Campus smart canteen management platform based on AI and IOT

Abstract

The invention discloses a campus smart canteen management platform based on AI and IOT. The campus intelligent canteen management platform comprises a global sensing system, an intelligent decision support system, a credible tracing system, a safety and supervision system and a user experience optimization system. The global sensing system is used for collecting multi-modal data and carrying out data fusion and preprocessing on the multi-modal data; the intelligent decision support system is used for performing multifunctional analysis on the fused and preprocessed multi-modal data to obtain a management decision; the credible traceability system is used for recording digital traceability information, storing reserved food samples and detecting pesticide residues in real time; the safety and supervision system is used for providing food reserved sample management, food quality event monitoring and analysis, supervision record management and parent applet functions; and the user experience optimization system is used for providing health inspection, health monitoring and personalized suggestions, identity verification and payment modes. In this way, the overall level of canteen management is improved.

Description

Technical Field

[0001] The present application relates to the field of canteen management technology, and in particular to a campus smart canteen management platform based on AI and IOT. Background Art

[0002] The current school catering management system still has room for improvement in the following technical aspects: Food quality traceability technology: The existing system is insufficient in terms of the integrity and real-time nature of food supply chain traceability information. More comprehensive data collection and more efficient information integration technologies are needed to achieve full-process traceability from the source of food to the table.

[0003] Food quality supervision technology: The existing system needs to be strengthened in terms of food quality risk warning and real-time monitoring. It is necessary to introduce more advanced sensor technology, image recognition technology and data analysis models to achieve timely detection and rapid response to food quality risks.

[0004] Nutrition and health management technology: The existing system does not fully analyze and utilize students' individual dining data. It is necessary to combine student health records and nutrition knowledge bases, use big data and artificial intelligence technology, and provide personalized dietary recommendations and nutritional guidance.

[0005] Intelligent operation and decision support technology: The existing system relies more on manual experience to make operational decisions. It is necessary to use data mining and machine learning technology to optimize resource allocation, predict dining demand, reduce food waste, and improve canteen operation efficiency.

[0006] System integration and data sharing: The data interoperability and collaborative working capabilities between the existing smart canteen subsystems (such as payment systems and ordering systems) need to be improved. It is necessary to establish unified data standards and interface specifications to form a complete, data-driven closed-loop management system.

[0007] User interaction technology: The information interaction between the existing system and students and parents is not sufficient. It is necessary to use technologies such as mobile Internet and information push to improve information transparency, meet users' right to know about food quality and nutritional information, and provide convenient feedback channels. Summary of the invention

[0008] This application provides a campus smart canteen management platform based on AI and IOT, which can improve the overall level of canteen management.

[0009] The present application provides a campus smart canteen management platform based on AI and IOT, which includes: a global perception system for collecting multimodal data, fusing the multimodal data to obtain target fusion data, and preprocessing the multimodal data to obtain target data; wherein the multimodal data includes: food traceability information, water quality monitoring data, energy consumption monitoring data, personnel health data, restaurant kitchen waste data, customer feedback data and inertial sensor data; an intelligent decision support system, which is connected to the global perception system for performing multifunctional analysis on the target fusion data and the target data to obtain corresponding management decisions; policies; management decisions include intelligent procurement management decisions, intelligent scheduling management decisions, intelligent dish recommendation decisions, intelligent health management decisions, and intelligent early warning management decisions; a trusted traceability system is used to record digital traceability information, food sample storage, real-time detection of pesticide residues, and morning inspection data recording and analysis; a safety and supervision system is used to provide food sample management, food quality event monitoring and analysis, supervision record management, parent app functions, background system management, and food inspection data recording; a user experience optimization system is used to provide hygiene inspections, health monitoring and personalized suggestions, identity authentication and payment methods, service experience, and information disclosure channels and content.

[0010] Among them, the global perception system includes: a food traceability information collection module, which is used to use IOT devices to automatically collect the source, manufacturer, and batch when the food is put into storage as the food traceability information; a water quality monitoring module, which is used to use water quality sensors to monitor the canteen water and obtain water quality monitoring data; an energy consumption monitoring module, which is used to use energy consumption sensors to monitor the energy consumption of kitchen equipment and obtain energy consumption monitoring data; a personnel health data collection module, which is used to collect personnel health data; a restaurant kitchen waste data collection module, which is used to use smart trash cans or weighing systems to record the amount and classification of restaurant kitchen waste and obtain restaurant kitchen waste data; a customer feedback collection module, which is used to collect customer feedback data; an inertial sensor data collection module, which is used to collect inertial sensor data using an inertial measurement unit; a data fusion and preprocessing module, which is used to fuse multimodal data to obtain target fusion data, and to preprocess multimodal data to obtain target data.

[0011] Among them, the data fusion and preprocessing module includes: a spatiotemporal data fusion unit, which is used to perform spatiotemporal association on multimodal data; an outlier detection and processing unit, which is used to automatically detect and process outliers in multimodal data using statistics, machine learning methods or local anomaly factor algorithms; a data enhancement unit, which is used to perform data enhancement on multimodal data; and a knowledge graph construction unit, which is used to construct a knowledge graph for multimodal data.

[0012] Among them, the spatiotemporal data fusion unit is also used to perform spatiotemporal association on multimodal data using the attention mechanism and the spatiotemporal dual encoding network.

[0013] Among them, outlier detection and processing is also used to delete, correct, retain and / or group outliers after they are detected.

[0014] Among them, the data enhancement unit is also used to perform geometric transformation, color transformation, add noise, and add bias to the multimodal data.

[0015] Among them, the intelligent decision support system includes: an intelligent procurement management decision module, which is used to evaluate and manage suppliers, dynamically price and negotiate, and manage multi-level inventory for target fusion data and target data; an intelligent scheduling management decision module, which is used to match employee skills, personalize scheduling preferences, and intelligently adjust emergencies for target fusion data and target data; an intelligent dish recommendation decision module, which is used to recommend seasonal ingredients, filter allergens and dietary taboos, and make interactive dish recommendations for target fusion data and target data; among them, the recommended dishes comply with dietary guidelines and nutritional balance principles; an intelligent health management decision module, which is used to analyze and generate nutritional components, generate labels for target fusion data and target data, evaluate and optimize dietary structure, predict health risks, and collect data from smart wearable devices; an intelligent early warning management decision module, which is used to perform multi-level early warnings, automatic analysis of early warning causes, system linkage responses, and receive customizable early warning rules for target fusion data and target data.

[0016] Among them, the trusted traceability system includes: a digital traceability information module for recording digital traceability information; a food sample storage module for automatic sample reminders, intelligent sample cabinet management, and association of sample information with traceability information; a real-time pesticide residue detection module for pesticide residue detection and linkage with the early warning system; a morning inspection data recording and analysis module for recording a variety of morning inspection item records, morning inspection data statistics and trend analysis, and integration with the parent communication platform.

