Marinating and boiling online monitoring method and system based on multi-source information fusion

Abstract

The embodiment of the invention relates to the technical field of marinating and boiling online monitoring, and particularly discloses a marinating and boiling online monitoring method and system based on multi-source information fusion. According to the embodiment of the invention, periodic sensing on-line monitoring is carried out on marinating and boiling to obtain sensing and monitoring data; carrying out online monitoring shooting to obtain monitoring shooting data; performing data cleaning, standardization processing and multi-source data fusion to generate fused monitoring data; carrying out online display, and carrying out abnormity identification and related verification; and when the abnormal condition is determined, generating an abnormal regulation instruction, and carrying out online abnormal regulation on the marinating cooking. Sensing on-line monitoring, on-line monitoring shooting and data processing can be carried out on marinating and boiling, then abnormity recognition and related verification are carried out, when abnormity is determined, on-line abnormity adjustment is carried out, intelligent multi-source monitoring and regulation in the marinating and boiling processing process are achieved, the marinating and boiling on-line monitoring efficiency is effectively improved, and the marinating and boiling quality is improved. And the processing process can be accurately controlled, so that the quality and the consistency of the marinated and boiled products are ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of online monitoring of marinated cooking, and particularly relates to an online monitoring method and system for marinated cooking based on multi-source information fusion. Background Art

[0002] Online monitoring of marinated cooking is a process of real-time monitoring and data collection of key parameters (such as temperature, time, humidity, ingredient content, etc.) during the marinating and cooking process of food, especially for meat, soy products, etc., so as to ensure the quality and safety of food, improve production efficiency, and at the same time ensure that the taste and flavor of food reach the best state.

[0003] In the prior art, the online monitoring means for the marinated cooking process are relatively limited, mainly limited to simple monitoring of temperature, and the level of intelligence is significantly insufficient. In the actual monitoring process, it still largely depends on direct manual observation and manual adjustment and control, which is not only inefficient, but also difficult to achieve precise control of the processing process, thus affecting the quality and consistency of marinated cooking products. Summary of the Invention

[0004] The purpose of the embodiments of the present invention is to provide an online monitoring method and system for marinated cooking based on multi-source information fusion, aiming to solve the problems proposed in the background art.

[0005] To achieve the above purpose, the embodiments of the present invention provide the following technical solutions: An online monitoring method for marinated cooking based on multi-source information fusion, the method specifically includes the following steps: Carry out periodic sensing online monitoring of marinated cooking according to a preset monitoring period to obtain sensing monitoring data; Carry out periodic online monitoring and shooting of marinated cooking according to a preset monitoring period to obtain monitoring shooting data; Perform data cleaning, standardization processing and multi-source data fusion on the sensing monitoring data and the monitoring shooting data to generate fusion monitoring data; Online display the fusion monitoring data, and perform anomaly recognition and related verification to determine whether there is a definite anomaly; When there is a definite anomaly, generate an anomaly adjustment instruction, and perform online anomaly adjustment on the marinated cooking according to the anomaly adjustment instruction.

[0006] An online monitoring system for marinated cooking based on multi-source information fusion, the system includes a sensing online monitoring unit, an online monitoring shooting unit, a data fusion processing unit, an anomaly recognition and verification unit and an online anomaly adjustment unit, wherein: The sensing online monitoring unit is used to carry out periodic sensing online monitoring of marinated cooking according to a preset monitoring period to obtain sensing monitoring data; An on-line monitoring and photographing unit for periodically on-line monitoring and photographing the marinated stew according to a preset monitoring period to obtain monitoring and photographing data; A data fusion processing unit for performing data cleaning, standardization processing, and multi-source data fusion on the sensing monitoring data and the monitoring and photographing data to generate fusion monitoring data; An anomaly recognition and verification unit for on-line displaying the fusion monitoring data, performing anomaly recognition and related verification, and determining whether there is a definite anomaly; An on-line anomaly adjustment unit for generating an anomaly adjustment instruction when there is a definite anomaly and performing on-line anomaly adjustment on the marinated stew according to the anomaly adjustment instruction.

[0007] Compared with the prior art, the beneficial effects of the present invention are as follows: In the embodiment of the present invention, periodic sensing on-line monitoring is performed on the marinated stew to obtain sensing monitoring data; on-line monitoring and photographing are performed to obtain monitoring and photographing data; data cleaning, standardization processing, and multi-source data fusion are performed to generate fusion monitoring data; on-line display is performed, and anomaly recognition and related verification are carried out; when there is a definite anomaly, an anomaly adjustment instruction is generated to perform on-line anomaly adjustment on the marinated stew. It is possible to perform sensing on-line monitoring, on-line monitoring and photographing, and data processing on the marinated stew, and then perform anomaly recognition and related verification. When there is a definite anomaly, on-line anomaly adjustment is performed, realizing intelligent multi-source monitoring and control in the processing process of the marinated stew, effectively improving the efficiency of on-line monitoring of the marinated stew, and being able to accurately control the processing process to ensure the quality and consistency of the marinated stew products. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0009] Figure 1 Shows a flowchart of the method provided by the embodiment of the present invention.

[0010] Figure 2 Shows a flowchart of periodic sensing on-line monitoring in the method provided by the embodiment of the present invention.

[0011] Figure 3 Shows a flowchart of periodic on-line monitoring and photographing in the method provided by the embodiment of the present invention.

[0012] Figure 4 Shows a flowchart of generating fusion monitoring data in the method provided by the embodiment of the present invention.

[0013] Figure 5The flowchart of anomaly recognition and related verification in the method provided by the embodiment of the present invention is shown.

[0014] Figure 6 The flowchart of performing online anomaly adjustment in the method provided by the embodiment of the present invention is shown.

[0015] Figure 7 The application architecture diagram of the system provided by the embodiment of the present invention is shown.

[0016] Figure 8 The structural block diagram of the sensing online monitoring unit in the system provided by the embodiment of the present invention is shown.

[0017] Figure 9 The structural block diagram of the data fusion processing unit in the system provided by the embodiment of the present invention is shown.

