Food management method, apparatus, and system

Abstract

This application discloses a food ingredient management method, apparatus, and system. The method includes: acquiring the identification information of a target food ingredient stored in a food ingredient storage device; acquiring multi-dimensional state information related to the target food ingredient based on the identification information; inputting the multi-dimensional state information into a pre-trained shelf-life determination model to output the shelf-life information of the target food ingredient; determining the target management operation corresponding to the target food ingredient based on the shelf-life information, and executing the target management operation to manage the target food ingredient. This method can improve the accuracy of shelf-life prediction, making the prediction results highly consistent with the actual storage scenario and quality change patterns of the food ingredient, reducing the food safety risks caused by expired food ingredients; furthermore, it can automatically and adaptively complete the dynamic determination of shelf life without relying on manual operation, making it highly applicable. In addition, it can achieve comprehensive, refined, and differentiated management of food ingredients, adapting to various food ingredient storage devices and scenarios, and reducing management costs.

Description

Technical Field

[0001] This application relates to the field of smart home technology, and in particular to a food management method, device, and system. Background Technology

[0002] With the fast pace of life, families often buy large quantities of food at once and store them in the refrigerator. Due to the variety of food and the messy storage, it is easy to forget to eat them or miss the expiration date, resulting in food waste and even food safety issues.

[0003] While some solutions for assisting in food management have emerged, they generally have shortcomings. Regarding shelf-life management, most rely on fixed preset countdowns, resulting in a simplistic prediction method that struggles to adapt to the diverse usage patterns of different households. The predicted results often deviate from the actual edible period of the food, easily leading to misjudgments. Furthermore, shelf-life management primarily depends on time, lacking effective perception of the food's actual condition. It cannot promptly identify spoiled food before its preset expiration date, posing a safety hazard. There are no specific management rules for opened, halved, or leftover food, leading to a high waste rate for these items. Additionally, when an odor is detected in the refrigerator, users struggle to quickly pinpoint the specific spoiled food. Existing intelligent solutions often employ an integrated design, requiring users to replace the entire unit, making low-cost adaptation to existing traditional refrigerators difficult and hindering widespread adoption.

[0004] There is currently no effective solution to the problem of the inability to manage ingredients accurately, comprehensively, and at low cost. Summary of the Invention

[0005] The purpose of this application is to provide a food ingredient management method, apparatus, and system to solve at least one of the problems of being unable to manage food ingredients accurately, comprehensively, and at low cost.

[0006] To solve the above-mentioned technical problems, the first aspect of this specification provides a food ingredient management method applied to a food ingredient storage device, the method comprising: Obtain the identification information of the target food stored in the food storage device; Based on the identification information, multidimensional status information related to the target ingredient is obtained. The multidimensional status information includes at least two of the following: attribute information of the target ingredient, first environmental data of the environment in which the target ingredient is located, ingredient status data of the target ingredient, operating status data of the ingredient storage device, and second environmental data of the environment in which the ingredient storage device is located. The multidimensional state information is input into a pre-trained shelf-life determination model, which outputs the shelf-life information of the target food ingredient. Based on the shelf life information, determine the target management operation corresponding to the target ingredient, and execute the target management operation to manage the target ingredient.

[0007] In some embodiments of this specification, the attribute information includes at least one of the following: ingredient type, ingredient name, ingredient quantity, and basic shelf life of the ingredient; The food ingredient status data includes at least one of the following: food ingredient contact status, food ingredient freshness, food ingredient processing status, and food ingredient integrity. The first environmental data includes at least one of the following: temperature, humidity, odor, and storage location; The operating status data includes at least one of the following: the frequency of door opening and closing of the food storage device, the duration of door opening, the refrigeration efficiency or operating power, and the food stacking status. The second environmental data includes at least one of the following: temperature, humidity, geographic location, seasonal information, and climate type.

[0008] In some embodiments of this specification, the shelf-life determination model includes a first sub-model and a second sub-model; The multidimensional state information is input into a pre-trained shelf-life determination model, which outputs the shelf-life information of the target food ingredient, including: The attribute information and the first environmental data from the multidimensional state information are input into the first sub-model, and the initial shelf life of the target food ingredient is output. The initial shelf life, as well as the ingredient status data, operating status data, and second environment data from the multidimensional status information, are input into the second sub-model to output the shelf life information of the target ingredient. The shelf life information includes at least one of the following: target shelf life, target shelf life confidence level, and quality decay data. The quality decay data is used to characterize the relationship between the quality of the target ingredient and time.

[0009] In some embodiments of this specification, the confidence level of the target shelf life is determined by the following method: The second sub-model determines the data integrity score of the target ingredient based on the input data of the second sub-model; The second sub-model determines the matching degree between the input data and the training data of the second sub-model, and determines the model matching degree score of the target ingredient relative to the second sub-model based on the matching degree; The second sub-model obtains the shelf life data corresponding to the target ingredient data that matches the target ingredient from the ingredient database, and determines the synergy score between the target shelf life and the shelf life data; The second sub-model determines the confidence level of the target shelf life based on the data integrity score, the model matching score, the synergy score, and the preset weights of each score dimension.

[0010] In some embodiments of this specification, before inputting the initial shelf life, as well as the food ingredient status data, operational status data, and second environmental data from the multidimensional status information, into the second sub-model and outputting the shelf life information of the target food ingredient, the method further includes: Based on the attribute information in the multidimensional state information, the model parameters of the second sub-model are determined, and the second sub-model is adjusted based on the determined model parameters; The initial shelf life, along with the ingredient status data, operational status data, and second environmental data from the multi-dimensional status information, are input into the second sub-model to output the shelf life information of the target ingredient, including: The initial shelf life, along with the ingredient status data, operational status data, and second environmental data from the multidimensional status information, are input into the adjusted second sub-model to output the shelf life information of the target ingredient.

[0011] In some embodiments of this specification, after inputting the initial shelf life, as well as the food ingredient status data, operational status data, and second environmental data from the multidimensional status information, into the second sub-model and outputting the shelf life information of the target food ingredient, the method further includes: Obtain the shelf-life data corresponding to the target ingredient data that matches the target ingredient from the ingredient database; The target shelf life is corrected based on the shelf life data; the correction includes at least one of the following: The process involves: filtering the shelf-life data to find first shelf-life data that matches the geographical location and name of the target ingredient; correcting the target shelf-life based on the first shelf-life data; filtering the shelf-life data to find second shelf-life data that matches the name and storage device of the target ingredient; correcting the target shelf-life based on the second shelf-life data; and filtering the shelf-life data to find third shelf-life data that matches the name and target user of the target ingredient; and correcting the target shelf-life based on the third shelf-life data.

[0012] In some embodiments of this specification, the shelf-life information further includes the shelf-life confidence level at which the target shelf-life confidence level is located; Performing the target management operation to manage the target ingredients includes: Determine the execution method for the target management operation that matches the confidence level of the shelf life; The target management operation is performed based on the execution method to manage the target ingredients; the target management operation includes at least one of the following: displaying the freshness status of the target ingredients on the door display of the ingredient storage device, pushing early warning information to the user terminal corresponding to the ingredient storage device, generating and displaying the quality change trend information of the target ingredients, and updating the ingredient data associated with the target ingredients in the ingredient database.

[0013] In some embodiments of this specification, the method further includes: Obtain abnormal environmental data collected by the sensing module of the food storage device; Based on the location information of the sensing unit corresponding to the abnormal environment data, the abnormal environment data, and the attribute information of each food ingredient, candidate spoiled food ingredients are identified from the food ingredients stored in the food ingredient storage device. Based on the target shelf life and food status data of the candidate spoiled food ingredients, a target spoiled food ingredient is determined from the candidate spoiled food ingredients, so as to generate a spoilage warning information of the target spoiled food ingredient and push it to the user terminal that matches the food storage device.

[0014] A second aspect of this specification provides a food management device for use in a food storage device, the device comprising: The first acquisition module is used to acquire the identification information of the target food stored in the food storage device; The second acquisition module is used to acquire multi-dimensional status information related to the target ingredient based on the identification information. The multi-dimensional status information includes at least two of the following: attribute information of the target ingredient, first environmental data of the environment in which the target ingredient is located, ingredient status data of the target ingredient, operating status data of the ingredient storage device, and second environmental data of the environment in which the ingredient storage device is located. The determination module is used to input the multidimensional state information into a pre-trained shelf-life determination model and output the shelf-life information of the target food ingredient. The operation module is used to determine the target management operation corresponding to the target ingredient based on the shelf life information, and execute the target management operation to manage the target ingredient.

[0015] A third aspect of this specification provides a food management system, including: a food storage device, a sensing module disposed in the food storage device, and a management module; The sensing module is used to collect multi-dimensional sensing data of the target ingredients stored in the food storage device, and send the multi-dimensional sensing data to the management module; The management module is used to: process the multidimensional sensing data to obtain multidimensional state information related to the target ingredient, wherein the multidimensional state information includes at least two of the following: attribute information of the target ingredient, first environmental data of the environment in which the target ingredient is located, ingredient state data of the target ingredient, operating state data of the ingredient storage device, and second environmental data of the environment in which the ingredient storage device is located; input the multidimensional state information into a pre-trained shelf-life determination model and output shelf-life information of the target ingredient; determine the target management operation corresponding to the target ingredient based on the shelf-life information, and execute the target management operation to manage the target ingredient.

