are deployed on the grilling line to capture the temperature field distribution and high-temperature area ratio on the surface of the steak in real time. Before packaging, the surface components of the steak are sampled with swabs, and the RLU value is output by the ATP biofluorescence detector.

3. The intelligent monitoring method for the entire steak processing process based on the Internet of Things according to claim 1, characterized in that, The thawing time of raw beef is predicted based on its physical parameters, and quality inspection is conducted during the raw material processing stage. The specific method is as follows: During the raw material processing stage, each piece of raw beef is assigned a unique number, and its density, specific heat capacity, and geometric thickness are linked to it. The raw beef pieces are sequentially numbered 1, 2, ..., n. Starting from the moment the raw beef enters the thawing chamber, the temperature gradient of the raw beef is collected every t1 time interval. N temperature gradients are collected for each piece of raw beef. By combining the temperature gradient, beef density, specific heat capacity, and thickness of the raw beef, the thawing time is predicted. A preset time deviation threshold p is set, and the difference between the actual and predicted thawing times for all steaks is determined and denoted as the thawing time difference. An absolute value for the thawing time difference is set. p represents the condition for the steak to meet the thawing quality standards. Conversely, if the conditions are not met, the robotic arm is triggered to sort the raw material to the quality inspection line for inspection, while the remaining raw beef enters the preliminary processing stage for processing.

4. The intelligent monitoring method for the entire steak processing process based on the Internet of Things according to claim 3, characterized in that, The specific method for predicting the thawing time of beef is as follows: Using formula Predict the thawing time of beef, where i represents the raw beef with the number i. This indicates the predicted thawing time of raw beef i. This represents the density of raw beef i. This indicates the specific heat capacity of the raw beef i. Indicates the core temperature of the target. This represents the initial core temperature of raw beef i. This represents the thickness of raw beef i, specifically the geometric thickness of the beef chunk in the thawing direction. Let i represent the temperature gradient of the raw beef, and k be the thermal conductivity.

5. The intelligent monitoring method for the entire steak processing process based on the Internet of Things according to claim 1, characterized in that, Based on the 3D point cloud data of the cut steak, a preliminary quality inspection is performed. The specific method is as follows: The bounding box dimensions output by the YOLOv8 segmentation model are jointly verified with the weighing data. In the 3D point cloud, the hemisphere is divided into 10×10 grids, with each grid corresponding to the intervals of polar angle θ and azimuth angle φ. The distribution of all normal vectors in each grid is counted, and the proportion of normal vectors in each grid to the total number of normal vectors is calculated, denoted as the grid probability p(x), where x represents the x-th grid. This is achieved using the formula... Let x represent the entropy value, where x represents the x-th grid and j represents the j-th steak after cutting. This represents the grid probability of the x-th grid of steak j after cutting. Steaks with an entropy value ≤ 1.2 are judged as having minor fiber damage and enter the finished product processing stage. Steaks with an entropy value > 1.2 are judged as having excessive fiber breakage and trigger sorting. The robotic arm transfers steaks with H > 1.2 to the manual inspection station or rework line based on the entropy value judgment result. If 10 consecutive steaks have H > 1.2, the system pushes a tool replacement prompt.

6. The intelligent monitoring method for the entire steak processing process based on the Internet of Things according to claim 1, characterized in that, Based on the high-temperature ratio and RLU value of the steak, quality inspection is carried out during the finished product processing stage. The specific method is as follows: An infrared thermal imager deployed along the grilling line captures the surface temperature distribution of the steak at a rate of 10 frames per second, re-dividing the steak surface into grid regions, divided into U... A V-shaped grid is used to determine the average temperature of each grid cell, and the percentage of high-temperature steaks is monitored in real time. when When 85% of the surface charring of the steak is determined to be up to standard, the robotic arm sorts the steaks with up to standard to the automatic packaging area. In the automatic packaging area, a sampling swab is used to smear the surface area S of the steak, which is then dissolved in buffer solution to measure the RLU value. A dynamic safety threshold is set according to the type of steak product. If the actual measured RLU value is greater than the dynamic safety threshold, automatic packaging is triggered, and the UV-C sterilization device is started. After a 30-second delay, the test is repeated. Otherwise, automatic packaging is performed.

7. The intelligent monitoring method for the entire steak processing process based on the Internet of Things according to claim 6, characterized in that, The specific method for real-time monitoring of the high-temperature percentage of steak is as follows: Using formula This represents the percentage of high-temperature samples, where y represents the y-th grid cell and j represents the steak numbered j. This represents the average temperature of the y-th grid cell. This indicates the critical temperature at which the Maillard reaction occurs completely. This represents the area of ​​the y-th grid cell. This represents the surface area of ​​the steak numbered j. This indicates the uniformity of the surface temperature gradient of the steak numbered j. This is an indicator function that returns 1 if the condition * is met, and 0 otherwise. To adjust the weights.

8. The intelligent monitoring method for the entire steak processing process based on the Internet of Things according to claim 7, characterized in that, The uniformity of the surface temperature gradient of the steak includes: Using formula This represents the uniformity of the temperature gradient on the surface of the steak, where y represents the y-th grid and j represents the steak numbered j. This represents the average temperature of the y-th grid cell. This represents the overall average temperature of steak j.

9. An IoT-based intelligent monitoring system for the entire steak processing process, applied to the IoT-based intelligent monitoring method for the entire steak processing process as described in any one of claims 1-8, characterized in that, include: The phased data acquisition module collects data on the steak processing process in stages, based on IoT sensing devices and the steak processing technology. The raw material processing stage quality inspection module predicts the thawing time of beef based on the physical parameters of the raw beef and conducts quality inspection during the raw material processing stage. The quality inspection module for the rough processing stage performs quality inspection based on the three-dimensional point cloud data of the cut steak. The finished product processing stage quality inspection module performs quality inspections based on the high-temperature ratio and RLU value of the steak.

Images