[0018] Figure 10 The structural block diagram of the anomaly recognition and verification unit in the system provided by the embodiment of the present invention is shown. Detailed implementation manners

[0019] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0020] It can be understood that in the prior art, the online monitoring means for the marinated cooking process are relatively limited, mainly limited to the simple monitoring of temperature, and the intelligent level is significantly insufficient. In the actual monitoring process, it still largely depends on the direct observation and manual adjustment and control by humans, which is not only inefficient, but also difficult to achieve precise control of the processing process, thus affecting the quality and consistency of marinated cooking products.

[0021] To solve the above problems, the embodiment of the present invention performs periodic sensing online monitoring on marinated cooking according to a preset monitoring period to obtain sensing monitoring data; performs periodic online monitoring and shooting on marinated cooking according to a preset monitoring period to obtain monitoring shooting data; performs data cleaning, standardization processing and multi-source data fusion on the sensing monitoring data and the monitoring shooting data to generate fusion monitoring data; performs online display on the fusion monitoring data, and performs anomaly recognition and related verification to determine whether there is a definite anomaly; when there is a definite anomaly, generates an anomaly adjustment instruction, and performs online anomaly adjustment on marinated cooking according to the anomaly adjustment instruction. It can perform sensing online monitoring, online monitoring shooting and data processing on marinated cooking, and then perform anomaly recognition and related verification. When there is a definite anomaly, it performs online anomaly adjustment, realizes intelligent multi-source monitoring and control in the marinated cooking process, effectively improves the efficiency of online monitoring of marinated cooking, and can precisely control the processing process to ensure the quality and consistency of marinated cooking products.

[0022] Figure 1 The flowchart of the method provided by the embodiment of the present invention is shown.

[0023] Specifically, for the online monitoring method of stewed pig offal based on multi-source information fusion, the method specifically includes the following steps: Step S101, perform periodic sensing online monitoring on the stewed pig offal according to a preset monitoring period to obtain sensing monitoring data.

[0024] In the embodiment of the present invention, temperature monitoring signals, pH monitoring signals, conductivity monitoring signals, and gas monitoring signals are periodically generated according to a preset monitoring period. Then, according to the temperature monitoring signal, periodic temperature online monitoring is performed on the stewed pig offal to obtain temperature monitoring data. According to the pH monitoring signal, periodic pH online monitoring is performed on the stewed pig offal to obtain pH monitoring data. According to the conductivity monitoring signal, periodic conductivity online monitoring is performed on the stewed pig offal to obtain conductivity monitoring data. According to the gas monitoring signal, periodic gas online monitoring is performed on the stewed pig offal to obtain gas monitoring data. The temperature monitoring data, pH monitoring data, conductivity monitoring data, and gas monitoring data are comprehensively processed to generate sensing monitoring data.

[0025] Specifically, Figure 2 The flowchart of the periodic sensing online monitoring in the method provided by the embodiment of the present invention is shown.

[0026] Among them, in the preferred embodiment provided by the present invention, the step of performing periodic sensing online monitoring on the stewed pig offal according to a preset monitoring period to obtain sensing monitoring data specifically includes the following steps: Step S1011, perform periodic temperature online monitoring on the stewed pig offal according to a preset monitoring period to obtain temperature monitoring data; Step S1012, perform periodic pH online monitoring on the stewed pig offal according to a preset monitoring period to obtain pH monitoring data; Step S1013, perform periodic conductivity online monitoring on the stewed pig offal according to a preset monitoring period to obtain conductivity monitoring data; Step S1014, perform periodic gas online monitoring on the stewed pig offal according to a preset monitoring period to obtain gas monitoring data; Step S1015, comprehensively process the temperature monitoring data, the pH monitoring data, the conductivity monitoring data, and the gas monitoring data to generate sensing monitoring data.

[0027] Furthermore, the online monitoring method of stewed pig offal based on multi-source information fusion further includes the following steps: Step S102: Perform periodic online monitoring and shooting of the stewed pig offal according to a preset monitoring period, and obtain monitoring and shooting data.

[0028] In an embodiment of the present invention, a monitoring and shooting instruction is periodically generated according to a preset monitoring period. According to the monitoring and shooting instruction, online monitoring and shooting of the stewed pig offal is performed to obtain monitoring and shooting images, and multiple monitoring and shooting images are sorted out to generate monitoring and shooting data.

[0029] Specifically, Figure 3 FIG. shows a flowchart of periodic online monitoring and shooting in the method provided by the embodiment of the present invention.

[0030] Among them, in a preferred embodiment provided by the present invention, the step of performing periodic online monitoring and shooting of the stewed pig offal according to a preset monitoring period and obtaining monitoring and shooting data specifically includes the following steps: Step S1021: Periodically generate a monitoring and shooting instruction according to a preset monitoring period; Step S1022: Perform online monitoring and shooting of the stewed pig offal according to the monitoring and shooting instruction to obtain monitoring and shooting images; Step S1023: Sort out multiple monitoring and shooting images to generate monitoring and shooting data.

[0031] Furthermore, the online monitoring method for stewed pig offal based on multi-source information fusion further includes the following steps: Step S103: Perform data cleaning, standardization processing, and multi-source data fusion on the sensing monitoring data and the monitoring and shooting data to generate fused monitoring data.

[0032] In an embodiment of the present invention, by performing data cleaning of removing outliers and noise on the sensing monitoring data and the monitoring and shooting data, sensing cleaning data and shooting cleaning data are generated. Then, according to a preset standardization reference information, the sensing cleaning data and the shooting cleaning data are subjected to standardization processing to generate sensing standard data and shooting standard data. By extracting monitoring time series data from the sensing standard data and the shooting standard data, and then performing time series fusion processing of multi-source data on the sensing standard data and the shooting standard data according to the monitoring time series data, fused monitoring data is generated.

[0033] Specifically, Figure 4 FIG. shows a flowchart of generating fused monitoring data in the method provided by the embodiment of the present invention.