[0016] Based on the food ingredient management method, apparatus, and system provided in the embodiments of this specification, the following steps are taken: 1) Obtain the identification information of the target food ingredient stored in the food ingredient storage device; 2) Obtain multi-dimensional state information related to the target food ingredient based on the identification information, the multi-dimensional state information including at least two of the following: attribute information of the target food ingredient, first environmental data of the environment in which the target food ingredient is located, food ingredient state data of the target food ingredient, operating state data of the food ingredient storage device, and second environmental data of the environment in which the food ingredient storage device is located; 3) Input the multi-dimensional state information into a pre-trained shelf-life determination model, and output the shelf-life information of the target food ingredient; 4) Determine the target management operation corresponding to the target food ingredient based on the shelf-life information, and execute the target management operation to manage the target food ingredient. In the embodiments of this specification, by fusing multidimensional state information covering at least two categories from dimensions such as target ingredient attributes, storage environment, and operating status, the shelf-life information corresponding to the target ingredient is obtained. Based on this shelf-life information, corresponding target management operations are performed, which improves the accuracy of shelf-life prediction. This ensures that the prediction results closely match the actual storage scenario and quality change patterns of the ingredients, reducing the food safety risks caused by expired ingredients. Furthermore, based on a pre-trained shelf-life determination model, the dynamic determination of shelf-life can be completed automatically and adaptively without relying on manual operation, making it highly applicable. In addition, by combining shelf-life prediction, operational decision-making, and automatic execution, comprehensive, refined, and differentiated management of ingredients can be achieved without structural modifications to existing ingredient storage devices. This allows for adaptation to various ingredient storage devices and scenarios, reducing ingredient management costs. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1The diagram shown is a schematic representation of a food ingredient management method provided in an embodiment of this specification. Figure 2 The diagram shown is a schematic representation of a method for determining shelf life provided in an embodiment of this specification. Figure 3 The diagram shown is a schematic representation of a shelf-life correction method provided in an embodiment of this specification. Figure 4 The diagram shown is a schematic representation of a method for determining the confidence level of the target shelf life provided in an embodiment of this specification. Figure 5 The diagram shown is a schematic representation of a method for determining target spoiled food provided in an embodiment of this specification. Figure 6 The diagram shown is a schematic of a compliance verification method provided in an embodiment of this specification. Figure 7 The diagram shown is a schematic of an electronic device provided in an embodiment of this specification. Detailed Implementation

[0019] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0020] It should be noted that the information and data related to users involved in the embodiments of this specification are all information and data authorized by the user or fully authorized by the relevant parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of the relevant data all comply with relevant laws, regulations, and standards, and necessary confidentiality measures have been taken. They do not violate public order and good morals, and corresponding operation entry points are provided for users or relevant parties to choose to authorize or refuse.

[0021] It should also be noted that in the embodiments of this specification, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0022] The food management method provided in the embodiments of this specification will be described below with reference to the accompanying drawings. It is understood that the methods described in the embodiments of this specification can be applied to food storage devices, but this application is not limited to execution by the processor of the food storage device itself. The method can be executed by a hardware module independent of the food storage device (such as a smart speaker, smart control box, user terminal, or cloud server, etc.). This hardware module can achieve the technical solutions in the embodiments of this specification by interacting with the food storage device (including but not limited to acquiring sensor unit data, sending display commands, and pushing early warning information, etc.). Any technical solution capable of acquiring the identification information and multi-dimensional status information of the target food, and determining the shelf life information and performing management operations accordingly, falls within the protection scope of this invention. The food storage device can be, for example, a refrigerator, freezer, wine cabinet, fresh-keeping cabinet, smart storage cabinet, smart fresh-keeping box, etc. In the following embodiments, a refrigerator is used as an example for description.

[0023] Figure 1 The diagram illustrates a food ingredient management method provided in an embodiment of this specification. While this specification provides method operation steps or apparatus structures as shown in the following embodiments or figures, the method or apparatus may include more or fewer operation steps or module units, either combined or integrated, based on conventional or non-inventive effort. In steps or structures where there is no logically necessary causal relationship, the execution order of these steps or the module structure of the apparatus is not limited to the execution order or module structure shown in the embodiments or figures of this specification. When the method or module structure is applied in actual devices, servers, or terminal products, it can be executed sequentially or in parallel according to the method or module structure shown in the embodiments or figures (e.g., in a parallel processor or multi-threaded processing environment, or even in a distributed processing or server cluster implementation environment). Figure 1 As shown, it may include: S101: Obtain the identification information of the target food stored in the food storage device.

[0024] The target food items can be any storage compartment in the refrigerator, such as the refrigerator compartment, variable temperature compartment, or freezer compartment, or food items stored at room temperature outside the refrigerator. The identification information for the target food items can be unique identification information, which may include, but is not limited to, food name, unique food ID, category code, batch information, and user's estimated shelf life.

[0025] In practice, when a user opens the refrigerator door, the door's open / close sensor detects the door opening and switches the voice recognition module in the food management system to a high-sensitivity pickup mode. The user can activate the voice interaction system using a preset wake-up word, issuing a fixed-format command, such as "Wake-up word: Add 5 Fuji apples, shelf life 7 days." The voice recognition module collects voice signals through a directional microphone, first undergoing hardware anti-interference processing (filtering low-frequency noise from the refrigerator compressor using a low-frequency noise filtering circuit), then software noise reduction (spectral subtraction + wavelet denoising algorithm), and finally parsing the command through a pre-trained voice recognition model to extract identifying information such as the food name, quantity, and the user's estimated shelf life, generating a unique ID for each food item. Therefore, during food management, the system can obtain the identification information for each item and, based on this information, obtain multi-dimensional status information for subsequent food management.

[0026] In other embodiments, users can also manually input food identification information through the food input interface on the touch screen display set on the refrigerator; or scan the product barcode, production date, and shelf life on the food packaging through the user terminal's APP, and the system will automatically parse and generate identification information.

[0027] S102: Based on the identification information, obtain multi-dimensional status information related to the target ingredient, the multi-dimensional status information including at least two of the following: attribute information of the target ingredient, first environmental data of the environment in which the target ingredient is located, ingredient status data of the target ingredient, operating status data of the ingredient storage device, and second environmental data of the environment in which the ingredient storage device is located.

[0028] Multidimensional state information can be linked data for multiple dimensions that affect the shelf life and freshness decay of target ingredients. It can cover the inherent properties of the ingredients themselves, the storage microenvironment, the operating status of equipment, the external environment, and other dimensions. Based on multidimensional state information, accurate prediction of shelf life can be achieved.

[0029] Multidimensional status information can be obtained from a pre-stored food database based on the identification information of the target food, or it can be obtained by processing multidimensional sensing data collected by the refrigerator's sensing module through methods such as data preprocessing, feature extraction, and data conversion. After data processing, the multidimensional sensing data can be stored in the food database for easy management of the food in the future.

[0030] In some embodiments of this specification, the multidimensional state information may be obtained by processing multidimensional sensing data collected by a sensing unit in the sensing module of the food storage device that matches the target food. The sensing unit may include at least one of the following: a temperature sensing unit, a humidity sensing unit, a distance sensing unit, an odor sensing unit, a door opening / closing sensing unit, and a whole-machine temperature control sensing unit.

[0031] In some embodiments of this specification, the attribute information may include at least one of the following: ingredient type, ingredient name, ingredient quantity, and basic shelf life of the ingredient. Ingredient type can be a category classified according to the quality decay characteristics of the ingredient, such as perishable (fresh fruits and vegetables, raw meat, seafood, fresh milk, etc.), moderately perishable (cooked food, processed meat, yogurt, soy products, etc.), and shelf-stable (dried goods, seasonings, frozen meat, canned food, etc.); the basic shelf life of the ingredient refers to the safe consumption period of the ingredient under standard storage conditions.

[0032] The food ingredient status data may include at least one of the following: food ingredient contact status, food ingredient freshness, food ingredient processing status, and food ingredient integrity. Food ingredient contact status refers to whether the food ingredient was stored with raw meat, seafood, or other ingredients prone to microbial growth. Food ingredient integrity refers to whether the food ingredient is in a complete, unopened state, or in a cut, opened, or remaining, incomplete state.

[0033] The first environmental data may include at least one of the following: temperature, humidity, odor, and storage location. The storage location may be a storage compartment in the refrigerator where the food is stored.

[0034] The operational status data may include at least one of the following: the frequency of door opening and closing of the food storage device, the duration of door opening, the cooling efficiency or operating power, and the food stacking status. The food stacking status can be the food stacking density within the food storage zone, which can be determined based on the proportion of the zone's effective volume.

[0035] The second environmental data may include at least one of the following: temperature, humidity, geographic location, seasonal information, and climate type.

[0036] Specifically, the sensing module can be a mesh structure. Sensing units can be deployed in various compartments of the refrigerator based on their acquisition accuracy and sensing range. Each sensing unit is bound to its physical location. Each compartment can include multiple sensing units that collect data from different dimensions. Furthermore, based on the characteristics of the food stored in different compartments, high-priority matching units and auxiliary matching units can be set for each compartment. These compartments can be further subdivided into areas such as the refrigerator compartment, variable temperature compartment, and freezer compartment. For example, in special scenarios such as incomplete food or mixed raw and cooked food, after the user marks the specific storage location of the target food via voice / touchscreen, the system sets the sensing unit closest to that location as the high-priority core matching unit, and the other sensing units in the same compartment as auxiliary matching units. For instance, if the user places a cut watermelon on the left side of the upper shelf in the refrigerator compartment, the system sets the temperature and odor sensing units on the left side of that shelf as core matching units, prioritizing their data collection to further improve the data's relevance to the target food.

[0037] In cases where there is no data output due to sensor unit failure, sensor data from adjacent partitions of the same type can be used to supplement the data by interpolation with the overall temperature control data of the refrigerator. At the same time, the confidence level of the data in this dimension is marked, and maintenance prompts are pushed to users through apps to avoid data dimension loss due to a single sensor failure.

[0038] In practical implementation, the temperature sensing unit can employ a micro-temperature sensor. After dividing the storage compartments, one sensor can be deployed in pairs on the inner wall of each compartment to collect the micro-environmental temperature of the corresponding compartment. The system can preset the sampling frequency of the sensor and set different sampling frequencies for different application scenarios (such as after the refrigerator door is opened or closed, or when the refrigerator door is continuously closed). The data collected by the temperature sensing unit can be processed through outlier filtering, smoothing filtering, and eigenvalue calculation, and then stored as at least one dimension of the multi-dimensional state information (such as temperature data, temperature stability data, and cooling efficiency in the first environmental data).