[0017] Among them, the safety and supervision system includes: a food sample management module, which is used to conduct compliance audits on the sample retention process and real-time monitoring of the sample retention status; a food quality event monitoring and analysis module, which is used to perform event correlation analysis and automatically generate event reports; and supports multi-department collaborative processing; a supervision record management module, which is used for digital signatures and timestamps of supervision records, and supports the export and sharing of supervision records; a parent mini-program function module, which is used to provide the function of viewing food inspection reports, viewing kitchen operation live videos, and receiving personalized health advice; a background system management module, which is used for multi-dimensional data statistics and analysis, authority management and auditing; a food inspection data recording module, which is used to support the recording of multiple inspection items, abnormal inspection result alarms, and inspection data trend analysis.

[0018] Among them, the user experience optimization system includes: a hygiene inspection module, which is used to use AI to assist in hygiene inspections, visualize and track inspection results; a health monitoring and personalized recommendation module, which is used for dietary intake analysis, nutrient intake assessment and recommendations, and integration with exercise data; an identity authentication and payment method module, which is used to provide a contactless payment function and support multiple identity authentication methods; a service experience module, which is used to learn and predict taste preferences, recommend intelligent nutritional meal plans, and provide voice ordering and interaction, dish customization and DIY functions; an information disclosure channel and content module, which is used for multi-platform information synchronization, data visualization, push personalized information, and provide interactive information query functions.

[0019] The beneficial effects of this application are as follows: Different from the existing technology, this application provides a campus smart canteen management platform based on AI and IOT, which monitors all aspects of catering in real time through a global perception system, realizes intelligent management, and reduces manual intervention. Through multimodal data fusion analysis and intelligent early warning system, potential food quality risks are discovered and warned in time to reduce the probability of food quality incidents. Through user health data analysis and personalized recommendations, students are provided with dishes and dietary plans that are more in line with their nutritional needs, promoting the healthy growth of students. Through big data analysis and artificial intelligence technology, scientific decision-making basis is provided for canteen managers, management efficiency is improved, and operating costs are reduced. Through blockchain technology, the whole chain of food can be traced, so that schools, students and parents can clearly understand the source and safety information of food. Through the integration and analysis of data, data support is provided for the refined management of the canteen, and the management level is continuously improved. Through convenient services and personalized recommendations, a better and more comfortable dining experience is provided to improve the satisfaction of students and parents. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them: Figure 1 This is a structural diagram of an embodiment of a campus smart canteen management platform based on AI and IOT provided by this application; Figure 2 It is a structural diagram of an embodiment of a global perception system provided by the present application; Figure 3 It is a structural diagram of an embodiment of a data fusion and preprocessing module provided by the present application; Figure 4 It is a structural diagram of an embodiment of an intelligent decision support system provided by the present application; Figure 5 It is a structural diagram of an embodiment of an intelligent decision support system provided by the present application; Figure 6 It is a structural diagram of an embodiment of a safety and supervision system provided by the present application; Figure 7 It is a structural diagram of an embodiment of a user experience optimization system provided in this application. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It will be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only some but not all structures related to the present application are shown in the drawings. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0022] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0023] The current school catering management system still has room for improvement in the following technical aspects: Food quality traceability technology: The existing system is insufficient in terms of the integrity and real-time nature of food supply chain traceability information. More comprehensive data collection and more efficient information integration technologies are needed to achieve full-process traceability from the source of food to the table.

[0024] Food quality supervision technology: The existing system needs to be strengthened in terms of food quality risk warning and real-time monitoring. It is necessary to introduce more advanced sensor technology, image recognition technology and data analysis models to achieve timely detection and rapid response to food quality risks.

[0025] Nutrition and health management technology: The existing system does not fully analyze and utilize students' individual dining data. It is necessary to combine student health records and nutrition knowledge bases, use big data and artificial intelligence technology, and provide personalized dietary recommendations and nutritional guidance.

[0026] Intelligent operation and decision support technology: The existing system relies more on manual experience to make operational decisions. It is necessary to use data mining and machine learning technology to optimize resource allocation, predict dining demand, reduce food waste, and improve canteen operation efficiency.

[0027] System integration and data sharing: The data interoperability and collaborative working capabilities between the existing smart canteen subsystems (such as payment systems and ordering systems) need to be improved. It is necessary to establish unified data standards and interface specifications to form a complete, data-driven closed-loop management system.

[0028] User interaction technology: The information interaction between the existing system and students and parents is not sufficient. It is necessary to use technologies such as mobile Internet and information push to improve information transparency, meet users' right to know about food quality and nutritional information, and provide convenient feedback channels.

[0029] That is, the prior art has the following problems: How to achieve comprehensive and real-time perception of the entire catering process in school cafeterias (including food procurement, storage, processing, meal distribution, dining, and post-meal), and effectively integrate multi-source heterogeneous data.

[0030] How to achieve intelligent supervision of all aspects of catering without increasing personnel costs, and timely discover and warn of potential risks.

[0031] How to use intelligent analysis technology to analyze user health data and dining behaviors to provide data support for personalized dietary guidance and health management.

[0032] How to build a reliable food traceability system so that schools, students and parents can clearly understand the source and safety information of food.

[0033] How to use big data and artificial intelligence technologies to achieve intelligent decision-making and optimization of canteen operations, improve efficiency and reduce costs.

[0034] How to fully consider user experience in system design and provide convenient and personalized services.

[0035] Based on this, the present application proposes to monitor all aspects of catering in real time through a global perception system, realize intelligent management, and reduce manual intervention. Through multimodal data fusion analysis and intelligent early warning system, potential food quality risks can be discovered and warned in time to reduce the probability of food quality incidents. Through user health data analysis and personalized recommendations, students can be provided with dishes and dietary plans that are more in line with their nutritional needs to promote their healthy growth. Through big data analysis and artificial intelligence technology, scientific decision-making basis can be provided for canteen managers, management efficiency can be improved, and operating costs can be reduced. Through blockchain technology, the whole chain of food can be traced, so that schools, students and parents can clearly understand the source and safety information of food. Through the integration and analysis of data, data support is provided for the refined management of the canteen to achieve continuous improvement of management level. Through convenient services and personalized recommendations, a better and more comfortable dining experience is provided to improve the satisfaction of students and parents. Solve at least one of the above technical problems. For details, refer to any of the following embodiments or a combination of any embodiments.

[0036] See also Figure 1 , Figure 1 This is a schematic diagram of the structure of an embodiment of a campus smart canteen management platform based on AI and IOT provided by this application. The campus smart canteen management platform 100 includes: a global perception system 10, an intelligent decision support system 20, a trusted traceability system 30, a security and supervision system 40 and a user experience optimization system 50.