[0034] Among them, in a preferred embodiment provided by the present invention, the step of performing data cleaning, standardization processing, and multi-source data fusion on the sensing monitoring data and the monitoring and shooting data to generate fused monitoring data specifically includes the following steps: Step S1031: Perform data cleaning on the sensing monitoring data and the monitoring shooting data to remove outliers and noise, generating cleaned sensing data and cleaned shooting data; Step S1032: Perform standardization processing on the cleaned sensing data and the cleaned shooting data, generating standardized sensing data and standardized shooting data; Step S1033: Extract monitoring time-series data from the standardized sensing data and the standardized shooting data; Step S1034: Perform multi-source data fusion on the standardized sensing data and the standardized shooting data according to the monitoring time-series data, generating fused monitoring data.

[0035] Among them, in the preferred embodiment provided by the present invention, the performing multi-source data fusion on the standardized sensing data and the standardized shooting data according to the monitoring time-series data to generate fused monitoring data specifically includes the following steps: Calculate the instantaneous change rate of each sensor parameter at the current moment based on the numerical difference between adjacent time points of the standardized sensing data, obtaining change rate data; Extract a key time window from the monitoring time-series data according to the key reaction period of the marinated cooking process, and perform exponentially decaying weighted integration on the change rate data within the key time window to obtain a weighted change rate integral value; Combine the weighted change rate integral value with the standardized sensing data to obtain multi-dimensional sensor features including the standardized sensing data and time-varying characteristics; Calculate the difference amplitude between the current image feature and the previous moment image feature in the standardized shooting data, obtaining a difference value; Perform non-linear transformation on the difference value through the hyperbolic tangent function to calculate a modulation coefficient; Multiply the modulation coefficient with the image of the standardized shooting data element by element to suppress visual interference of non-significant changes, obtaining visually stable parameters with adaptive modulation; Establish a pairing relationship for each parameter between the multi-dimensional sensor features and the visual parameters, obtaining parameter pairs; Calculate the product of the parameters between the parameter pairs, and calculate the absolute value of the difference between the parameters between the parameter pairs. Combine the product of the parameters between the parameter pairs and the absolute value of the difference between the parameters between the parameter pairs to obtain a combined value; Filter negative interference and retain positive correlation features for the combined value through a non-linear activation function to obtain a pairing result; Sum all the pairing results to obtain a cross-modal coupling strength value; Calculate the variance statistical value of the multi-dimensional sensor features and calculate the change gradient value of the visual parameters; The variance statistical value of multi-dimensional sensor features is used to generate sensor weight values by an exponential function, and corresponding visual weights are set according to the change gradient value of visual parameters; The sensor weight values and visual weights are normalized and harmonized to balance the contribution ratios of different data sources, obtaining the coupling term weights; The multi-dimensional sensor features are weighted with the sensor weight values, obtaining the weighted result of multi-dimensional sensor features; The visual parameters are weighted with the visual weights, obtaining the weighted result of visual parameters; the cross-modal coupling strength values are weighted with the coupling term weights, obtaining the weighted result of cross-modal coupling strength values; the weighted result of multi-dimensional sensor features, the weighted result of visual parameters, and the weighted result of cross-modal coupling strength values are linearly superimposed to generate the fused monitoring data.

[0036] In the above solution, since the integration operation naturally suppresses high-frequency random interference, therefore, in the present invention, by exponentially decaying and weighting the time-varying gradient of sensor parameters, the normal process fluctuations (low low-frequency integral values) and abnormal mutations (high high-frequency integral values) can be effectively distinguished. The integral value is ahead of the absolute value of the parameter to reflect the change of the system state, so the sensitivity to the change trend of the parameter is improved. And, during the dynamic modulation of visual parameters, soft threshold filtering of noise is realized through hyperbolic tangent non-linear conversion to achieve element-level modulation and maintain the independent adjustment ability of different visual parameters. And a parameter pairing mechanism is constructed, corresponding combinations are made according to ways such as temperature and chromaticity, pH value and turbidity, etc., to reveal the internal relationship between sensor data and visual features, and the combined abnormality of temperature and chromaticity can be identified, and the ability to capture the composite features of the spice penetration process is improved. Finally, the three-dimensional weight space realizes the precise ratio of the contribution degrees of different data sources, and the linear superposition structure ensures the calculation efficiency and meets the real-time requirement. The constructed fused monitoring data breaks through the one-sided limitation of single-dimensional analysis.

[0037] Among them, in the preferred embodiment provided by the present invention, the establishment of a pairing relationship for each parameter in the multi-dimensional sensor features and visual parameters to obtain parameter pairs specifically includes the following steps: The multi-dimensional sensor features and visual parameters are synchronized on the time axis, obtaining a multi-source data sequence synchronized on the time axis; According to the three-dimensional coordinate model of the pot body, a mapping relationship between the positions of temperature sensors and image pixels is established, and according to the image ROI coordinates in the visual parameters, the readings of temperature sensors at corresponding spatial positions are matched from the multi-source data sequence synchronized on the time axis, obtaining a spatial parameter matrix with a spatial correspondence relationship; Retrieve the parameter association rules in the knowledge base of the marinating process, for example: Temperature↑ → Maillard reaction intensifies → Chromaticity value↓ (browning deepens); pH value↓ → Protein denaturation rate↑ → Liquid turbidity↑; Conductivity↑ → Ion concentration change → Change in surface bubble morphology (change in texture features); According to the parameter association rule, calculate the real-time Pearson correlation coefficient between the sensor part and the vision part in the multi-source data sequence synchronized with the time axis. Based on the real-time Pearson correlation coefficient, establish an association pairing to obtain a list of association parameter pairs with weight tags; Perform PCA dimensionality reduction on the multi-dimensional sensor features and vision parameters to obtain the dimensionality-reduced multi-dimensional sensor features and dimensionality-reduced vision parameters; In the three-dimensional principal component space, calculate the Mahalanobis distance between the dimensionality-reduced multi-dimensional sensor features and the dimensionality-reduced vision parameters to obtain a distance matrix of the similarity of multi-source data distribution; According to the distance matrix of the similarity of multi-source data distribution, use the Hungarian algorithm to solve the optimal pairing scheme for the list of association parameter pairs with weight tags. After optimization, obtain an optimized parameter association mapping table containing several parameter pairs.