[0039] In practice, humidity sensing units can be paired with temperature sensing units in the same zone for deployment, and their acquisition frequency can be synchronized with that of the temperature sensing units. The data collected by the humidity sensing units can undergo condensation anomaly correction (for example, if humidity data at five consecutive collection points is ≥95% RH, it can be determined that condensation may occur within the zone; dew point calculation can be performed using synchronized temperature data from the same zone to eliminate false full-scale data caused by condensation), drift calibration, eigenvalue calculation, etc., to extract at least one dimension of the multidimensional state information (such as humidity data from the first environmental data) for storage.

[0040] In practical implementation, the ranging sensing unit can use an infrared ranging sensor. One sensor can be centrally located at the top of the inner wall of each storage compartment to collect the stacking height of the food within the compartment vertically downwards, and the collection frequency can be preset. The data collected by the ranging sensing unit can be filtered, stacking density calculated, and density levels classified (e.g., three levels can be preset: low density (≤30%), medium density (30%-60%), and high density (>60%). High-density stacking can lead to poor air circulation, and the shelf life of the food will be reduced proportionally (e.g., 10%-20% reduction for high-density stacking)). At least one dimension of the multi-dimensional status information (e.g., food stacking status data from operational status data) is then extracted and stored. The stacking density can be calculated based on the effective ranging value: food stacking volume = compartment length × compartment width × (total compartment height - measured ranging value), and then the volume percentage is calculated as: food stacking volume / total compartment volume.

[0041] In practical implementation, the odor sensing unit can be, for example, a semiconductor odor sensor. It can be deployed according to refrigerator zones, installed in areas with good airflow (below shelves, inside drawer walls, etc.), avoiding direct obstruction by food, to identify characteristic gases such as ammonia and hydrogen sulfide released during food spoilage. The sensor unit can be preset with a fixed sampling frequency, and different sampling frequencies can be preset for different states of food. The data collected by the odor sensing unit can be processed through temperature drift compensation, baseline calibration, cross-interference filtering, and feature value calculation to extract at least one dimension of the multi-dimensional state information (e.g., odor data from the first environmental data) for storage.

[0042] In practical implementation, one door opening / closing sensor unit can be deployed on each door of each storage compartment of the refrigerator to detect the door's open / closed status, opening time, and closing time in real time. Real-time data acquisition is triggered when the door status changes, recording the start time, end time, and duration of each door opening. The data collected by the door sensor unit can undergo invalid event filtering, statistical feature calculation, and influence coefficient conversion (e.g., converting the frequency and duration of door openings into a temperature fluctuation influence coefficient; the more frequent and longer the opening, the higher the coefficient). After these processes, at least one dimension of the multi-dimensional status information (e.g., door opening / closing frequency and opening duration data from the operating status data) is extracted and stored.

[0043] In practical implementation, the whole-machine temperature control sensor unit can interface with the refrigerator's own temperature control system to collect the refrigerator's set temperature, actual cooling power, operating current, and the deviation between the set temperature and the actual temperature. The sampling frequency can also be preset. The data collected by this sensor unit can eliminate abnormal data with fluctuating power, calculate the cooling power deviation and temperature control deviation, and convert them into a cooling stability coefficient (the larger the deviation, the worse the cooling stability and the shorter the shelf life of food). At least one dimension of the multi-dimensional state information is extracted and stored.

[0044] Some dimensions of the multidimensional state information can be obtained by connecting to a meteorological platform. For example, real-time temperature, relative humidity, seasonal information, climate type, and geographical location data of the user's location can be obtained. The meteorological data can be standardized and converted to match the shelf-life correction benchmarks of food products in different regions and seasons in the food database. After processing, the temperature, humidity, geographical location, seasonal information, and climate type data of the second environmental data can be output as multidimensional state information.

[0045] Some dimensions of the multidimensional status information can be determined by pre-entered ingredient attribute data or user input data when storing ingredients. For example, regarding the contact status of ingredients, a voice interaction system can provide auxiliary questions when the user enters the ingredients (such as "Should it be stored with raw meat / seafood?"). After the user confirms via voice, the system marks the risk factor of mixed storage, and the shelf life of ingredients stored with raw and cooked ingredients is reduced by 5%-15%, in line with food safety storage requirements. For example, regarding the initial freshness of ingredients, a voice command for freshness selection can be provided when the user enters the ingredients via voice (such as "Wake-up word, add strawberries, shelf life 5 days, freshness excellent / good / average"). The system divides freshness into four levels: excellent (100%), good (80%), average (60%), and poor (40%). The lower the freshness, the lower the initial shelf life is reduced proportionally. For example, regarding the processing status of ingredients, a voice interaction system can collect information on whether the ingredients are in a raw / processed / cooked state. The shelf life of processed / cooked ingredients can be reduced by 30%-50% according to the system default value, in line with the quality decay pattern of ingredients in different processing states.

[0046] After processing the data collected by each sensor unit, time-series synchronization and alignment, food association and binding, and multi-dimensional feature fusion and extraction can be performed to obtain multi-dimensional status information stored in the food database or directly used for subsequent food management. The refrigerator's internal temperature / humidity and door opening status from the multi-dimensional sensor data can be collected in real time, with a preset collection frequency of once every 10 seconds; door opening frequency / duration and temperature fluctuation coefficient can be statistically analyzed in near real-time, with a statistical frequency of once per hour. Food stacking density, regional temperature and humidity, and other multi-dimensional status information can be updated periodically, for example, once per day. For other data that requires manual or voice input by the user, such as the initial freshness of the food, processing status, and mixed storage status, the user can input this information all at once when the food is put into storage.

[0047] S103: Input the multidimensional state information into the pre-trained shelf-life determination model and output the shelf-life information of the target food ingredient.

[0048] The shelf-life determination model can be an artificial intelligence model pre-trained with a large number of food shelf-life samples, used to predict food shelf-life related data based on multi-dimensional state information. For example, the shelf-life determination model can adopt an edge-cloud two-tier architecture, where the edge corresponds to the local food storage device and the cloud corresponds to the cloud server, thus balancing offline usage needs with accurate cloud-based predictions. In other embodiments, the two-tier architecture of the shelf-life determination model can also be deployed on the same device, or the shelf-life determination module can be a single model; this specification does not impose any limitations on this.

[0049] Shelf life information can be data related to the food safety and freshness of the target ingredient in terms of time, reliability, and quality changes, including but not limited to target shelf life, shelf life confidence level, and quality degradation data.

[0050] In practical implementation, pre-collected sample data covering various common ingredients, different storage environments, and different usage habits can be input into the constructed artificial intelligence model. Sample data can include, for example, the actual shelf life of the ingredients, freshness degradation data, and spoilage time points. Based on the sample data, a lightweight local edge model and a high-precision cloud-based fusion model can be trained separately. After training, the lightweight model can be deployed in the refrigerator's intelligent management module (it can be independent of the refrigerator or as part of it), while the high-precision model can be deployed on a cloud server. During application, pre-processed multi-dimensional state information can be input into the pre-trained shelf-life determination model. The model outputs the shelf-life information of the target ingredient through multi-feature weighted fusion calculations. In offline mode, basic prediction can be completed through the local lightweight model; in network mode, high-precision optimized prediction can be completed through the cloud-based model.

[0051] S104: Based on the shelf life information, determine the target management operation corresponding to the target ingredient, and execute the target management operation to manage the target ingredient.

[0052] Target management operations can be a full-process management action based on shelf-life information to ensure food safety and reduce food waste. This may include, but is not limited to, visual display of food status, tiered early warning, consumption planning suggestions, data statistics updates, and dietary optimization recommendations.

[0053] In practice, the system can pre-store management operation mapping rules corresponding to different shelf-life information. Based on the remaining shelf life, it can classify products into three levels: fresh, near-expiry, and expired, corresponding to different display methods and warning strategies. Based on the shelf-life confidence level, it can adjust the frequency and method of warnings. Based on quality degradation data, it can generate suggestions for the order in which ingredients should be consumed. The system can match corresponding target management operations based on the output shelf-life information and execute them synchronously through the refrigerator and mobile app.

[0054] In the embodiments of this specification, by fusing multidimensional state information covering at least two categories from dimensions such as target ingredient attributes, storage environment, and operating status, the shelf-life information corresponding to the target ingredient is obtained. Based on this shelf-life information, corresponding target management operations are performed, which improves the accuracy of shelf-life prediction. This ensures that the prediction results closely match the actual storage scenario and quality change patterns of the ingredients, reducing the food safety risks caused by expired ingredients. Furthermore, based on a pre-trained shelf-life determination model, the dynamic determination of shelf-life can be completed automatically and adaptively without relying on manual operation, making it highly applicable. In addition, by combining shelf-life prediction, operational decision-making, and automatic execution, comprehensive, refined, and differentiated management of ingredients can be achieved without structural modifications to existing food storage devices. This allows for adaptation to various food storage devices and scenarios, reducing food management costs.

[0055] refer to Figure 2 As shown, in some embodiments of this specification, the shelf-life determination model may include a first sub-model and a second sub-model, and the multi-dimensional state information may include at least attribute information, first environmental data, ingredient state data, operating state data, and second environmental data. Further, inputting the multi-dimensional state information into a pre-trained shelf-life determination model and outputting the shelf-life information of the target ingredient may include: S201: Input the attribute information and the first environmental data from the multidimensional state information into the first sub-model, and output the initial shelf life of the target food ingredient.

[0056] S202: Input the initial shelf life, as well as the food status data, operation status data, and second environment data in the multi-dimensional status information, into the second sub-model, and output the shelf life information of the target food. The shelf life information includes at least one of the following: target shelf life, target shelf life confidence level, and quality decay data. The quality decay data is used to characterize the relationship between the quality of the target food and the change over time.

[0057] The first sub-model can be a lightweight edge prediction model deployed locally on the food storage device (such as an intelligent management module), for example, a lightweight XGBoost regression model that can achieve initial shelf life prediction without network connection; the second sub-model can be a high-precision fusion prediction model deployed on a cloud server, for example, an LSTM long short-term memory network + XGBoost fusion model that can achieve accurate prediction and optimization based on full-dimensional big data.