[0037] The global perception system 10 is used to collect multimodal data, fuse the multimodal data to obtain target fusion data, and preprocess the multimodal data to obtain target data; wherein the multimodal data includes: food traceability information, water quality monitoring data, energy consumption monitoring data, personnel health data, restaurant kitchen waste data, customer feedback data and inertial sensor data.

[0038] The intelligent decision support system 20 is communicatively connected with the global perception system 10, and is used to perform multifunctional analysis on the target fusion data and the target data to obtain corresponding management decisions; wherein, the management decisions include intelligent procurement management decisions, intelligent scheduling management decisions, intelligent dish recommendation decisions, intelligent health management decisions, and intelligent early warning management decisions.

[0039] The trusted traceability system 30 is used to record digital traceability information, store food samples, detect pesticide residues in real time, and record and analyze morning inspection data.

[0040] The safety and supervision system 40 is used to provide food sample management, food quality event monitoring and analysis, supervision record management, parent app functions, background system management, and food inspection data recording.

[0041] The user experience optimization system 50 is used to provide health checks, health monitoring and personalized recommendations, identity authentication and payment methods, service experience, information disclosure channels and content.

[0042] See also Figure 2 , Figure 2 It is a structural diagram of an embodiment of the global perception system provided by the present application. The global perception system 10 includes: food traceability information collection module 11, water quality monitoring module 12, energy consumption monitoring module 13, personnel health data collection module 14, restaurant kitchen waste data collection module 15, customer feedback collection module 16, inertial sensor data collection module 17 and data fusion and preprocessing module 18. This part aims to build a comprehensive, multi-modal data collection and fusion system to provide a solid data foundation for subsequent intelligent analysis and decision-making.

[0043] The food traceability information collection module 11 is used to use IOT devices to automatically collect the origin, manufacturer, batch, etc. of the food when it enters the warehouse as the food traceability information. It uses QR codes, RFID and other technologies to automatically collect information such as the origin, manufacturer, batch, etc. when the food enters the warehouse, and links with the traceability system to ensure that the food information is traceable.

[0044] The water quality monitoring module 12 is used to monitor the canteen water using water quality sensors to obtain water quality monitoring data. Water quality sensors such as pH value and residual chlorine are deployed to monitor the safety of canteen water in real time to ensure that it meets drinking water standards.

[0045] The energy consumption monitoring module 13 is used to monitor the energy consumption of kitchen equipment using energy consumption sensors to obtain energy consumption monitoring data. Energy consumption sensors such as electric meters, water meters, and gas meters are integrated to monitor the energy consumption of kitchen equipment and provide data support for energy saving optimization.

[0046] The personnel health data collection module 14 is used to collect personnel health data, integrate the data of the morning inspection system and the temperature monitoring equipment, grasp the health status of employees in real time, and prevent the spread of infectious diseases.

[0047] The kitchen waste data collection module 15 is used to use a smart trash can or a weighing system to record the amount of kitchen waste generated and the classification to obtain kitchen waste data. Using a smart trash can or a weighing system to record the amount of kitchen waste generated and the classification provides data support for reducing waste.

[0048] The customer feedback collection module 16 is used to collect customer feedback data, integrate customer feedback information from online evaluations, suggestion boxes and other channels, and understand user needs and satisfaction.

[0049] The inertial sensor data acquisition module 17 is used to collect inertial sensor data using an inertial measurement unit. Inertial measurement unit (IMU) data, such as accelerometer and gyroscope data, is used to monitor the behavior patterns of kitchen staff, for example, it can be used to analyze the standardization of chef operation steps.

[0050] The data fusion and preprocessing module 18 is used to perform data fusion on the multimodal data to obtain target fused data, and to perform preprocessing on the multimodal data to obtain target data.

[0051] See also Figure 3 , Figure 3 The data fusion and preprocessing module 18 includes: a spatiotemporal data fusion unit 181, an outlier detection and processing module 182, a data enhancement unit 183 and a knowledge graph construction unit 184.

[0052] The spatiotemporal data fusion unit 181 is used to perform spatiotemporal association on multimodal data. It associates the sensory data at different times and locations, for example, tracking the complete path of ingredients from the base to the table, or analyzing the behavior patterns of employees at different time periods. It uses the attention mechanism and the spatiotemporal dual encoding network to process these data.

[0053] In some embodiments, food traceability information collection uses technologies such as scanning QR codes and RFID to collect information such as the origin, manufacturer, and batch of food. This data can not only be used to trace food quality issues, but also to build an "attention map" of food ingredients, for example, focusing on high-risk ingredients or suppliers, and predicting possible problems in the future based on historical data.

[0054] Water quality sensors collect data such as pH value and residual chlorine to ensure drinking water safety. Water quality data is regarded as "environmental attention". When water quality is abnormal, the platform can issue an alarm and take action.

[0055] Integrate energy consumption sensors to monitor the energy consumption of kitchen equipment. Energy consumption data can be regarded as "attention to the operating status of equipment", and abnormal energy consumption may indicate equipment failure.

[0056] Integrate the data of the morning check-up system and temperature monitoring equipment to grasp the health status of employees in real time. The health data of employees can be regarded as a kind of "individual attention". When abnormal situations occur, the system can take corresponding measures to prevent the spread of infectious diseases.

[0057] Use smart trash cans or weighing systems to record the amount and classification of kitchen waste and analyze food waste. Garbage data can be regarded as "attention to consumer behavior" and can help optimize dishes and reduce waste.

[0058] Integrate customer feedback information from online reviews, suggestion boxes and other channels to understand user needs and satisfaction. Customer feedback is "user attention" and can help the platform quickly identify problems and make improvements.

[0059] Inertial sensor data integration can capture the movement and posture of personnel to analyze their work efficiency, standardize operations, and even identify potential safety risks.

[0060] Spatiotemporal data fusion can link sensory data from different times and locations. For example, by tracking the complete path of ingredients from the base to the table, we can simulate the "trajectory of ingredients" and use time series data to predict future health risks.

[0061] Automatically detect and process outliers in data using statistical or machine learning methods. Preprocess input data to improve data quality.

[0062] Expand the amount of data through technical means. For example, rotating and flipping images can enhance the training effect of AI models. The collected data is constructed into a knowledge graph, for example, linking ingredients, suppliers, nutritional content, user preferences, etc. This helps provide deeper insights.

[0063] Through the attention mechanism, a “risk attention map” is constructed: Build an "attention map" for each link of the health management platform. This attention map can dynamically highlight areas or factors with higher risks, such as: High-risk ingredients: Mark high-risk ingredients based on historical data and traceability information and increase inspection frequency.

[0064] High-risk personnel: Based on health data, employees with abnormal health conditions are marked and timely measures are taken.

[0065] High-risk operations: Mark employees’ working status based on their movements and postures to remind them to avoid wrong operations.

[0066] High-risk areas: Mark areas with poor cleanliness or sanitation conditions for increased supervision.