[0038] Furthermore, the online monitoring method for stewed pig offal based on multi-source information fusion further includes the following steps: Step S104, perform online display on the fused monitoring data, and perform anomaly recognition and correlation verification to determine whether there is a definite anomaly.

[0039] In the embodiment of the present invention, obtain the online display address, transmit the fused monitoring data to the online display address, perform online display on the fused monitoring data, which is convenient for management personnel to intuitively view the fused monitoring data, and perform anomaly recognition on the fused monitoring data, extract the monitoring anomaly data, and then according to the monitoring anomaly data, match the monitoring-related data in the fused monitoring data. Through relevant analysis and recognition of the monitoring-related data, perform relevant verification on the monitoring anomaly data to determine whether there is a definite anomaly. Specifically, when the relevant verification is successful, it is determined that there is a definite anomaly; when the relevant verification fails, it is determined that there is no definite anomaly.

[0040] It can be understood that among temperature, pH, conductivity, gas, and image, when an item is abnormal, other items will be affected and there will be some corresponding changes. For example, when there is an anomaly of too high temperature, it will cause an increase in gas and a darker color of the stewed pig offal in the image, etc.

[0041] Specifically, Figure 5 Shows the flow chart of anomaly recognition and correlation verification in the method provided by the embodiment of the present invention.

[0042] Among them, in the preferred embodiment provided by the present invention, the step of performing online display on the fused monitoring data, performing anomaly recognition and correlation verification, and determining whether there is a definite anomaly specifically includes the following steps: Step S1041: Obtain the online display address; Step S1042: According to the online display address, perform online display on the fused monitoring data; Step S1043: Perform anomaly recognition on the fused monitoring data and extract the monitoring anomaly data; Step S1044: According to the monitoring anomaly data, match the monitoring-related data from the fused monitoring data; Step S1045: According to the monitoring-related data, perform relevant verification on the monitoring anomaly data to determine whether there is a definite anomaly.

[0043] Among them, in the preferred embodiment provided by the present invention, the relevant verification of the monitoring anomaly data according to the monitoring-related data to determine whether there is a definite anomaly specifically includes the following steps: Select historical parameter pairs that match the current raw material batch feature code; Calculate the variance of sensor values and the variance of visual parameters at each time point in the historical parameter pairs respectively; Calculate the covariance according to the variance of sensor values and the variance of visual parameters at each time point to obtain a dynamic correlation coefficient that changes with time, and convert the dynamic correlation coefficient into an attenuation coefficient to obtain a time series attenuation coefficient; Set the integration window length according to the current process stage, and perform exponential weighted integration operation on the time series attenuation coefficient within the window according to the integration window length, and then perform normalization to obtain the historical attenuation term quantization value; Calculate the dynamic association strength quantization index between sensor values and visual parameters according to the spatial parameter matrix to obtain a real-time correlation coefficient matrix; Perform singular value decomposition on the real-time correlation coefficient matrix, and take the largest singular value as the spectral radius, and record the change curve of the spectral radius with time to obtain the spectral radius time series change curve; Select several data points for spline interpolation on the spectral radius time series change curve, and calculate the first derivative at the current moment according to the interpolation result to obtain a derivative result; perform moving average filtering on the derivative result and output the smoothed dynamic gradient value; Obtain the distribution coordinate points of the sensor in the three-dimensional pot body model, and perform spatial interpolation on the visual parameters to generate a continuous distribution field, and fuse the distribution coordinate points with the continuous distribution field to form a three-dimensional fused data field; Calculate the parameter change rates in the transverse, longitudinal, and axial directions of each spatial unit in the three-dimensional fused data field; And take the vector norm of the parameter change rates in the transverse, longitudinal, and axial directions as the comprehensive gradient to statistically calculate the gradient average value of the entire three-dimensional fused data field to obtain the spatial fused field gradient intensity value; Fuse the historical decay term quantization value, the smoothed dynamic gradient value, and the spatial fusion field gradient intensity value to obtain a fusion result, perform logarithmic compression processing on the fusion result, and then map it to a standard decision interval to obtain a standardized coupling factor; Obtain the standardized coupling factor for the historical period, input the standardized coupling factor for the historical period into the LSTM model to predict the baseline value of the standardized coupling factor for the current process stage, and adjust the baseline value of the standardized coupling factor according to the current real-time production rhythm to obtain a dynamic decision threshold; Obtain the standardized coupling factor time series data according to the time sequence, and use the second derivative verification for the standardized coupling factor time series data to obtain an acceleration overrun flag; Compare the relative magnitudes of the standardized coupling factor and the dynamic decision threshold, and combine the acceleration flag to judge the abnormality persistence to obtain an abnormality decision result.

[0044] In the above solution, the present invention realizes a cross-modal collaborative verification mechanism. Through the real-time interaction verification of sensor data (temperature, pH, etc.) and visual parameters (chromaticity, turbidity, etc.), it breaks through the limitations of a single data source. In scenarios such as sensor drift or visual occlusion (such as steam interference), the system can still maintain a very high abnormality recognition accuracy. And the construction of the three-dimensional data field integrates the spatial distribution characteristics (such as the temperature gradient in different regions of the pot body) and the time evolution law (such as the parameter change curve during the stewing process), and can accurately locate local abnormalities (such as the bottom of the pot being burnt). And a multi-level filtering mechanism is designed to suppress accidental interference through the historical decay term and exclude instantaneous fluctuations through the second derivative verification, forming a double insurance.

[0045] Further, the online monitoring method for stewing based on multi-source information fusion further includes the following steps: Step S105, when a definite abnormality occurs, generate an abnormality adjustment instruction, and perform online abnormality adjustment on the stewing according to the abnormality adjustment instruction.

[0046] In the embodiment of the present invention, in the case of a definite abnormality, plan the abnormality adjustment for the monitored abnormal data, generate adjustment plan data, and then generate a corresponding abnormality adjustment instruction according to the adjustment plan data, and then perform online abnormality adjustment on the stewing according to the abnormality adjustment instruction to realize the automatic monitoring and control of the abnormality.

[0047] Specifically, Figure 6 The flowchart of performing online abnormality adjustment in the method provided by the embodiment of the present invention is shown.