[0058] In practice, the first sub-model can be pre-trained using a large number of basic food samples and then lightweighted and stored in the refrigerator's intelligent management module. The input features of the model can be food attribute information (such as food category and basic shelf life) and first environmental data (such as real-time temperature, humidity and storage area of ​​the storage partition). After calculation, the model can output the initial shelf life of the target food in the current local micro-environment.

[0059] In practical implementation, with the refrigerator connected to the network, the intelligent management module can upload initial shelf life, food condition data, operational status data, and secondary environment data to the cloud server via Bluetooth or other communication modules, inputting this data into the second sub-model deployed in the cloud. The second sub-model can perform high-precision calculations based on full-dimensional input features (such as attribute information, primary environment data, initial shelf life, food condition data, operational status data, and secondary environment data), combined with anonymized food storage big data from users across the network. It outputs a corrected target shelf life, a precise confidence level for the target shelf life, and quality degradation data of the food over time (e.g., presented as a quality score-time curve). Simultaneously, the cloud-based second sub-model can continuously collect actual food consumption and expiration data from users across the network, dynamically optimizing model parameters through online learning, and pushing incremental update packages to the local primary sub-model monthly to continuously improve local prediction accuracy. Furthermore, the quality score-time curve can be displayed as a line graph in the mobile app, with the horizontal axis representing storage time and the vertical axis representing the food quality score (0-100 points), while also indicating the safe consumption threshold (60 points) and the optimal consumption range. Users can visually see the decline in food quality over time through curves, such as "Strawberries' quality score drops to 60 points (safe threshold) on the 3rd day of storage, and it is recommended to eat them on the 1st or 2nd day for best results," which helps users plan the order of food consumption more rationally and improve their dining experience.

[0060] In the embodiments described in this specification, the two-level sub-model architecture design balances offline stability with high online prediction accuracy, enabling continuous iterative optimization and improving prediction accuracy. Simultaneously, the multi-dimensional output of shelf-life information enriches the dimensions of food management, enhancing practicality and user experience.

[0061] In some embodiments of this specification, before inputting the initial shelf life, as well as the food status data, operating status data, and second environment data from the multidimensional status information into the second sub-model and outputting the shelf life information of the target food, the method may further include: determining the model parameters of the second sub-model based on the attribute information in the multidimensional status information, and adjusting the second sub-model based on the determined model parameters.

[0062] Furthermore, inputting the initial shelf life, as well as the ingredient status data, operational status data, and second environmental data from the multidimensional status information, into the second sub-model and outputting the shelf life information of the target ingredient may include: inputting the initial shelf life, as well as the ingredient status data, operational status data, and second environmental data from the multidimensional status information, into the adjusted second sub-model and outputting the shelf life information of the target ingredient.

[0063] The model parameters can be the core computational parameters such as the weight coefficients, quality decay coefficients, and correction rule thresholds of each input feature in the second sub-model. Different types of ingredients can correspond to different model parameters.

[0064] In practice, a model parameter mapping table corresponding to the food type can be pre-stored. Based on the food type (perishable / moderately perishable / storable) in the target food's attribute information, the corresponding model parameters are matched, and the second sub-model is adaptively adjusted. Specific parameter adjustment rules may include: for perishable foods (such as fresh fruits and vegetables, raw meat, seafood, fresh milk, etc.), the quality decay coefficient is high, and the weight of environmental features such as temperature and door opening frequency is increased to 50%, with model prediction accuracy calibrated at the daily / hourly level; for moderately perishable foods (such as cooked food, processed meat, yogurt, soy products, etc.), the quality decay coefficient is moderate, the environmental feature weight is set to 35%, and calibration is performed at the daily level; for storable foods (such as dried goods, seasonings, frozen meat, canned food, etc.), the quality decay coefficient is low, the environmental feature weight is set to 20%, calibration is performed at the weekly / monthly level, and the shelf life is only corrected when there are significant changes in the environment.

[0065] In the embodiments of this specification, by using a model parameter adaptive adjustment mechanism based on the attribute information of ingredients, it is possible to achieve stratified dynamic prediction for different types of ingredients, adapt to the quality decay characteristics of different ingredients, make the prediction results more consistent with the inherent characteristics of different ingredients, and improve the accuracy of shelf life prediction for different types of ingredients.

[0066] refer to Figure 3 As shown, in some embodiments of this specification, after inputting the initial shelf life, as well as the food ingredient status data, operating status data, and second environment data from the multi-dimensional status information, into the second sub-model and outputting the shelf life information of the target food ingredient, the method may further include: S301: Obtain the shelf life data corresponding to the target ingredient data that matches the target ingredient in the ingredient database.

[0067] The food ingredient database can be a complete database of food storage, consumption, and expiration information for multiple users stored in the cloud. The data can be categorized and stored according to region, refrigerator type, and user habits. Multidimensional status information obtained after processing multidimensional sensor data can also be stored in the food ingredient database.

[0068] In practice, the cloud server can filter all data from the food database that match the target food in terms of name and category, categorize them by region, refrigerator type, and user ID, and extract the corresponding actual shelf life data to provide a data foundation for subsequent corrections.

[0069] S302: Correct the target shelf life based on the shelf life data.

[0070] The embodiments in this specification allow for adjustments to the shelf life of food ingredients in at least one of the following three dimensions: (1) Filter the first shelf life data of the food data that matches the geographical location and food name of the target food in the shelf life data, and correct the target shelf life based on the first shelf life data.

[0071] The above corrections can be made for the same ingredients in the same region. In practice, the region can be divided according to the user's geographical location. The cloud will calculate the average actual shelf life of the same ingredients in the same region. If the local predicted shelf life of a single user deviates from the regional average by more than 20%, the cloud will make a small correction (correction range ≤15%) based on the regional average to adapt to the climate differences of different regions (such as the difference in shelf life of the same ingredients in humid southern regions and dry northern regions).

[0072] (2) Filter the second shelf life data in the shelf life data that matches the ingredient name of the target ingredient and the ingredient storage device, and correct the target shelf life based on the second shelf life data.

[0073] The above corrections can be made for refrigerators operating in the same environment. In practice, the equipment can be categorized by refrigerator brand / model / cooling type. The cloud platform will then calculate the actual shelf life of the same type of refrigerator and the same food. If the predicted shelf life of a user's refrigerator deviates from the average of refrigerators of the same type by more than 15%, the cloud platform will make corrections based on the characteristics of the equipment to adapt to the differences in cooling effects between different refrigerators.

[0074] (3) Filter the third shelf life data in the shelf life data that matches the ingredient name of the target ingredient and the target user corresponding to the target ingredient, and correct the target shelf life based on the third shelf life data.

[0075] The above modifications can be made based on user habits. In practice, a user's personal usage habit model can be established based on the user's historical food storage data. The average frequency of opening the door, food stacking habits, and actual consumption period of the food can be statistically analyzed to make personalized adjustments to the shelf life of the same type of food (for example, if a user frequently opens the door, the system will uniformly reduce the shelf life of all their food by 5%-10%).

[0076] Furthermore, for the above corrections, correction rules and priorities can be set, including: (1) Correction priority: can be set to personal usage habit correction > same refrigerator environment correction > same area and same food correction, to ensure that the prediction results are in line with the user's individual usage scenario; (2) Correction range limit: the correction range of a single dimension is ≤15%, and the total correction range is ≤30%, to avoid excessive correction that causes the prediction results to deviate from reality; (3) Correction result feedback: the cloud will synchronize the corrected shelf life and core correction basis to the user's mobile APP, such as "Your Apple recommended shelf life has been corrected from 7 days to 6 days, because: your refrigerator opens the door more frequently than the average of the same model of refrigerator by 30%".

[0077] In the embodiments described in this specification, through the collaborative correction of three dimensions, a full-dimensional shelf-life correction can be achieved, from the commonalities of the group to the individual characteristics. This can adapt to differentiated factors such as regional climate, refrigeration equipment, and user habits, so that the shelf-life prediction results not only conform to the characteristics of the food itself, but also adapt to the actual usage scenarios of users, improving the personalization and accuracy of shelf-life prediction. At the same time, by limiting the correction priority and magnitude, over-correction is avoided, ensuring the stability and security of the prediction results.

[0078] refer to Figure 4As shown, in some embodiments of this specification, the confidence level of the target shelf life can be determined by the following methods: S401: The second sub-model determines the data integrity score of the target ingredient based on the input data of the second sub-model.

[0079] S402: The second sub-model determines the matching degree between the input data and the training data of the second sub-model, and determines the model matching degree score of the target ingredient relative to the second sub-model based on the matching degree.

[0080] S403: The second sub-model obtains the shelf life data corresponding to the target ingredient data that matches the target ingredient in the ingredient database, and determines the synergy score between the target shelf life and the shelf life data.

[0081] S404: The second sub-model determines the confidence level of the target shelf life based on the data integrity score, the model matching score, the synergy score, and the preset weights of each score dimension.

[0082] The target shelf-life confidence score quantifies the reliability of shelf-life predictions; a higher score indicates stronger reliability. The data integrity score assesses the completeness of feature data collected upon food entry into the warehouse. The model matching score evaluates the match between the feature data of the current food and the training samples of the second sub-model. The synergy score assesses the synergy between the prediction results based on the current food and large datasets from the same scenario.

[0083] In practice, the confidence level of the target shelf life can be calculated using a weighted summation method. That is, based on three dimensions—data integrity, model matching degree, and big data collaboration degree—different weights are set for each dimension, and the total weight sum is 100%. The calculation formula can be: Shelf life confidence level (C) = Data integrity score (C1) × 40% + Model matching degree score (C2) × 35% + Big data collaboration degree score (C3, i.e., collaboration score) × 25%.

[0084] Among them, the data integrity score (C1) is 0-100 points. It can be scored based on the completeness of the feature data collected when the ingredients are put into storage. The more complete the data dimensions, the higher the score (for example, only the ingredient name and shelf life are entered, and 40 points are obtained; all dimensions of data are entered, and 100 points are obtained).