[0067] High-risk time periods: Adjust staffing according to peak dining times to avoid food quality risks.

[0068] The “risk attention map” is presented to managers in a visual way so that they can quickly identify and respond to risks.

[0069] Use spatial attention mechanisms to selectively focus on data: Just as the driver's attention mechanism selectively focuses on important areas in the driving scene, the health management platform can use spatial attention mechanisms to selectively focus on the most relevant health data. For example, when analyzing personnel health data, it prioritizes abnormal body temperature and respiratory symptoms rather than other minor indicators.

[0070] Enhanced perception efficiency: Improve the system's perception efficiency of health risks by prioritizing key data.

[0071] Utilize the time attention mechanism: Pay attention to the time series changes to show that the platform needs to focus on the dynamic changes of health data or employee work status. For example, if it is found that an employee's body temperature continues to rise over a period of time, or the employee operates abnormally, the platform should pay close attention and take timely action. Predict time trends: Based on historical data, predict future trends so that timely prevention and intervention can be carried out. For example, it can predict the risk of food poisoning or employee misoperation in a school cafeteria.

[0072] And apply the space-time encoder to build a "health management space-time encoder": various data collected by the health management platform are space-time encoded.

[0073] Feature fusion module: integrates data from different sources (such as food traceability, water quality monitoring, personnel health, employee work status, etc.) to form a comprehensive "health status feature".

[0074] Time coding module: Analyze the changes of these data over time to capture the dynamic patterns of health risks.

[0075] A feed-forward network (FFN) is used to project the fused features into a high-dimensional space.

[0076] Batch Normalization (BN) is used to standardize features and improve training stability.

[0077] Time series information modeling: Analyze the time series changes of data and use it for prediction.

[0078] Behavior prediction module: Based on the characteristics after spatiotemporal encoding, it predicts possible health risks (such as food poisoning, infectious diseases, employee misoperation, etc.) and provides corresponding early warning and decision support.

[0079] Use CNN to extract high-level features.

[0080] Adaptive Average Pooling and Fully Connected Layer (MLP) are used to output the predicted health risk.

[0081] The specific application scenarios are as follows: Food safety warning: “Attention Map”: Through food traceability information, high-risk ingredients are marked and their flow is continuously monitored. Spatiotemporal Coding: Integrate information such as the source, batch, production date, and transportation time of ingredients to analyze their potential risks. Behavior Prediction: Predict food quality risks and issue early warnings in a timely manner.

[0082] Personnel health monitoring: “Attention map”: Based on the morning checkup data, employees with abnormal health conditions are marked and given special attention. Spatiotemporal coding: Analyze the changing trends of employees’ body temperature, respiratory symptoms and other data to capture potential infectious disease risks. Behavior prediction: Predict the risk of infectious disease outbreaks and recommend isolation or disinfection measures. Misoperation prediction: Predict the risk of employee misoperation and issue early warnings.

[0083] Environmental sanitation supervision: “Attention map”: Mark areas with poor cleanliness or substandard sanitation conditions and conduct key monitoring. Spatiotemporal coding: Analyze the changes in environmental sanitation indicators over time to identify potential problem areas. Behavior prediction: Predict environmental sanitation risks and recommend cleaning or disinfection measures.

[0084] By integrating multimodal data, the health management platform can more effectively identify, predict and manage health risks. Among them: the attention mechanism helps the platform focus on key risk points. Spatiotemporal coding helps the platform understand the dynamic change pattern of risks. The prediction model can provide early warning and support decision-making.

[0085] This integrated solution can significantly improve the intelligence and preventive nature of the health management platform, ultimately ensuring the health and safety of students and faculty.

[0086] The outlier detection and processing module 182 is used to automatically detect and process outliers in multimodal data using statistics, machine learning methods or local outlier factor algorithms. Using statistics or machine learning methods, the local outlier factor algorithm (LOF) and its improved versions (ILOF, EILOF) are used to automatically detect and process outliers in data. For example, it can be used to detect anomalies in food purchase volume, energy consumption data, morning inspection data, etc. When processing data streams, the EILOF algorithm only calculates the LOF score of new data points, which improves calculation efficiency.

[0087] Outliers are data points that are significantly different from other data points in a data set. These data points may be due to measurement errors, data entry errors, fraudulent behavior, or truly unusual events. Outlier detection is an important step in data analysis, and its goal is to identify these data points that deviate from the normal pattern and deal with them appropriately.

[0088] Outlier detection methods can be divided into many types, mainly including: statistical methods and machine learning methods.

[0089] Statistical methods: Methods based on normal distribution: For example, the mean and standard deviation of the data are calculated, and data points outside a certain standard deviation range are considered outliers. Methods based on box plots: Quartiles and interquartile ranges are used to define the normal range of data, and data points outside the range are considered outliers. Methods based on distance: For example, the distance between data points is calculated, and data points that are too far away from other data points are considered outliers.

[0090] Machine learning methods: Clustering algorithms: For example, the K-Means algorithm can cluster data points into different clusters, and data points far away from the cluster center can be regarded as outliers. Classification algorithms: For example, support vector machines (SVM) can learn the normal pattern of data and classify data points that do not conform to the normal pattern as outliers. Density methods: For example, the local outlier factor (LOF) algorithm and its improved versions (ILOF, EILOF).

[0091] Local Outlier Factor (LOF) algorithm and its improved versions (ILOF, EILOF) LOF algorithm (Local Outlier Factor) Core idea: Calculate the Local Outlier Factor (LOF) score for each data point, which measures the difference in local density between the data point and its neighbors. If the local density of a data point is much smaller than the local density of its neighbors, the data point is considered an outlier.

[0092] Calculation steps: 1. Calculate the k-nearest neighbors of data point p.

[0093] 2. Calculate the local reachability density (LRD) of data point p. LRD can be regarded as an approximate kernel density estimate of point p.

[0094] 3. Calculate the LOF score of data point p. LOF(p) is the ratio of the average LRD of p’s k nearest neighbors to the LRD of p.

[0095] Advantages: It can effectively detect local outliers. Even if the data distribution is relatively uniform globally, LOF can still find outliers with low local density.

[0096] ILOF algorithm (Incremental Local Outlier Factor) Core idea: For data stream scenarios, an incremental LOF calculation method is proposed to avoid recalculating the LOF scores of all data points every time new data arrives.

[0097] Calculation steps: When a new data point arrives, only the LRD and LOF scores of the data points affected by the new data point are updated. The affected data points include: 1. The new data point itself.

[0098] 2. k-nearest neighbors of the new data point.

[0099] 3. Include the new data point in the data points of its k-nearest neighbors.

[0100] Advantages: Reduces computational complexity and improves the efficiency of online anomaly detection.