[0048] Among them, in the preferred embodiment provided by the present invention, the step of generating an abnormality adjustment instruction when a definite abnormality occurs and performing online abnormality adjustment on the stewing according to the abnormality adjustment instruction specifically includes the following steps: Step S1051, when a definite anomaly exists, perform anomaly adjustment planning on the monitored anomaly data to generate adjustment planning data; Step S1052, generate an anomaly adjustment instruction according to the adjustment planning data; Step S1053, perform on-line anomaly adjustment on the marinated cooked food according to the anomaly adjustment instruction.

[0049] Among them, in the preferred embodiment provided by the present invention, the step of performing anomaly adjustment planning on the monitored anomaly data to generate adjustment planning data when a definite anomaly exists specifically includes the following steps: Obtain the monitored anomaly data, extract the numerical deviation, change rate, and spatial distribution characteristics of the anomaly parameters in the monitored anomaly data to obtain a structured anomaly feature vector; Obtain the historical adjustment case library, and screen and match cases according to the raw material RFID tag corresponding to the current anomaly to obtain historical case features; Respectively perform feature fusion on the structured anomaly feature vector and the historical case features by using a pre-trained dual-channel encoder to obtain a real-time feature vector and a historical case feature vector; Calculate the matrix similarity scores between the real-time feature vector and each historical case feature vector to generate a similarity score list; Obtain all process parameters at the current anomaly, and calculate the real-time deviation integral of each process parameter from the standard process curve to obtain a multi-dimensional process deviation quantization index; Perform non-linear compression processing on the deviation value of the multi-dimensional process deviation quantization index by using the hyperbolic tangent function to obtain a normalized penalty coefficient; Obtain the sensor energy consumption data and the production beat timing, and perform quality evaluation on the monitored shooting data at the current anomaly according to the standard monitored shooting data to obtain a visual quality evaluation result; Use the sensor energy consumption data, the production beat timing, and the visual quality evaluation result to perform multi-objective loss modeling to obtain a multi-objective loss value and the current process stage timing; Calculate the change rate of the multi-objective loss value with time, and perform exponential decay in combination with the current process stage timing to obtain a dynamic optimization weight coefficient; Sort the similarity score list in descending order, and select the historical cases corresponding to the top several similarity scores to form a candidate strategy pool; Establish a three-dimensional evaluation matrix according to the candidate strategy pool, the normalized penalty coefficient, and the dynamic optimization weight coefficient; Perform strategy sorting on the three-dimensional evaluation matrix to generate a strategy priority list, and select the top several from the strategy priority list as the adjustment planning data.

[0050] In the above solution, the present invention constructs a multi-stage dynamic evaluation system. By splitting a single anomaly judgment into three stages: feature embedding, offset evaluation, and optimization decision-making, compared with the traditional single-threshold alarm method, the false positive rate can be effectively reduced. And through the instant cross-validation of the current abnormal data with the historical case library, the problem of strategy failure caused by raw material batch differences is solved through feature space mapping technology. By adopting the hyperbolic tangent function to compress the process deviation amount, the frequent oscillation near the critical value of the linear penalty mechanism is avoided, and in the scenario of overcooked and burnt stew, the adjustment times can be effectively reduced. To ensure better effects, a dynamic optimization weight mechanism is set, which can automatically adjust the optimization weights of energy consumption, quality, and timeliness according to the production stage.

[0051] Furthermore, Figure 7 The application architecture diagram of the system provided by the embodiment of the present invention is shown.

[0052] Among them, in another preferred embodiment provided by the present invention, the online monitoring system for stew based on multi-source information fusion includes: The sensing online monitoring unit 101 is used to perform periodic sensing online monitoring on the stew according to a preset monitoring period and obtain sensing monitoring data.

[0053] In the embodiment of the present invention, the sensing online monitoring unit 101 periodically generates a temperature monitoring signal, a pH monitoring signal, a conductivity monitoring signal, and a gas monitoring signal according to a preset monitoring period. Then, according to the temperature monitoring signal, periodic temperature online monitoring is performed on the stew to obtain temperature monitoring data. According to the pH monitoring signal, periodic pH online monitoring is performed on the stew to obtain pH monitoring data. According to the conductivity monitoring signal, periodic conductivity online monitoring is performed on the stew to obtain conductivity monitoring data. According to the gas monitoring signal, periodic gas online monitoring is performed on the stew to obtain gas monitoring data. The temperature monitoring data, pH monitoring data, conductivity monitoring data, and gas monitoring data are comprehensively processed to generate sensing monitoring data.

[0054] Specifically, Figure 8 The structural block diagram of the sensing online monitoring unit 101 in the system provided by the embodiment of the present invention is shown.

[0055] Among them, in the preferred embodiment provided by the present invention, the sensing online monitoring unit 101 specifically includes: The temperature sensor 1011 is used to perform periodic temperature online monitoring on the stew according to a preset monitoring period and obtain temperature monitoring data; The pH sensor 1012 is used to perform periodic pH online monitoring on the stew according to a preset monitoring period and obtain pH monitoring data; A conductivity sensor 1013 is used to perform periodic on-line conductivity monitoring on the stewed cooked food according to a preset monitoring period, and obtain conductivity monitoring data; A gas sensor 1014 is used to perform periodic on-line gas monitoring on the stewed cooked food according to a preset monitoring period, and obtain gas monitoring data; A data integration module 1015 is used to integrate the temperature monitoring data, the pH monitoring data, the conductivity monitoring data and the gas monitoring data, and generate sensing monitoring data.

[0056] Furthermore, the on-line monitoring system for stewed cooked food based on multi-source information fusion further includes: An on-line monitoring photographing unit 102 is used to perform periodic on-line monitoring photographing on the stewed cooked food according to a preset monitoring period, and obtain monitoring photographing data.

[0057] In an embodiment of the present invention, the on-line monitoring photographing unit 102 periodically generates a monitoring photographing instruction according to a preset monitoring period, performs on-line monitoring photographing on the stewed cooked food according to the monitoring photographing instruction, obtains monitoring photographing images, and arranges a plurality of monitoring photographing images to generate monitoring photographing data.