[0085] Among them, the model matching score (C2) is 0-100 points. It can be scored based on the matching degree between the feature data of the current food and the training samples in the cloud model. The higher the matching degree, the higher the score (for example, the matching degree of the common food apple is 95 points, and the matching degree of the niche food prickly pear is 60 points). Among them, the big data synergy score (C3) is 0-100 points. It can be scored based on the synergy between the current food prediction results and the big data of the same region / same refrigerator / same user. The higher the synergy, the higher the score (for example, if the prediction result deviates from the regional average by 5%, you get 90 points; if the deviation is 30%, you get 50 points).

[0086] In some embodiments of this specification, the first sub-model may output an initial shelf life confidence score along with the initial shelf life. This initial shelf life confidence score can be calculated in a similar manner to the target shelf life confidence score described above; that is, it can be a weighted sum of the data integrity score, model matching score, and synergy score. In some embodiments, if the first sub-model is set up locally and is offline, and the synergy score cannot be calculated, then a weighted sum of the data integrity score and the model matching score can be calculated as the initial shelf life.

[0087] Furthermore, when the output of the first sub-model includes the initial shelf-life confidence score, this initial shelf-life confidence score can also be input into the second sub-model. Then, when calculating the target shelf-life confidence score, the second sub-model can combine the initial shelf-life confidence score and its weights with the weighted sum of the above three dimensions, resulting in four dimensions, with a total weight sum of 100%.

[0088] In some embodiments of this specification, the shelf-life information may further include the shelf-life confidence level at which the target shelf-life confidence level is located. For example, it may include extremely high, high, medium, low, and extremely low, with each level corresponding to a different management operation execution method.

[0089] Furthermore, performing the target management operation to manage the target ingredient may include: determining an execution method for the target management operation that matches the shelf-life confidence level; and performing the target management operation based on the execution method to manage the target ingredient. The target management operation may include at least one of the following: displaying the freshness status of the target ingredient on the door display of the ingredient storage device; pushing early warning information to the user terminal corresponding to the ingredient storage device; generating and displaying the quality change trend information of the target ingredient; and updating the ingredient data associated with the target ingredient in the ingredient database.

[0090] The system can pre-store mapping rules between confidence levels and management operation execution methods. Different confidence levels correspond to different execution methods, which may include, for example, the way prediction results are displayed, early warning trigger nodes, early warning frequency, and reminder content.

[0091] In practice, five levels can be defined based on the confidence level values. Different levels correspond to different prediction result display methods and early warning strategies. These levels are visually indicated (e.g., by color or star rating) on ​​the refrigerator door touchscreen display and mobile app, allowing users to easily identify them. An example mapping rule between confidence levels, display methods, and early warning strategies is shown in Table 1 below: Table 1

[0092] Furthermore, when the confidence level of the shelf life is B or below, the system pushes a confidence level improvement guide to the user in the mobile APP, prompting the user to collect relevant data, such as "Your banana shelf life confidence level is 65 points (B level). After supplementing the freshness and stacking status, the confidence level can be improved to above 85 points", guiding the user to improve the data and further improve the accuracy of subsequent predictions.

[0093] refer to Figure 5 As shown, in some embodiments of this specification, the method may further include: S501: Obtain abnormal environmental data collected by the sensing module of the food storage device.

[0094] Abnormal environmental data can be the concentration data of food spoilage characteristic gases (ammonia, hydrogen sulfide, etc.) exceeding the preset safety threshold, collected by the odor sensing unit. The odor sensing unit can capture objective signals of the presence of spoilage gases inside the refrigerator, rather than directly identifying a single food item as the source of the gas. Furthermore, through the linkage analysis of multi-dimensional state data and shelf-life information, the most likely spoiled food items can be determined.

[0095] In practice, odor sensors deployed in each storage compartment of the refrigerator collect data on the concentration of characteristic gases of spoilage in the corresponding compartment at a preset frequency. The data, carrying the compartment ID, is uploaded to the intelligent management module. The system can pre-store safe concentration thresholds for each characteristic gas. When the gas concentration in a compartment exceeds the threshold, it is determined to be abnormal environmental data, triggering subsequent investigation procedures. For incomplete food items, the sensor collection frequency is increased to once every 30 minutes, and the gas concentration exceeding the standard threshold is lowered by 30%, ensuring early detection and early warning.

[0096] S502: Based on the location information of the sensing unit corresponding to the abnormal environment data, the abnormal environment data, and the attribute information of each ingredient, candidate spoiled ingredients are determined from the ingredients stored in the ingredient storage device.

[0097] The location information of the sensing unit can be the partition ID corresponding to the odor sensor that collects abnormal data, that is, the uniquely bound physical monitoring partition. Based on this location information, ingredients with a high probability of gas release (i.e., candidate spoiled ingredients) can be identified.

[0098] In practical implementation, a zoned gas diffusion time model can be established. This model determines the time window from gas release to detection based on the gas release rate of different food spoilage foods and the airflow circulation pattern inside the refrigerator. Only food stored in that zone within that time window is listed as a "suspected spoiled food list." Furthermore, by combining the gas sensitivity ranking of food categories (perishable > moderately perishable > storable), food with a high probability of gas release can be prioritized.

[0099] S503: Based on the target shelf life and food status data of the candidate spoiled food ingredients, determine the target spoiled food ingredient from the candidate spoiled food ingredients, generate spoilage warning information of the target spoiled food ingredient, and push it to the user terminal that matches the food storage device.

[0100] In practice, the system can be based on a list of suspected spoiled ingredients, combined with multi-dimensional data linkage for screening. This can include: linking with shelf-life information to filter ingredients with a remaining shelf life of ≤3 days, expired, or marked as incomplete; linking with data collected by temperature sensors to double the probability of spoilage if the local temperature at the storage location exceeds the suitable range; and linking with user operation records to prioritize ingredients marked as "remaining" before abnormal data is detected. Furthermore, the probability of spoilage can be calculated using multi-dimensional data weighting to identify the ingredients with the highest probability of spoilage as target spoiled ingredients. A spoilage warning message containing the ingredient name, storage location, abnormal gas concentration, and spoilage risk level can be generated and pushed to the user via a red flashing indicator on the refrigerator touchscreen, real-time push notifications on a mobile app, and local voice alerts. Simultaneously, high-frequency secondary monitoring is initiated if the spoilage is not confirmed, with corresponding zone sensors collecting data every 10 minutes; the warning is escalated if the concentration continues to rise.

[0101] Furthermore, to manage food spoilage, cross-interference filtering can be implemented. That is, if multiple zone sensor units simultaneously show slight exceedances, it's judged as an environmental gas fluctuation, and no warning is triggered. It can also correct for food sealing status; when the user enters information, a voice prompt asks whether the food is sealed, and sealed food is downgraded by 50% in the suspected spoilage list. A sensor unit self-calibration mechanism is also included, automatically performing monthly self-calibration through detection in blank zones without food, correcting detection errors.

[0102] In some embodiments of this specification, considering that incomplete ingredients (such as cut fruits / meats, opened milk / beverages, leftover cooked food, etc.) experience a much faster decline in freshness than complete ingredients due to exposure to air, accelerated microbial growth, and moisture loss, differentiated management can be implemented for these types of ingredients to achieve precise freshness monitoring and early warning. Management rules for incomplete ingredients may include: (1) Half shelf life rule: The remaining shelf life of incomplete ingredients will be directly halved based on subsequent cloud dynamic prediction. For example, if the remaining shelf life of a whole cucumber is 3 days, the remaining shelf life after cutting it will be adjusted to 1.5 days. (2) Environmental sensitivity enhancement rules: Incomplete food is more sensitive to the frequency of refrigerator door opening and local temperature fluctuations. For every additional time the door is opened per day, the freshness score is reduced by 2 points. If the local temperature exceeds the suitable range by 1℃, the score is reduced by 5 points. (3) High-frequency sensing and monitoring rules: Odor sensors and micro-temperature sensors use high-frequency real-time monitoring for incomplete ingredients, with a monitoring frequency 4 times that of complete ingredients, to ensure timely capture of spoilage signals; (4) Freshness base deduction rules: When an ingredient is marked as "leftover", its freshness base score will be reduced by 10-20 points (20 points for perishable items and 10 points for moderately perishable / storable items). (5) Warning level upgrade rules: The warning point for incomplete ingredients is advanced. For complete ingredients, the warning is "2 days before the expiration date". For incomplete ingredients, the warning is "1 day before the expiration date / when the freshness score drops to 70 points". The warning method is "red flashing on the refrigerator touch screen + real-time push on the mobile APP + local voice reminder". Multiple warnings ensure that users pay attention in time. (6) Manual marking mandatory rule: When a user takes a whole ingredient but does not replenish the remaining amount information, the system will continuously pop up voice / APP reminders until the user marks the ingredient as "taken all" or "remaining XX", to avoid incomplete ingredients being managed as complete ingredients due to not being marked, which would lead to errors in freshness judgment.

[0103] In some embodiments of this specification, when users enter food data, they can also simultaneously enter the estimated consumption time. After predicting the target shelf life, the estimated consumption time can be compared with the target shelf life. If the estimated consumption time is less than or equal to the target shelf life, the estimated consumption time can be used as the countdown benchmark for subsequent food management operations, aligning with users' lifestyles. Simultaneously, the system will display a message in the app stating, "The current estimated time is within the safe consumption range for the food." If the estimated consumption time is greater than the target shelf life, the target shelf life can be used as the final countdown benchmark. The system will then display a pop-up message in the app showing the calculation basis for the recommended shelf life (e.g., "Current refrigerator temperature is 8°C; the default shelf life of apples is 7 days, but it has been corrected to 5 days due to the higher temperature"), and remind users that "exceeding the recommended time may pose a risk of food spoilage." Furthermore, users can manually confirm whether to adhere to the original estimated time. If the user confirms, the system will use the user's time as the benchmark but will increase the warning level (e.g., starting the warning 3 days before the expiration date, instead of the default 2 days).