[0101] EILOF algorithm (Efficient Incremental Local Outlier Factor) Core idea: The ILOF algorithm is further optimized to only calculate the LOF scores of new data points without changing the LOF scores of existing data points.

[0102] Calculation steps: 1. Calculate the k-nearest neighbors of the new data point.

[0103] 2. Update the LRD and LOF scores of new data points only.

[0104] 3. Do not update the LOF score of any existing data points.

[0105] Advantages: It significantly reduces the computational cost and improves the efficiency of processing data streams. Since noise is common in data sets and the LOF algorithm is sensitive to parameter k, not updating the LOF scores of existing data points may improve detection performance in practical applications.

[0106] Specifically, the outlier processing method is as follows: After outliers are detected, they need to be properly processed. Common processing methods include: Deletion: If outliers are due to data entry errors or measurement errors, they can be directly deleted.

[0107] Correction: If the outliers are due to measurement error, you can try to correct them, for example, using interpolation or regression techniques to estimate their correct values.

[0108] Keep: If the outliers represent true abnormal events and are critical to the analysis goal, they can be kept and further analyzed.

[0109] Group treatment: In some cases, outliers can be grouped, for example, treating them as a special category.

[0110] Use robust methods: Use statistical or machine learning methods that are insensitive to outliers to reduce the impact of outliers.

[0111] For example, in the scenario of detecting abnormal food purchase volume, the food inventory information is automatically collected by scanning QR codes, RFID, etc., including food name, quantity, supplier, etc.

[0112] Anomalies: If the purchase volume of a particular ingredient is much higher than the historical average, or does not match the current menu, there may be an anomaly (e.g., a purchase error).

[0113] Processing: The system can use the EILOF algorithm to detect abnormal purchase volumes and issue an alarm to remind managers to verify the situation.

[0114] Energy consumption data anomaly detection scenario: By integrating energy consumption sensors, the energy consumption of kitchen equipment (such as electricity meters, water meters, and gas meters) can be monitored in real time.

[0115] Abnormal: If the energy consumption data for a certain period of time is significantly higher or lower than the normal level, there may be abnormal conditions (for example, equipment failure, improper human operation).

[0116] Processing: The system can use the ILOF or EILOF algorithm to identify energy consumption anomalies and automatically adjust the kitchen temperature and humidity, ventilation system, etc. to achieve refined environmental control, or trigger maintenance requests to achieve predictive maintenance of equipment.

[0117] Morning checkup data anomaly detection scenario: Integrate the data from the morning checkup system and temperature monitoring equipment to understand the health status of employees in real time.

[0118] Abnormal: If an employee's body temperature is abnormally high, there may be a risk of infectious diseases.

[0119] Processing: The system can detect employees with abnormal body temperature and take timely isolation measures to prevent the spread of infectious diseases.

[0120] Among them, the advantage of the above-mentioned EILOF is that when processing data streams, the EILOF algorithm only calculates the LOF scores of new data points without changing the scores of existing data points. This makes it more computationally efficient when processing real-time data streams such as food purchase volume, energy consumption data, morning inspection data, etc., and may provide better detection results.

[0121] The data enhancement unit 183 is used to perform data enhancement on the multimodal data. The image data volume is expanded by operations such as rotation, flipping, and adding noise. For example, the inertial sensor data of kitchen staff and the images in the kitchen monitoring video can be enhanced to improve the AI ​​model training effect. At the same time, the bias addition method can also be used to simulate sensor errors.

[0122] Data augmentation is an effective method to improve the performance of machine learning models on limited data sets. By performing various transformations on existing data, new samples with the same category labels as the original data are generated, thereby expanding the training data set. This helps the model learn more robust features, reduce the risk of overfitting, and improve the generalization ability of the model.

[0123] Data augmentation techniques are for inertial sensor data, but they are also applicable to image data. The following are several commonly used data augmentation techniques: geometric transformation, color transformation, adding noise, and adding bias.

[0124] Geometric transformations: Rotation: Rotate the image around the center point by a certain angle. This transformation can help the model recognize objects at different angles and enhance the model's rotation invariance. For example, for images in a kitchen surveillance video, you can rotate them at different angles to simulate the camera shooting at different angles.

[0125] Flipping: Flip the image horizontally or vertically. This transformation can enhance the model's ability to recognize the left-right or top-down symmetry of an object. It can simulate the situation where the camera is installed in a different position and the picture is reversed left and right.

[0126] Scaling: Enlarging or reducing the image. This transformation can help the model recognize objects of different sizes.

[0127] Cropping: Randomly crop a portion of the image. This transformation can simulate the situation where the object is in different positions in the picture, and it can also increase the diversity of training samples.

[0128] Color Transformation: Brightness Adjustment: Randomly adjust the brightness of the image.

[0129] Contrast Adjustment: Randomly adjust the contrast of the image.

[0130] Saturation Adjustment: Randomly adjust the color saturation of the image.

[0131] Color Jittering: Randomly change the color channels of an image.

[0132] These transformations can enhance the robustness of the model to changes in lighting conditions. For example, a kitchen surveillance video may have different light intensities at different time periods, and these color transformations can make the model better adapt to different lighting conditions.

[0133] Additive Noise: Add random noise to the image to simulate interference in real scenes. Noise is usually sampled from a Gaussian distribution and added to the pixel values ​​of the original image. This can improve the robustness and noise resistance of the model. For example, the noise in a camera sensor can be simulated to make the model better adapt to the image quality of the real world.

[0134] Additive Bias: Adding bias is mainly used to enhance inertial sensor data. However, in image data, it can be understood by analogy as adding a constant offset to the pixel value of the image. This can simulate the systematic error of the sensor in different situations. For example, a camera sensor may have a certain brightness offset. By adding a bias, the model can better adapt to this error.

[0135] Specific applications of data enhancement technology: Rotation: Use rotation matrices (T1, T2, T3) to define rotations about different axes. These rotation matrices can be applied to image data to rotate the image in different directions.

[0136] T1: Rotate π / 6 radians around the Z axis. T2: Rotate π / 6 radians around the X axis. T3: Rotate π / 6 radians around the Y axis. In practical applications, a suitable rotation angle can be selected according to specific circumstances, and a larger angle or a random angle rotation can be considered.

[0137] Additive Bias: In inertial sensor data augmentation, the bias is randomly sampled from a Gaussian distribution and added to the original data. In image data, this can be analogously understood as adding a fixed brightness offset to the entire image, which can be randomly sampled from a Gaussian distribution. For example, for images captured by a camera, the overall brightness offset caused by lighting changes or sensor errors can be simulated. The standard deviation of the bias is usually set according to different data, for example, 0.1 m / s² for accelerometer data and 0.001 rad / s for gyroscope data, and can also be adjusted according to the characteristics of the image data.