[0058] A data fusion processing unit 103 is used to perform data cleaning, standardization processing and multi-source data fusion on the sensing monitoring data and the monitoring photographing data, and generate fusion monitoring data.

[0059] In an embodiment of the present invention, the data fusion processing unit 103 performs data cleaning on the sensing monitoring data and the monitoring photographing data to remove outliers and noise, generates sensing cleaning data and photographing cleaning data, then performs standardization processing on the sensing cleaning data and the photographing cleaning data according to a preset standardization reference information, generates sensing standard data and photographing standard data, extracts monitoring time series data from the sensing standard data and the photographing standard data, and then performs time series fusion processing on the sensing standard data and the photographing standard data according to the monitoring time series data to generate fusion monitoring data.

[0060] Specifically, Figure 9 FIG. shows a structural block diagram of the data fusion processing unit 103 in the system provided by the embodiment of the present invention.

[0061] Among them, in a preferred embodiment provided by the present invention, the data fusion processing unit 103 specifically includes: A data cleaning module 1031 is used to perform data cleaning on the sensing monitoring data and the monitoring photographing data to remove outliers and noise, and generate sensing cleaning data and photographing cleaning data; A normalization processing module 1032 for normalizing the sensed cleaning data and the captured cleaning data to generate sensed standard data and captured standard data; A timing extraction module 1033 for extracting monitored timing data from the sensed standard data and the captured standard data; A multi-source data fusion module 1034 for performing multi-source data fusion on the sensed standard data and the captured standard data according to the monitored timing data to generate fused monitoring data.

[0062] Furthermore, the stew online monitoring system based on multi-source information fusion further includes: An anomaly recognition and verification unit 104 for online displaying the fused monitoring data and performing anomaly recognition and related verification to determine whether there is a definite anomaly.

[0063] In an embodiment of the present invention, the anomaly recognition and verification unit 104 obtains an online display address, transmits the fused monitoring data to the online display address, and performs online display on the fused monitoring data, facilitating managers to intuitively view the fused monitoring data. Moreover, it performs anomaly recognition on the fused monitoring data, extracts monitored anomaly data, and then matches monitored related data from the fused monitoring data according to the monitored anomaly data. Through relevant analysis and recognition of the monitored related data, it performs relevant verification on the monitored anomaly data to determine whether there is a definite anomaly. Specifically, when the relevant verification is successful, it is determined that there is a definite anomaly; when the relevant verification fails, it is determined that there is no definite anomaly.

[0064] Specifically, Figure 10 FIG. shows a structural block diagram of the anomaly recognition and verification unit 104 in the system provided by an embodiment of the present invention.

[0065] Among them, in a preferred embodiment provided by the present invention, the anomaly recognition and verification unit 104 specifically includes: An address acquisition module 1041 for acquiring an online display address; An online display module 1042 for performing online display on the fused monitoring data according to the online display address; An anomaly recognition module 1043 for performing anomaly recognition on the fused monitoring data and extracting monitored anomaly data; A relevant matching module 1044 for matching monitored related data from the fused monitoring data according to the monitored anomaly data; A relevant verification module 1045 for performing relevant verification on the monitored anomaly data according to the monitored related data to determine whether there is a definite anomaly.

[0066] Furthermore, the stew online monitoring system based on multi-source information fusion further includes: The on-line anomaly regulation unit 105 is configured to generate an anomaly regulation instruction when a determined anomaly occurs, and perform on-line anomaly regulation on the marinated cooking according to the anomaly regulation instruction.

[0067] In an embodiment of the present invention, in the case of a determined anomaly, the on-line anomaly regulation unit 105 performs anomaly regulation planning on the monitored anomaly data to generate regulation planning data, and then generates a corresponding anomaly regulation instruction according to the regulation planning data, and further performs on-line anomaly regulation on the marinated cooking according to the anomaly regulation instruction, so as to achieve automatic monitoring and control of the anomaly.

[0068] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limitation, and these steps can be executed in other orders. Moreover, at least a part of the steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

[0069] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. The non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. The volatile memory can include random access memory (RAM) or an external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0070] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as falling within the scope described in this specification.

[0071] The above-described embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation to the scope of the patent for the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent for the present invention shall be subject to the appended claims.

[0072] The foregoing is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An online monitoring method for stewed pork offal based on multi-source information fusion, characterized in that, The method specifically includes the following steps: According to a preset monitoring period, perform periodic sensing on-line monitoring of the stewed pig offal to obtain sensing monitoring data; According to a preset monitoring period, perform periodic on-line monitoring and shooting of the stewed pig offal to obtain monitoring shooting data; Perform data cleaning, standardization processing, and multi-source data fusion on the sensing monitoring data and the monitoring shooting data to generate fused monitoring data; Perform on-line display on the fused monitoring data, and perform anomaly identification and related verification to determine whether there is a definite anomaly; When there is a definite anomaly, generate an anomaly adjustment instruction, and perform on-line anomaly adjustment on the stewed pig offal according to the anomaly adjustment instruction; The step of performing periodic on-line monitoring and shooting of the stewed pig offal according to a preset monitoring period to obtain monitoring shooting data specifically includes the following steps: Generate monitoring shooting instructions periodically according to a preset monitoring period; According to the monitoring shooting instructions, perform on-line monitoring and shooting of the stewed pig offal to obtain monitoring shooting images; Sort out multiple monitoring shooting images to generate monitoring shooting data.

2. The online monitoring method for stewed pork intestines based on multi-source information fusion according to claim 1, wherein The step of performing periodic sensing on-line monitoring of the stewed pig offal according to a preset monitoring period to obtain sensing monitoring data specifically includes the following steps: Perform periodic temperature on-line monitoring of the stewed pig offal according to a preset monitoring period to obtain temperature monitoring data; Perform periodic pH on-line monitoring of the stewed pig offal according to a preset monitoring period to obtain pH monitoring data; Perform periodic conductivity on-line monitoring of the stewed pig offal according to a preset monitoring period to obtain conductivity monitoring data; Perform periodic gas on-line monitoring of the stewed pig offal according to a preset monitoring period to obtain gas monitoring data; Integrate the temperature monitoring data, the pH monitoring data, the conductivity monitoring data, and the gas monitoring data to generate sensing monitoring data.