[0104] Based on the foregoing food management method, one or more embodiments of this specification also provide a food management device. The device may include an apparatus (including a distributed system), software (application), module, plug-in, server, client, etc., using the method described in the embodiments of this specification, combined with necessary hardware implementation. Based on the same innovative concept, the devices in one or more embodiments provided in this specification are as described in the following embodiments. Since the implementation schemes and methods for solving the problem are similar, the implementation of specific devices in the embodiments of this specification can refer to the implementation of the foregoing method, and repeated details will not be elaborated. As used below, the terms "unit" or "module" can refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated. Figure 6 The diagram shown is a schematic representation of the food ingredient management device provided in an embodiment of this specification. Figure 6 As shown, the food ingredient management device 600 may include: The first acquisition module 601 is used to acquire the identification information of the target food stored in the food storage device.

[0105] The second acquisition module 602 is used to acquire multi-dimensional status information related to the target ingredient based on the identification information. The multi-dimensional status information includes at least two of the following: attribute information of the target ingredient, first environmental data of the environment in which the target ingredient is located, ingredient status data of the target ingredient, operating status data of the ingredient storage device, and second environmental data of the environment in which the ingredient storage device is located.

[0106] The determination module 603 is used to input the multidimensional state information into a pre-trained shelf-life determination model and output the shelf-life information of the target food ingredient.

[0107] The operation module 604 is used to determine the target management operation corresponding to the target ingredient based on the shelf life information, and execute the target management operation to manage the target ingredient.

[0108] The descriptions and functions of the above modules can be found in the section on food ingredient management methods, and will not be repeated here.

[0109] This application also provides an electronic device, which can be a module / device for performing the aforementioned food ingredient management method, such as... Figure 7 As shown, the electronic device may include a processor 701 and a memory 702, wherein the processor 701 and the memory 702 may be connected via a bus or other means. Figure 7 Taking the bus connection as an example, this electronic device can be applied to each node of each geographic subsystem.

[0110] Processor 701 can be a central processing unit (CPU). Processor 701 can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0111] The memory 702, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the food ingredient management method in the embodiments of the present invention. The processor 701 executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions, and modules stored in the memory 702, thereby implementing the food ingredient management method in the above method embodiments.

[0112] The memory 702 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor 701, etc. Furthermore, the memory 702 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 702 may optionally include memory remotely located relative to the processor 701, and these remote memories may be connected to the processor 701 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0113] The one or more modules are stored in the memory 702, and when executed by the processor 701, the following food ingredient management method is performed: The system acquires the identification information of the target ingredient stored in the ingredient storage device; based on the identification information, it acquires multi-dimensional state information related to the target ingredient, the multi-dimensional state information including at least two of the following: attribute information of the target ingredient, first environmental data of the environment in which the target ingredient is located, ingredient state data of the target ingredient, operating state data of the ingredient storage device, and second environmental data of the environment in which the ingredient storage device is located; it inputs the multi-dimensional state information into a pre-trained shelf-life determination model and outputs the shelf-life information of the target ingredient; based on the shelf-life information, it determines the target management operation corresponding to the target ingredient and executes the target management operation to manage the target ingredient.

[0114] The specific details of the aforementioned electronic device can be understood by referring to the relevant descriptions and effects in the above method embodiments, and will not be repeated here.

[0115] This specification also provides a computer storage medium storing computer program instructions, which, when executed, implement the steps of the above-described food ingredient management method.

[0116] This specification also provides a computer program product, which includes a computer program that, when executed, implements the steps of the above-described food ingredient management method.

[0117] Based on the aforementioned food management method, this specification also provides a food management system, which may include: a food storage device, a sensing module disposed on the food storage device, and a management module.

[0118] The sensing module is used to collect multidimensional sensing data of the target ingredients stored in the food storage device and send the multidimensional sensing data to the management module.

[0119] The management module is used to: process the multidimensional sensing data to obtain multidimensional state information related to the target ingredient, wherein the multidimensional state information includes at least two of the following: attribute information of the target ingredient, first environmental data of the environment in which the target ingredient is located, ingredient state data of the target ingredient, operating state data of the ingredient storage device, and second environmental data of the environment in which the ingredient storage device is located; input the multidimensional state information into a pre-trained shelf-life determination model and output shelf-life information of the target ingredient; determine the target management operation corresponding to the target ingredient based on the shelf-life information, and execute the target management operation to manage the target ingredient.

[0120] Based on the aforementioned food management method, this specification also provides a voice-interactive-based intelligent food management and early warning method for refrigerators. This method may include a hardware module and a software system, with the hardware module and software system communicatively connected.

[0121] (1) The hardware module may include: Intelligent Management Module: This module processes data, manages food information, and handles alerts and notifications. Specifically, it integrates data processing, computational analysis, food information management, alert triggering, and notification push functions. It also possesses command parsing, algorithm execution, and data storage capabilities, enabling data interaction and command execution with various associated hardware modules and software systems. This module is installed on the outer back of the refrigerator and connects to other hardware modules through corresponding interfaces, providing core control support for the entire system.

[0122] Refrigerator door touchscreen display: Displays and manages food information in separate zones, such as the refrigerator compartment, variable temperature compartment, and freezer compartment. It allows users to manually input, modify, and delete key data such as food name, quantity, and expiration date. The device connects to the intelligent management module via an adapter interface. Installation is designed to avoid the refrigerator door's sealing structure and storage space, ensuring it does not affect the refrigerator's sealing performance or storage capacity.

[0123] Refrigerator door open / close sensor: Used to detect the open / closed status of refrigerator doors in different areas. The voice recognition module is switched to a high-sensitivity pickup mode only when the refrigerator door is open (when the user needs to access food or interact via voice). When the refrigerator door is closed, the module enters a low-power sleep / low-sensitivity mode to reduce noise collection during periods of no interaction and prevent invalid noise from being misidentified by the algorithm. It provides data support for judging food access behavior and controlling system wake-up timing. It uses an adaptive voltage power supply and connects to the intelligent management module through a dedicated interface.

[0124] Low-power Bluetooth communication module: Used to enable data transmission between hardware modules and between hardware modules and software systems. Specifically, a low-power Bluetooth communication component can be used, which has stable communication capabilities beyond a preset distance and low-power operation characteristics. Its core function is to enable data transmission between various hardware modules and between hardware modules and software systems (cloud, mobile APP), ensuring real-time synchronization of data such as food information, warning signals, and control commands. It connects to the intelligent management module through corresponding interfaces.

[0125] The voice recognition anti-interference module consists of two parts: a directional microphone with a near-field directional pickup design, which highly sensitively collects voice signals only from the upper inner side of the refrigerator door (where the user speaks, within a preset pickup distance), naturally attenuating ambient noise from the compressor at the back of the refrigerator and distant kitchen areas, reducing noise intake at the source; and a hardware anti-interference processing unit, which integrates a low-frequency noise filtering circuit. The refrigerator compressor's operating noise is a low-frequency continuous noise (fixed frequency band), and the hardware circuit directly filters this frequency band, retaining only the effective mid-to-high frequency signals of human speech, reducing the noise reduction burden on the software algorithm. It can be powered by an adaptive voltage and connected to the intelligent management module through a corresponding interface, installed on the upper inner side of the refrigerator door (close to the user's opening position, without affecting the refrigerator door seal or storage space).

[0126] Infrared ranging sensor: A low-cost, lightweight infrared ranging sensor can be selected and installed on the inner wall of each area of ​​the refrigerator (such as the upper and lower shelves of the refrigerator compartment and the drawer area). One sensor is deployed in each area. Data transmission is achieved through the expansion interface of the intelligent management module. It is used to collect the volume ratio of food stacking and to provide a basis for determining the density level of food stacking. The installation method is adhesive and no drilling or modification is required.

[0127] Odor Sensor: Employs semiconductor odor sensors (low cost, low power consumption), precisely deployed according to the refrigerator's storage areas (1-2 sensors are installed on the inner walls of different areas in the refrigerator compartment, variable temperature compartment, and freezer compartment, with a total of ≤5 sensors). Each sensor is assigned a unique "zone ID" and installed at the corresponding zone's airflow points (such as under shelves or inside drawers). It can identify characteristic gases such as ammonia and hydrogen sulfide released during food spoilage, and the collected gas concentration data, along with the zone ID, is uploaded to the intelligent management module.

[0128] Micro-temperature sensor: A micro-temperature sensor with a resolution of ±0.1℃ can be selected and deployed on the inner wall of each storage compartment, corresponding to the odor sensor. It accurately collects the micro-environmental temperature of the food storage location and transmits data through the intelligent management module expansion interface. This provides a basis for monitoring local temperature fluctuations and determining the suitable storage temperature for food. The installation method is adhesive, requiring no drilling or modification.