[0138] Additive Noise: Similar to bias, noise is randomly sampled from a Gaussian distribution and added to the original data. For image data, the random noise generated by the camera sensor can be simulated. The document mentions that the standard deviation of noise is usually set according to different data, such as 0.1 m / s² and 0.001rad / s for accelerometer and gyroscope data. It can also be adjusted according to the characteristics of the image data, for example, the statistical characteristics of the pixel value distribution can be used to set the standard deviation of noise.

[0139] The application of data enhancement in kitchen surveillance video is as follows: Enhancement of kitchen staff inertial sensor data: Add bias to capture their movements and postures. This data can be used to analyze their work efficiency, standardize operations, and even identify potential safety risks.

[0140] Image enhancement: Perform geometric transformations such as rotation, flipping, scaling, and cropping on images in kitchen surveillance videos to increase the diversity of training data and enable the model to recognize objects at different angles and locations.

[0141] Adjust the brightness, contrast, and color saturation of the image to enhance the model’s robustness to lighting changes.

[0142] Add Gaussian noise to simulate the noise in the camera sensor and improve the model's noise resistance.

[0143] Adding a bias simulates possible brightness shifts in camera sensors, making the model more resilient to such errors.

[0144] Video enhancement: For video data, in addition to enhancing each frame of the image, some timing enhancements can also be performed, such as fast forward, slow forward, time warp, etc., to increase the generalization ability of the model.

[0145] The impact of data enhancement on AI model training is as follows: Improve model performance: Data augmentation can significantly improve the generalization and robustness of the model. By increasing the diversity of training data, the model can learn more robust features and reduce its dependence on specific training samples, thereby improving performance on unseen data.

[0146] Reduce overfitting: Data augmentation can help the model better adapt to various situations and reduce the risk of overfitting, thereby improving the overall performance of the model.

[0147] Simulating real-world scenarios: By adding noise and bias, the errors that may occur in sensors in real-world scenarios can be simulated, thereby improving the applicability of the model in real environments.

[0148] Reduce data dependence: In many cases, it is very costly to obtain large-scale, high-quality data. Data augmentation technology can effectively utilize existing small-scale data to obtain model performance comparable to that trained with large-scale data.

[0149] The knowledge graph construction unit 184 is used to construct a knowledge graph for multimodal data. The collected data is constructed into a knowledge graph, for example, by associating ingredients, suppliers, nutritional components, user preferences, etc., to provide deeper insights.

[0150] Among them, the spatiotemporal data fusion unit 181 is also used to use the attention mechanism and the spatiotemporal dual encoding network to perform spatiotemporal association on multimodal data.

[0151] Among them, outlier detection and processing is also used to delete, correct, retain and / or group outliers after they are detected.

[0152] The data enhancement unit 183 is also used to perform geometric transformation, color transformation, add noise, and add bias to the multimodal data.

[0153] See also Figure 4 , Figure 4 It is a structural diagram of an embodiment of an intelligent decision support system provided by the present application. The intelligent decision support system 20 includes: an intelligent procurement management decision module 21, an intelligent scheduling management decision module 22, an intelligent dish recommendation decision module 23, an intelligent health management decision module 24 and an intelligent early warning management decision module 25. This part aims to use AI technology to conduct in-depth analysis of the collected data and provide intelligent support for the operation management and decision-making of the catering platform.

[0154] The intelligent procurement management decision module 21 is used to evaluate and manage suppliers, dynamically price and negotiate, and manage multi-level inventory for target fusion data and target data. Supplier evaluation and management can intelligently evaluate suppliers based on historical procurement data, quality inspection data, on-time delivery rate, etc., and optimize supplier selection. Dynamic pricing and negotiation can analyze market conditions and supplier quotations to provide intelligent pricing suggestions and negotiation support for procurement personnel. Multi-level inventory management realizes refined inventory management of raw materials, semi-finished products, and finished products to prevent expiration and shortage.

[0155] The intelligent scheduling management decision module 22 is used to match the target fusion data and the target data with employee skills, personalized scheduling preferences, and intelligently adjust emergencies. Employee skill matching is mainly based on the skills and qualifications of employees to more reasonably allocate work tasks. Personalized scheduling preferences mainly consider employees' personal preferences and rest needs to improve employee satisfaction. Intelligent adjustment of emergencies enables the system to automatically adjust the schedule in the event of emergencies such as staff absences to ensure that operations are not affected.

[0156] The intelligent dish recommendation decision module 23 is used to recommend seasonal ingredients, filter allergens and dietary taboos, and make interactive dish recommendations for the target fusion data and the target data; wherein the recommended dishes comply with the dietary guidelines and the principle of nutritional balance. Seasonal ingredient recommendations can take seasonal ingredients into consideration and recommend seasonal dishes based on the supply of seasonal ingredients. Combining nutritional ingredients with dietary guidelines ensures that the recommended dishes comply with dietary guidelines and the principle of nutritional balance. Supporting allergen and dietary taboo filtering can automatically filter ingredients and dietary taboos that users are allergic to when making recommendations. Interactive dish recommendations allow users to input taste preferences or health needs, and the system provides more accurate recommendations.

[0157] The intelligent health management decision module 24 is used to perform nutritional analysis and label generation, dietary structure evaluation and optimization suggestions, health risk prediction on the target fusion data and target data, and collect data from smart wearable devices. Nutritional analysis and label generation can automatically analyze the nutritional components of dishes and generate clear nutrition labels. Dietary structure evaluation and optimization suggestions can analyze students' long-term dietary structure and provide more comprehensive dietary optimization suggestions. Health risk prediction can predict potential health risks based on health data and dining data, and provide personalized intervention suggestions. Integration with smart wearable devices can integrate data from students' smart wearable devices to gain a more comprehensive understanding of students' health status.

[0158] The intelligent early warning management decision module 25 is used to perform multi-level early warning on the target fusion data and target data, automatically analyze the reasons for the early warning, respond to the system linkage, and receive customizable early warning rules. The multi-level early warning system can set early warnings of different levels and take different response measures according to the degree of risk. The automatic analysis of the reasons for the early warning enables the system to preliminarily analyze the reasons for the early warning, for example, analyze which supplier's ingredients have problems, and combine the fusion of spatiotemporal data to track the entire process of the ingredients, thereby helping managers to quickly locate the problem. The linkage response with other systems can automatically link other systems after the early warning is triggered, such as triggering camera recording, sending notifications, etc. And the customizable early warning rules allow managers to customize the early warning rules according to actual needs.

[0159] See also Figure 5 , Figure 5 It is a structural diagram of an embodiment of a trusted traceability system provided by the present application. The trusted traceability system 30 includes: a digital traceability information module 31, a food sample storage module 32, a pesticide residue real-time detection module 33, and a morning inspection data recording and analysis module 34. This part aims to build a reliable food traceability system to ensure that food quality information is transparent and traceable.