3. The online monitoring method of stewed pig offal based on multi-source information fusion according to claim 2, characterized in that, The step of performing data cleaning, standardization processing, and multi-source data fusion on the sensing monitoring data and the monitoring shooting data to generate fused monitoring data specifically includes the following steps: Perform data cleaning of removing outliers and removing noise on the sensing monitoring data and the monitoring shooting data to generate sensing cleaning data and shooting cleaning data; Perform standardization processing on the sensing cleaning data and the shooting cleaning data to generate sensing standard data and shooting standard data; Extract monitoring time series data from the sensing standard data and the shooting standard data; According to the monitoring time series data, perform multi-source data fusion on the sensing standard data and the shooting standard data to generate fused monitoring data.

4. The online monitoring method for stewed pork intestines based on multi-source information fusion according to claim 3, wherein, The step of performing multi-source data fusion on the sensing standard data and the shooting standard data according to the monitoring time series data to generate fused monitoring data specifically includes the following steps: Calculate the instantaneous change rate of each sensor parameter at the current moment according to the numerical difference between adjacent time points of the sensing standard data to obtain change rate data; Extract a key time window from the monitoring time series data according to the key reaction period of the stewed pig offal process, and perform exponentially decaying weighted integration on the change rate data within the key time window to obtain a weighted change rate integral value; Combine the integrated value of the weighted change rate with the sensing standard data to obtain multi-dimensional sensor features including the sensing standard data and time-varying features; Calculate the difference magnitude between the current image feature and the previous image feature in the shooting standard data to obtain a difference value; Perform a non-linear transformation on the difference value through the hyperbolic tangent function to calculate the modulation coefficient; Multiply the modulation coefficient element-wise with the image of the shooting standard data to suppress the visual interference of non-significant changes and obtain the visually parameter of adaptive modulation stabilization; Establish a pairing relationship for each parameter in the multi-dimensional sensor features and the visually parameter to obtain parameter pairs; Calculate the product of the parameters between the parameter pairs, and calculate the absolute value of the difference between the parameters between the parameter pairs. Combine the product of the parameters between the parameter pairs and the absolute value of the difference between the parameters between the parameter pairs to obtain a combined value; Filter the negative interference and retain the positive correlation features of the combined value through a non-linear activation function to obtain a pairing result; Sum all the pairing results to obtain the cross-modal coupling strength value; Calculate the variance statistical value of the multi-dimensional sensor features and calculate the change gradient value of the visually parameter; Generate a sensor weight value for the variance statistical value of the multi-dimensional sensor features using an exponential function, and set the corresponding visual weight according to the change gradient value of the visually parameter; Normalize and reconcile the sensor weight value and the visual weight to balance the contribution ratio of different data sources and obtain the coupling term weight; Weight the multi-dimensional sensor features with the sensor weight value to obtain the weighted result of the multi-dimensional sensor features; Weight the visually parameter with the visual weight to obtain the weighted result of the visually parameter; Weight the cross-modal coupling strength value with the coupling term weight to obtain the weighted result of the cross-modal coupling strength value; Linearly superimpose the weighted result of the multi-dimensional sensor features, the weighted result of the visually parameter, and the weighted result of the cross-modal coupling strength value to generate the fusion monitoring data.

5. The online monitoring method for stewed pig offal based on multi-source information fusion according to claim 4, characterized in that, The step of establishing a pairing relationship for each parameter in the multi-dimensional sensor features and the visually parameter to obtain parameter pairs specifically includes the following steps: Synchronize the multi-dimensional sensor features and the visually parameter on the time axis to obtain a multi-source data sequence synchronized on the time axis; According to the three-dimensional coordinate model of the pot body, establish a mapping relationship between the temperature sensor position and the image pixels, and match the temperature sensor readings at the corresponding spatial positions from the multi-source data sequence synchronized on the time axis according to the image ROI coordinates in the visually parameter to obtain a spatial parameter matrix with a spatial correspondence relationship; Retrieve the parameter association rules in the stewing process knowledge base; According to the parameter association rules, calculate the real-time Pearson correlation coefficient between the sensor part and the visual part in the multi-source data sequence synchronized on the time axis, and establish an association pairing according to the real-time Pearson correlation coefficient to obtain a list of association parameter pairs with weight tags; Perform PCA dimensionality reduction on the multi-dimensional sensor features and the visually parameter to obtain the multi-dimensional sensor features after dimensionality reduction and the visually parameter after dimensionality reduction; In the three-dimensional principal component space, calculate the Mahalanobis distance between the multi-dimensional sensor features after dimensionality reduction and the visually parameter after dimensionality reduction to obtain a distance matrix of the similarity of the multi-source data distribution; Based on the distance matrix of multi-source data distribution similarity, the Hungarian algorithm is used to solve the optimal pairing scheme for the list of associated parameter pairs with weight marks. After optimization, an optimized parameter association mapping table containing several parameter pairs is obtained.

6. The online monitoring method for stewed pork offal based on multi-source information fusion according to claim 5, characterized in that, The online display of the fusion monitoring data is performed, and anomaly recognition and related verification are carried out to determine whether there is a definite anomaly, which specifically includes the following steps: Obtain the online display address; According to the online display address, perform online display of the fusion monitoring data; Perform anomaly recognition on the fusion monitoring data and extract the monitoring anomaly data; According to the monitoring anomaly data, match the monitoring-related data from the fusion monitoring data; According to the monitoring-related data, perform relevant verification on the monitoring anomaly data to determine whether there is a definite anomaly.