[0129] (2) The software system may include: Voice interaction system: Users can interact with the refrigerator by using a wake-up word. When placing food items in the refrigerator, users can verbally tell the refrigerator the name, quantity, and expiration date of the food. It can be used in conjunction with a voice recognition module, pre-stores a database of common food names (covering over a thousand common foods including vegetables, fruits, meats, snacks, and beverages), and stores their corresponding default expiration dates. Different names for the same food item in different regions are also included in the food name database, supporting accurate recognition of fixed-format commands such as "put in xxx" and "take out xxx". To address low-frequency compressor noise from refrigerators and random background noise such as water flow or clattering kitchen utensils, feature sample databases of two types of noise are pre-collected. The mixed speech signal is then processed using spectral subtraction: separating the spectral features of the speech and noise signals, removing the noise spectrum to restore the pure speech signal. For random and irregular environmental noise, a wavelet denoising algorithm is used to accurately filter non-stationary noise. The speech signal after hardware filtering and spectral subtraction is then enhanced in amplitude and features to highlight the core features of human speech (such as tone and syllables), weakening residual noise and ensuring accurate capture by the subsequent recognition model. The system determines in real-time whether the collected signal is valid human speech or noise, only transmitting it to the subsequent command recognition algorithm when valid speech is detected. If it is pure noise, it is directly shielded and not transmitted to avoid misidentification as a command, further improving anti-interference capabilities and achieving a recognition accuracy of ≥92% and a response time of ≤1 second. If the food name in the command does not match the database, a confirmation prompt can be pushed to the user via a mobile app, allowing the user to manually add the food information. Food ingredient management and early warning system: Based on food ingredient data transmitted through the voice interaction system, it automatically generates a food ingredient information table containing ingredient name, quantity, storage time, remaining shelf life, and freshness status, and activates a shelf-life countdown function. The system uses a three-color labeling system to classify and manage food ingredients: fresh ingredients are automatically classified and displayed in green; ingredients nearing their expiration date (2 days before expiration) are automatically classified and displayed in orange; and ingredients past their expiration date are automatically classified and displayed in red, triggering an early warning mechanism that sends warning reminders to users through various means, including a flashing red indicator on the refrigerator door touchscreen and notifications via a mobile app. The cloud-based dynamic shelf-life prediction system dynamically adjusts shelf-life predictions by combining data on the user's local temperature and humidity, the properties of the food itself, and the environment surrounding the refrigerator. For example, it can be deployed on a cloud server, integrating an LSTM (Long Short-Term Memory) network and an XGBoost fusion model, communicating with the intelligent management module and a mobile app via an interface. The system acquires real-time temperature and humidity data from the user's location, temperature / humidity data from sensors inside the refrigerator, dynamic operating data from the refrigerator door open / close sensor, and food density data from an infrared ranging sensor. It combines this data with various food standard shelf-lifespans, properties, and quality degradation coefficients stored in a food database, dynamically adjusting the shelf-life prediction results through a multi-feature weighted fusion algorithm. It also establishes a regional gas diffusion time model, supporting the generation and sorting of lists of suspected spoiled food. Simultaneously, it collects anonymized data on food storage, consumption, and expiration from users across the network, enabling online model learning and iteration. Monthly incremental model updates are pushed to the local intelligent management module to continuously improve local prediction accuracy. Layered prediction rules are applied to different categories of food, outputting accurate personalized shelf-life predictions, confidence values, and quality degradation trend curves, which are then displayed on the user's mobile app. For example, milk with a shelf life of 7 days at room temperature will have a shelf life of 14 days when refrigerated (4°C). If the relative humidity in the region is >80%, the shelf life will be further adjusted to 12 days. If the user opens the refrigerator door too frequently, the shelf life will be slightly reduced again.

[0130] The user's mobile app (supports iOS / Android / HarmonyOS, etc.) receives and displays food information, freshness status, and food management alerts. It shows the types, quantities, and expiration dates of food in the refrigerator, facilitating user planning of meals and purchases. Users can also manually modify food data, view historical records, waste logs, carbon reduction data, and dietary optimization suggestions. For example, it can display a real-time list of all food items in the refrigerator, variable temperature compartment, and freezer, categorizing and managing them dynamically. The cloud-based dynamic shelf-life prediction system adjusts the shelf-life model based on real-time regional temperature and humidity, refrigerator environmental parameters, and food characteristics, outputting accurate shelf-life predictions. It pushes alerts to users when food is nearing its expiration date or has spoiled. Users can also view all food information on their phones and make timely purchase plans.

[0131] Dynamic shelf-life prediction allows users to estimate the time they will consume food within a certain number of days based on their lifestyle habits, rather than the actual shelf-life of the food under specific temperature and humidity conditions. Simultaneously, the system will also provide recommendations based on the actual shelf-life of relevant foods under specific temperature and humidity conditions. When the user's estimated shelf-life exceeds the system's recommended shelf-life, the system's default shelf-life will be used, and a countdown will begin. The system can employ a multi-feature weighted fusion prediction model integrating AI to achieve intelligent dynamic prediction of recommended shelf-life, specifically including: 1) Local edge prediction layer: A lightweight XGBoost regression model is employed. After training on a massive dataset of food shelf-life samples, the model undergoes lightweight trimming and is pre-recorded in the intelligent management module. It can achieve rapid predictions based on basic environmental data collected locally within the refrigerator and basic food characteristic data, with a response time of ≤0.5s. Model input features include: food category / subcategory, initial shelf-life (system default), real-time temperature / humidity inside the refrigerator, and food storage area (refrigerated / frozen / variable temperature). Model output results include: a basic recommended shelf-life for the food under the current local environment + an initial confidence value for the shelf-life. Basic predictions can be performed without an internet connection, meeting offline usage requirements, with low computational load and fast response speed.

[0132] 2) Cloud-based Big Data Optimization Layer: A combined LSTM (Long Short-Term Memory) network and XGBoost model is deployed on a cloud server. Based on locally uploaded, multi-dimensional data, anonymized user data from across the internet, and a large database of food characteristics, accurate and optimized shelf-life predictions are achieved, with incremental updates to the local model. Model input features include: all features from the local prediction layer, refrigerator door opening frequency / duration, food stacking density, initial food freshness, temperature and humidity in the user's region, and large-scale data on the shelf-life of the same food in the same region. Model outputs include: accurate recommended shelf-life (correcting local prediction results), shelf-life confidence score, and food quality degradation trend curves. Model iteration mechanism: The cloud model continuously collects actual food consumption / expiration data from users across the internet, dynamically optimizing model parameters through online learning (OL), and pushing incremental update packages to the local edge model monthly to continuously improve local prediction accuracy.

[0133] Furthermore, the system can integrate the user's estimated consumption time with the system's recommended time. If the user's estimated consumption time is less than or equal to the system's accurately recommended shelf life, the user's estimated consumption time will be used as the final countdown benchmark, aligning with the user's lifestyle. Simultaneously, the system will display a message in the app stating, "The current estimated time is within the safe consumption range for the food." If the user's estimated consumption time exceeds the system's accurately recommended shelf life, the system's accurately recommended shelf life will be used as the final countdown benchmark. The system will also display a pop-up message in the app showing the calculation basis for the recommended shelf life (e.g., "Current refrigerator temperature is 8℃, apples have a default shelf life of 7 days, corrected to 5 days due to the higher temperature"), and remind the user that "Exceeding the recommended time may pose a risk of food spoilage." Users can manually confirm whether to stick to the original estimated time. If the user confirms, the system will use the user's time as the benchmark, but will increase the warning level (e.g., starting the warning 3 days before the expiration date, instead of the default 2 days).

[0134] In practice, before putting food into the refrigerator, you can select the storage area: refrigerator compartment, variable temperature compartment, freezer compartment, outside the refrigerator, etc. You can also intelligently select the storage area by opening and closing the refrigerator door. When no door is open, the default is to store the food outside the refrigerator.

[0135] Users can quickly update food data inside the refrigerator using fixed commands such as "wake word, put in / purchase xxx, shelf life XX days" or "wake word, take out / use / delete xxx", without manual operation or reliance on image recognition. The specific workflow is as follows: (1) Food Placement Stage. When the user prepares to place food, the refrigerator door switch sensor detects that the corresponding door is open and immediately transmits the signal to the intelligent management module. The system automatically enters the interactive state. The user starts the voice interaction system by using a "wake word". The voice recognition module collects and recognizes the food name, quantity and shelf life information provided by the user. After receiving the recognition results, the intelligent management module synchronizes the data to the cloud server through the low-power Bluetooth communication module. If the food has outer packaging, the user can scan the packaging barcode through the mobile APP to enter the production date and shelf life, or manually enter the relevant information through the refrigerator door touch screen display or the mobile APP. All data is synchronized to the cloud and the refrigerator door touch screen display through the intelligent management module and the low-power Bluetooth communication module to complete the food storage and filing.

[0136] (2) Freshness monitoring and shelf-life prediction stage. The cloud-based dynamic shelf-life prediction system acquires real-time temperature and humidity data of the user's location and the ambient temperature data inside the refrigerator (if the refrigerator has its own sensor, the regional ambient temperature data is used by default). Combined with the standard shelf-life and properties of the food in the food database, the system dynamically adjusts the shelf-life prediction results through a weighted algorithm. After receiving the prediction results from the cloud, the intelligent management module synchronizes them to the food ingredient management and early warning system. The system automatically generates a food ingredient information table, starts the shelf-life countdown function, and completes the classification and labeling according to the freshness status.

[0137] (3) Early Warning and Reminder Stage. The food management early warning system updates the freshness status and remaining shelf life data of food in real time, and synchronizes them to the mobile APP and refrigerator door touch screen display through the intelligent management module and low-power Bluetooth communication module: fresh food is displayed in green, food nearing its expiration date (2 days before expiration) is displayed in orange, and food that has exceeded its expiration date is displayed in red and triggers an early warning. When the remaining shelf life of food is ≤3 days, an expiration reminder is triggered (mobile APP push notification, enhanced orange indicator on refrigerator door touch screen display); when the food management early warning system determines that food has expired, a spoilage reminder is triggered (mobile APP push warning information, red flashing on refrigerator door touch screen display).

[0138] (4) Food Removal Stage

[0139] When a user removes food, the refrigerator door open / close sensor detects the door opening and sends a signal to the smart management module. The user then issues the command "Take out xxx" through the voice interaction system. After accurate recognition by the voice recognition module, the smart management module retrieves the data of the stored food items, matches the name, and executes the out-of-stock operation, moving the food item from "Stored Foods" to "History". The data is synchronized to the cloud server via the low-power Bluetooth communication module, and the mobile app and the refrigerator door touchscreen display refresh the list of stored food items in real time, completing the food out-of-stock record.

[0140] (5) Speech recognition anomaly handling

[0141] If the voice recognition module fails to recognize the command due to excessive ambient noise, the intelligent management module pushes a prompt to the mobile app via the low-power Bluetooth communication module, reminding the user to retry the voice command or manually operate via the refrigerator door touchscreen display or the mobile app. If the recognized food name does not match the food database of the voice interaction system or is ambiguous, the mobile app immediately pushes a confirmation prompt with similar food options. The user can confirm with one click or manually enter the food name, expiration date, and other data to ensure that the food management process is not interrupted.