[0160] The digital traceability information module 31 is used to record digital traceability information. For example, key parameters of the production process: record key parameters such as temperature, duration, and operators during food processing. Logistics and transportation environment monitoring data: record information such as temperature, humidity, and GPS tracks during transportation. Digitization of quality inspection reports: convert paper quality inspection reports into electronic data and upload them to the blockchain to ensure that the information cannot be tampered with.

[0161] The food sample storage module 32 is used for automatic sample reminder, intelligent sample cabinet management, and association of sample information with traceability information. For example, automatic sample reminder: the system automatically reminds the sample personnel to keep samples according to the type of dishes and the production time. Intelligent sample cabinet management: use a sample cabinet with a sensor to automatically record the sample time and status. Association of sample information with traceability information: associate the sample information with the corresponding food traceability information for easy query.

[0162] The real-time pesticide residue detection module 33 is used for pesticide residue detection and linkage with the early warning system. Support for more types of pesticide residue detection: expand the capabilities of portable pesticide residue detection equipment to detect more types of pesticide residues. Visual presentation of detection data: intuitively display the detection data in the form of charts, etc. Linkage with the early warning system: automatically trigger an early warning when excessive pesticide residues are detected.

[0163] The morning checkup data recording and analysis module 34 is used to record various morning checkup items, morning checkup data statistics and trend analysis, and integrate with the parent communication platform. Supports various morning checkup item records: such as mental state, respiratory symptoms, etc. Morning checkup data statistics and trend analysis, analyzes morning checkup data, and finds potential risks in students' health status. Integration with the parent communication platform facilitates the school to notify parents of abnormal morning checkups in a timely manner.

[0164] See also Figure 6 , Figure 6 It is a structural diagram of an embodiment of the safety and supervision system provided by the present application. The safety and supervision system 40 includes: a food sample management module 41, a food quality event monitoring and analysis module 42, a supervision record management module 43, a parent applet function module 44, a background system management module 45 and a food detection data recording module 46. This part aims to build a safe and reliable supervision system to ensure the compliance operation of the catering platform.

[0165] The food sample management module 41 is used to conduct compliance review of the sample retention process and real-time monitoring of the sample retention status. The compliance review of the sample retention process can use AI to analyze the sample retention video to check whether the sample retention operation complies with the regulations. The real-time monitoring of the sample retention status can monitor the temperature, humidity and other environmental parameters of the sample retention cabinet.

[0166] The food quality event monitoring and analysis module 42 is used to perform event correlation analysis and automatically generate event reports; and supports multi-department collaborative processing. Event correlation analysis can automatically correlate relevant data before and after the event occurs to help analyze the cause of the event. Automatic generation of event reports can automatically generate detailed event reports to facilitate post-event investigation and summary. Support for multi-department collaborative processing is used to support different departments (such as canteens, school medical offices, and security offices) to collaboratively handle food quality events.

[0167] The supervision record management module 43 is used for the digital signature and timestamp of the supervision record, and supports the export and sharing of the supervision record. The digital signature and timestamp of the supervision record can ensure the authenticity and non-tamperability of the supervision record. Supporting the export and sharing of the supervision record can facilitate the supervision department to review.

[0168] The parent applet function module 44 is used to provide functions for viewing food inspection reports, viewing kitchen operation live video, and receiving personalized health advice. Viewing food inspection reports can disclose the inspection reports of pesticide residues, heavy metals, microorganisms, etc. of food. Viewing kitchen operation live video can further improve transparency. Receiving personalized health advice can receive personalized health advice based on the child's dining data and health data.

[0169] The backend system management module 45 is used for multi-dimensional data statistics and analysis, authority management and auditing. Multi-dimensional data statistics and analysis can provide richer data analysis reports to help managers understand the canteen operation status. Authority management and auditing can strictly control the access rights of different users and record the user's operation behavior.

[0170] The food test data recording module 46 is used to support multiple test item records, abnormal test result alarms, and test data trend analysis. It supports multiple test item records: for example, allergen detection, genetically modified detection, etc. The abnormal test result alarm can automatically alarm when the test result is abnormal. The test data trend analysis can analyze historical test data and assess food quality risks.

[0171] See also Figure 7 , Figure 7 : is a structural diagram of an embodiment of the user experience optimization system provided by the present application. The user experience optimization system 50 includes: a health inspection module 51, a health monitoring and personalized suggestion module 52, an identity authentication and payment method module 53, a service experience module 54 and an information disclosure channel and content module 55. This part aims to improve the satisfaction of users (teachers, students, parents) and enhance the overall user experience of the platform.

[0172] The sanitation inspection module 51 is used to use AI to assist in sanitation inspections, and to visualize and track inspection results. AI-assisted sanitation inspections can use image recognition technology to assist inspectors in conducting sanitation inspections, such as identifying floor stains, tableware cleanliness, etc. Visual display and tracking of inspection results can display inspection results in the form of charts, etc., and track the rectification of problems.

[0173] The health monitoring and personalized advice module 52 is used for dietary intake analysis, nutrient intake assessment and advice, and integration with exercise data. Dietary intake analysis can combine smart plate data to more accurately analyze students' dietary intake. Nutrient intake assessment and advice can assess students' nutrient intake and provide more specific nutritional advice. Integration with exercise data: If conditions permit, integrate students' exercise data to provide more comprehensive health management advice.

[0174] The identity authentication and payment method module 53 is used to provide a touchless payment function and supports multiple identity authentication methods. Touchless payment can be combined with technologies such as face recognition to achieve touchless payment and further improve dining efficiency. Multiple identity authentication methods are supported: for example, fingerprint recognition, iris recognition, etc.

[0175] The service experience module 54 is used for learning and predicting taste preferences, recommending intelligent nutritional meal plans, and providing voice ordering and interaction, and dish customization and DIY functions. Taste preference learning and prediction can learn the user's taste preferences by analyzing the user's dining history, and predict the dishes they may like in the future. Dish customization and DIY can provide limited dish customization options, such as selecting side dishes, flavors, etc. Intelligent nutritional meal plan recommendations can recommend more personalized nutritional meal plans based on the user's health data and taste preferences. Voice ordering and interaction can provide voice ordering functions to facilitate user operation.

[0176] The information disclosure channel and content module 55 is used for multi-platform information synchronization, data visualization display, push personalized information, and provide interactive information query function. Multi-platform information synchronization can synchronize public information on multiple platforms such as mobile phone APP, LED display screen, website, etc. Richer data visualization display can display operation data, traceability information, test reports, etc. in a more intuitive way. Interactive information query allows users to query specific information according to their needs. Push personalized information, such as pushing children's dining records and health advice to parents.