7. The online monitoring method for stewed pig offal based on multi-source information fusion according to claim 6, characterized in that The relevant verification of the monitoring anomaly data according to the monitoring-related data to determine whether there is a definite anomaly specifically includes the following steps: Select historical parameter pairs that match the current raw material batch feature code; Calculate the variance of sensor values and the variance of visual parameters at each time point in the historical parameter pairs respectively; Calculate the covariance according to the variance of sensor values and the variance of visual parameters at each time point to obtain the dynamic correlation coefficient that changes with time, and convert the dynamic correlation coefficient into an attenuation coefficient to obtain the time series attenuation coefficient; Set the integral window length according to the current process stage, and perform exponential weighted integral operation on the time series attenuation coefficient within the window according to the integral window length, and then perform normalization to obtain the historical attenuation term quantization value; Calculate the dynamic association strength quantization index between sensor values and visual parameters according to the spatial parameter matrix to obtain the real-time correlation coefficient matrix; Perform singular value decomposition on the real-time correlation coefficient matrix, and take the largest singular value as the spectral radius, and record the change curve of the spectral radius with time to obtain the spectral radius time series change curve; Select several data points from the spectral radius time series change curve for spline interpolation, and calculate the first derivative at the current moment according to the interpolation result to obtain the derivative result; perform moving average filtering on the derivative result and output the smoothed dynamic gradient value; Obtain the distribution coordinate points of the sensor in the three-dimensional pot body model, and perform spatial interpolation on the visual parameters to generate a continuous distribution field, and fuse the distribution coordinate points with the continuous distribution field to form a three-dimensional fusion data field; Calculate the parameter change rates of each spatial unit in the three-dimensional fusion data field in the horizontal, vertical, and axial directions, and take the vector modulus of the parameter change rates in the horizontal, vertical, and axial directions as the comprehensive gradient to statistically calculate the gradient average value of the entire three-dimensional fusion data field to obtain the spatial fusion field gradient intensity value; Fuse the historical attenuation term quantization value, the smoothed dynamic gradient value, and the spatial fusion field gradient intensity value to obtain a fusion result, and perform logarithmic compression processing on the fusion result, and then map it to the standard decision interval to obtain the standardized coupling factor; Obtain the standardized coupling factor of the historical period, and input the standardized coupling factor of the historical period into the LSTM model to predict the baseline value of the standardized coupling factor in the current process stage, and adjust the baseline value of the standardized coupling factor according to the current real-time production rhythm to obtain the dynamic decision threshold; Obtain the standardized coupling factor time-series data according to the time sequence, and verify the standardized coupling factor time-series data using the second derivative to obtain the acceleration overrun flag; Compare the relative magnitudes of the standardized coupling factor and the dynamic determination threshold, and combine the acceleration flag to judge the abnormality persistence to obtain the abnormality determination result.

8. The online monitoring method for stewed pig offal based on multi-source information fusion according to claim 7, characterized in that, When there is a determined abnormality, generate an abnormality adjustment instruction, and according to the abnormality adjustment instruction, perform on-line abnormality adjustment on the marinated cooking specifically including the following steps: When there is a determined abnormality, perform an abnormality adjustment plan on the monitored abnormal data to generate adjustment plan data; Generate an abnormality adjustment instruction according to the adjustment plan data; According to the abnormality adjustment instruction, perform on-line abnormality adjustment on the marinated cooking.

9. The online monitoring method for stewed pig offal based on multi-source information fusion according to claim 8, wherein, When there is a determined abnormality, performing an abnormality adjustment plan on the monitored abnormal data to generate adjustment plan data specifically includes the following steps: Obtain the monitored abnormal data, extract the numerical deviation, change rate, and spatial distribution characteristics of the abnormal parameters in the monitored abnormal data to obtain a structured abnormal feature vector; Obtain the historical adjustment case library, and screen and match cases according to the raw material RFID tag corresponding to the current abnormality to obtain historical case characteristics; Perform feature fusion on the structured abnormal feature vector and the historical case characteristics respectively using a pre-trained dual-channel encoder to obtain a real-time feature vector and a historical case feature vector; Calculate the matrix similarity scores between the real-time feature vector and each historical case feature vector to generate a similarity score list; Obtain all process parameters at the current abnormality, and calculate the real-time deviation integral of each process parameter from the standard process curve to obtain a multi-dimensional process deviation quantification index; Perform non-linear compression processing on the deviation value of the multi-dimensional process deviation quantification index using the hyperbolic tangent function to obtain a normalized penalty coefficient; Obtain the sensor energy consumption data and the production beat timing, and perform quality evaluation on the monitored shooting data at the current abnormality according to the standard monitored shooting data to obtain the visual quality evaluation result; Use the sensor energy consumption data, the production beat timing, and the visual quality evaluation result to perform multi-objective loss modeling to obtain the multi-objective loss value and the current process stage timing; Calculate the change rate of the multi-objective loss value over time, and perform exponential decay in combination with the current process stage timing to obtain a dynamic optimization weight coefficient; Sort the similarity score list in descending order, and select the first several historical cases corresponding to the similarity scores to form a candidate strategy pool; Establish a three-dimensional evaluation matrix according to the candidate strategy pool, the normalized penalty coefficient, and the dynamic optimization weight coefficient; Perform strategy sorting on the three-dimensional evaluation matrix to generate a strategy priority list, and select the first several from the strategy priority list as the adjustment plan data.

10. An online monitoring system for stewed pig offal based on multi-source information fusion, the system is applied to the online monitoring method for stewed pig offal based on multi-source information fusion according to any one of claims 1 to 9, characterized in that, The system includes a sensing on-line monitoring unit, an on-line monitoring shooting unit, a data fusion processing unit, an abnormality identification and verification unit, and an on-line abnormality adjustment unit, where: The sensing on-line monitoring unit is used to perform periodic sensing on-line monitoring of the marinated cooking according to a preset monitoring period to obtain sensing monitoring data; The on-line monitoring shooting unit is used to perform periodic on-line monitoring shooting of the marinated cooking according to a preset monitoring period to obtain monitoring shooting data; A data fusion processing unit, which is used to perform data cleaning, standardization processing and multi-source data fusion on the sensing and monitoring data and the monitoring and shooting data to generate fused monitoring data; An abnormal recognition and verification unit, which is used to perform online display on the fused monitoring data, and perform abnormal recognition and related verification to determine whether there is a definite abnormality; An online abnormal adjustment unit, which is used to generate an abnormal adjustment instruction when there is a definite abnormality, and perform online abnormal adjustment on the marinated cooked food according to the abnormal adjustment instruction.

Images