[0177] In summary, the application provides a campus smart canteen management platform based on AI and IOT, which monitors all aspects of catering in real time through a global perception system, realizes intelligent management, and reduces manual intervention. Through multimodal data fusion analysis and intelligent early warning system, potential food quality risks are discovered and warned in time to reduce the probability of food quality incidents. Through user health data analysis and personalized recommendations, students are provided with dishes and dietary plans that are more in line with their nutritional needs to promote the healthy growth of students. Through big data analysis and artificial intelligence technology, scientific decision-making basis is provided for canteen managers, management efficiency is improved, and operating costs are reduced. Through blockchain technology, the whole chain of food can be traced, so that schools, students and parents can clearly understand the source and safety information of food. Through the integration and analysis of data, data support is provided for the refined management of the canteen, and the management level is continuously improved. Through convenient services and personalized recommendations, a better and more comfortable dining experience is provided to improve the satisfaction of students and parents.

[0178] In the several embodiments provided in this application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device implementation described above is only illustrative, for example, the division of the modules or units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0179] If the integrated units in the above other embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, RandomAccess Memory), disk or optical disk and other media that can store program codes.

[0180] The above description is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A campus smart canteen management platform based on AI and IOT, characterized by: The campus smart canteen management platform includes: A global perception system, used to collect multimodal data, fuse the multimodal data to obtain target fusion data, and preprocess the multimodal data to obtain target data; wherein the multimodal data includes: food traceability information, water quality monitoring data, energy consumption monitoring data, personnel health data, restaurant kitchen waste data, customer feedback data and inertial sensor data; An intelligent decision support system, which is in communication connection with the global perception system, is used to perform multifunctional analysis on the target fusion data and the target data to obtain corresponding management decisions; wherein the management decisions include intelligent procurement management decisions, intelligent scheduling management decisions, intelligent food recommendation decisions, intelligent health management decisions, and intelligent early warning management decisions; Trusted traceability system, used to record digital traceability information, store food samples, detect pesticide residues in real time, and record and analyze morning inspection data; Safety and supervision system, used to provide food sample management, food quality event monitoring and analysis, supervision record management, parent app functions, backend system management, and food testing data records; User experience optimization system, used to provide health checks, health monitoring and personalized recommendations, identity authentication and payment methods, service experience, information disclosure channels and content.

2. The campus smart canteen management platform according to claim 1 is characterized in that: The global perception system includes: The food traceability information collection module is used to use IOT devices to automatically collect the origin, manufacturer, and batch of food when the food is put into storage as the food traceability information; A water quality monitoring module, used to monitor the canteen water using a water quality sensor to obtain the water quality monitoring data; An energy consumption monitoring module, used to monitor the energy consumption of kitchen equipment using energy consumption sensors to obtain the energy consumption monitoring data; A personnel health data collection module, used to collect the personnel health data; A food waste data collection module is used to use an intelligent trash can or a weighing system to record the amount and classification of food waste and obtain the food waste data; A customer feedback collection module, used to collect the customer feedback data; An inertial sensor data acquisition module, used to acquire the inertial sensor data using an inertial measurement unit; The data fusion and preprocessing module is used to perform data fusion on the multimodal data to obtain target fused data, and to preprocess the multimodal data to obtain target data.

3. The campus smart canteen management platform according to claim 2 is characterized in that: The data fusion and preprocessing module includes: A spatiotemporal data fusion unit, used for performing spatiotemporal correlation on the multimodal data; An outlier detection and processing unit, used to automatically detect and process outliers in the multimodal data using statistics, machine learning methods or local outlier factor algorithms; A data enhancement unit, used to perform data enhancement on the multimodal data; A knowledge graph construction unit is used to construct a knowledge graph for the multimodal data.

4. The campus smart canteen management platform according to claim 3 is characterized in that: The spatiotemporal data fusion unit is also used to perform spatiotemporal association on the multimodal data using an attention mechanism and a spatiotemporal dual encoding network.

5. The campus smart canteen management platform according to claim 3 is characterized in that: The outlier detection and processing is also used to delete, correct, retain and / or group the outliers after they are detected.

6. The campus smart canteen management platform according to claim 3 is characterized in that: The data enhancement unit is also used to perform geometric transformation, color transformation, add noise, and add bias on the multimodal data.

7. The campus smart canteen management platform according to claim 1 is characterized in that: The intelligent decision support system comprises: An intelligent procurement management decision module, used for supplier evaluation and management, dynamic pricing and negotiation, and multi-level inventory management of the target fusion data and the target data; An intelligent scheduling management decision module is used to match employee skills, personalized scheduling preferences, and intelligently adjust emergencies based on the target fusion data and the target data; An intelligent dish recommendation decision module is used to recommend seasonal ingredients, filter allergens and dietary taboos, and make interactive dish recommendations for the target fusion data and the target data; wherein the recommended dishes comply with dietary guidelines and nutritional balance principles; An intelligent health management decision module, which is used to perform nutritional component analysis and label generation, dietary structure assessment and optimization suggestions, health risk prediction, and collect data from smart wearable devices on the target fusion data and the target data; The intelligent early warning management decision module is used to perform multi-level early warning on the target fusion data and the target data, automatically analyze the early warning causes, respond to system linkage, and receive customizable early warning rules.

8. The campus smart canteen management platform according to claim 1, characterized in that: The credible traceability system includes: Digital traceability information module, used to record digital traceability information; Food sample storage module, used for automatic sample reminder, intelligent sample cabinet management, and association of sample information with traceability information; Real-time pesticide residue detection module, used for pesticide residue detection and linkage with early warning system; The morning checkup data recording and analysis module is used to record various morning checkup items, morning checkup data statistics and trend analysis, and integrate with the parent communication platform.

9. The campus smart canteen management platform according to claim 1, characterized in that: The safety and supervision system includes: Food sample management module, used to conduct compliance review of the sample retention process and real-time monitoring of the sample retention status; Food quality event monitoring and analysis module, used for event correlation analysis and automatic generation of event reports; and supports multi-department collaborative processing; Regulatory records management module, which is used for digital signature and timestamp of regulatory records, and supports export and sharing of regulatory records; The parent mini-program function module is used to provide functions such as viewing food inspection reports, viewing kitchen operation live video, and receiving personalized health advice; Backend system management module, used for multi-dimensional data statistics and analysis, authority management and auditing; The food testing data recording module is used to support the recording of multiple testing items, abnormal test result alarms, and test data trend analysis.

10. The campus smart canteen management platform according to claim 1, characterized in that: The user experience optimization system comprises: The health inspection module is used to use AI to assist health inspections and visualize and track inspection results; Health monitoring and personalized advice module for dietary intake analysis, nutrient intake assessment and advice, and integration with exercise data; The authentication and payment method module is used to provide a touchless payment function and support multiple authentication methods; Service experience module, which is used for learning and predicting taste preferences, recommending intelligent nutritional meal plans, and providing voice ordering and interaction, dish customization and DIY functions; Information disclosure channels and content modules are used for multi-platform information synchronization, data visualization, push of personalized information, and interactive information query functions.

Images