[0142] If the speech recognition is ambiguous (such as "banana" and "cantaloupe" sounding similar) or fails to recognize the food name due to different regional names and dialects, the system will push a prompt through the mobile app, allowing users to quickly confirm or correct the information and ensure data accuracy.

[0143] The system also supports the management of food stored outside the refrigerator. The system can select the storage area as outside the refrigerator. Food condiments (salt, soy sauce, vinegar, Sichuan peppercorns, chili peppers, star anise, broad bean paste, etc.) and dried goods (dried bean curd sticks, vermicelli, wood ear mushrooms, mushrooms, noodles, etc.) are often stored outside the refrigerator, but are frequently forgotten. With the intelligent food management system of this invention, the management of such foods can be achieved. These foods generally have a long shelf life, and people are more concerned about the remaining quantity. Therefore, when storing outside the refrigerator, the voice command is "wake-up word, food name, estimated shelf life, quantity available." The quantity available is not a specific amount, but a general range: sufficient, small amount, purchase. When the quantity is "purchase," the user will be notified via the app, providing guidance for purchasing.

[0144] In the embodiments described in this specification, the lightweight upgrade of the local edge prediction model and the expansion of the data acquisition function are achieved through the local firmware upgrade of the intelligent management module (wireless OTA upgrade, no manual operation required); the deployment of the cloud model, big data collaborative correction, and confidence calculation are achieved through system iteration of the cloud server and continuous optimization; all local calculations adopt the lightweight model, the amount of calculation for a single shelf life prediction is ≤100KB, which does not occupy the core computing resources of the intelligent management module, ensuring that the overall system response speed remains unchanged (identification response time ≤1s, prediction response time ≤0.5s).

[0145] In this embodiment, dietary optimization suggestions can also be generated through AI big data analysis based on users' daily food types and user types (such as fitness and weight loss people, office workers, middle-aged and elderly people, etc.).

[0146] The voice-interactive-based intelligent management and early warning method for refrigerator food in the embodiments of this specification achieves accurate correlation between spoiled gas and designated food by adding low-cost odor sensors and micro-temperature sensors, combined with zoned deployment and multi-dimensional data linkage. This solves the problem of unclear gas source location when food is densely stored in the refrigerator. The method has low modification cost, high system fault tolerance, and timely information, thus realizing intelligent management of food throughout the kitchen.

[0147] By using voice recognition and fixed command interaction, it can solve the pain points of users holding food with both hands, making it inconvenient to operate a mobile phone or cooperate with image recognition. It further simplifies the process of updating food data and confirming spoilage. It is especially suitable for users who are not familiar with operating smart devices, such as the elderly and children, thus broadening the applicable scenarios of the product and improving the ease of operation and user experience.

[0148] By using a cloud-based dynamic shelf-life prediction system, the system dynamically adjusts the shelf-life model based on regional temperature and humidity, food characteristics, refrigerator environmental parameters, and sensor data. It innovatively designs six differentiated management rules for incomplete ingredients, making the prediction results more consistent with actual usage scenarios. This effectively reduces food waste caused by inaccurate shelf-life judgments and the rapid decline in freshness of incomplete ingredients. Through high-frequency monitoring by odor and micro-temperature sensors, it enables early detection of spoilage signals, ensuring food safety.

[0149] Adopting a modular hardware design, all new sensors are low-cost and universal, and the adhesive installation requires no drilling or modification. It can be installed on most existing refrigerators on the market, so users do not need to buy a new refrigerator, reducing the cost of use and improving product compatibility and consumer acceptance.

[0150] Furthermore, it enables intelligent management of the entire process, from food identification, freshness monitoring, shelf-life prediction, spoilage warning to data statistics and personalized suggestions, forming a complete behavioral closed loop. This not only ensures food safety but also promotes a green and low-carbon lifestyle. Through mechanisms such as cross-interference filtering and sensor self-calibration, it improves the accuracy of monitoring data and the reliability of the system.

[0151] In addition, it requires no camera and no manual input of information by users, making it easy to operate and convenient for people of different ages, such as the elderly and children. It also supports manual correction function, which balances intelligence and flexibility and improves the user experience. The special optimization for incomplete ingredients further covers the core pain points of family food use.

[0152] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium can also include combinations of the above types of memory.

[0153] The various embodiments in this specification are described in a progressive manner. For the same or similar parts between the various embodiments, please refer to each other. The focus of each embodiment is to describe the differences from other embodiments.

[0154] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions.

[0155] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0156] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute certain parts of the methods of various embodiments of this application.

[0157] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0158] This application can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0159] Although this application has been described through embodiments, those skilled in the art will know that this application has many modifications and variations without departing from the spirit of this application, and it is intended that the appended claims cover such modifications and variations without departing from the spirit of this application.

Claims

1. A method for managing food ingredients, characterized in that, The method, applied to a food storage device, includes: Obtain the identification information of the target food stored in the food storage device; Based on the identification information, multidimensional status information related to the target ingredient is obtained. The multidimensional status information includes at least two of the following: attribute information of the target ingredient, first environmental data of the environment in which the target ingredient is located, ingredient status data of the target ingredient, operating status data of the ingredient storage device, and second environmental data of the environment in which the ingredient storage device is located. The multidimensional state information is input into a pre-trained shelf-life determination model, which outputs the shelf-life information of the target food ingredient. Based on the shelf life information, determine the target management operation corresponding to the target ingredient, and execute the target management operation to manage the target ingredient.

2. The food ingredient management method according to claim 1, characterized in that, The attribute information includes at least one of the following: ingredient type, ingredient name, ingredient quantity, and basic shelf life of the ingredient; The food ingredient status data includes at least one of the following: food ingredient contact status, food ingredient freshness, food ingredient processing status, and food ingredient integrity. The first environmental data includes at least one of the following: temperature, humidity, odor, and storage location; The operating status data includes at least one of the following: the frequency of door opening and closing of the food storage device, the duration of door opening, the refrigeration efficiency or operating power, and the food stacking status. The second environmental data includes at least one of the following: temperature, humidity, geographic location, seasonal information, and climate type.

3. The food ingredient management method according to claim 1, characterized in that, The shelf-life determination model includes a first sub-model and a second sub-model; The multidimensional state information is input into a pre-trained shelf-life determination model, which outputs the shelf-life information of the target ingredient, including: The attribute information and the first environmental data from the multidimensional state information are input into the first sub-model, and the initial shelf life of the target food ingredient is output. The initial shelf life, as well as the ingredient status data, operating status data, and second environment data from the multidimensional status information, are input into the second sub-model to output the shelf life information of the target ingredient. The shelf life information includes at least one of the following: target shelf life, target shelf life confidence level, and quality decay data. The quality decay data is used to characterize the relationship between the quality of the target ingredient and time.

4. The food ingredient management method according to claim 3, characterized in that, The confidence level for the target shelf life was determined using the following method: The second sub-model determines the data integrity score of the target ingredient based on the input data of the second sub-model; The second sub-model determines the matching degree between the input data and the training data of the second sub-model, and determines the model matching degree score of the target ingredient relative to the second sub-model based on the matching degree; The second sub-model obtains the shelf life data corresponding to the target ingredient data that matches the target ingredient from the ingredient database, and determines the synergy score between the target shelf life and the shelf life data; The second sub-model determines the confidence level of the target shelf life based on the data integrity score, the model matching score, the synergy score, and the preset weights of each score dimension.

5. The food ingredient management method according to claim 3, characterized in that, Before inputting the initial shelf life, along with the food status data, operational status data, and second environmental data from the multidimensional status information, into the second sub-model and outputting the shelf life information of the target food ingredient, the process further includes: Based on the attribute information in the multidimensional state information, the model parameters of the second sub-model are determined, and the second sub-model is adjusted based on the determined model parameters; The initial shelf life, along with the ingredient status data, operational status data, and second environmental data from the multi-dimensional status information, are input into the second sub-model to output the shelf life information of the target ingredient, including: The initial shelf life, along with the ingredient status data, operational status data, and second environmental data from the multidimensional status information, are input into the adjusted second sub-model to output the shelf life information of the target ingredient.

6. The food ingredient management method according to claim 3, characterized in that, After inputting the initial shelf life, along with the food status data, operational status data, and second environment data from the multi-dimensional status information, into the second sub-model and outputting the shelf life information of the target food ingredient, the model further includes: Obtain the shelf-life data corresponding to the target ingredient data that matches the target ingredient from the ingredient database; The target shelf life is corrected based on the shelf life data; the correction includes at least one of the following: The process involves: filtering the shelf-life data to find first shelf-life data that matches the geographical location and name of the target ingredient; correcting the target shelf-life based on the first shelf-life data; filtering the shelf-life data to find second shelf-life data that matches the name and storage device of the target ingredient; correcting the target shelf-life based on the second shelf-life data; and filtering the shelf-life data to find third shelf-life data that matches the name and target user of the target ingredient; and correcting the target shelf-life based on the third shelf-life data.

7. The food ingredient management method according to claim 3, characterized in that, The shelf-life information also includes the shelf-life confidence level at which the target shelf-life confidence level is located; Performing the target management operation to manage the target ingredients includes: Determine the execution method for the target management operation that matches the confidence level of the shelf life; The target management operation is performed based on the execution method to manage the target ingredients; the target management operation includes at least one of the following: displaying the freshness status of the target ingredients on the door display of the ingredient storage device, pushing early warning information to the user terminal corresponding to the ingredient storage device, generating quality change trend information of the target ingredients, and updating the ingredient data associated with the target ingredients in the ingredient database.

8. The food ingredient management method according to claim 1, characterized in that, Also includes: Obtain abnormal environmental data collected by the sensing module of the food storage device; Based on the location information of the sensing unit corresponding to the abnormal environment data, the abnormal environment data, and the attribute information of each food ingredient, candidate spoiled food ingredients are identified from the food ingredients stored in the food ingredient storage device. Based on the target shelf life and food status data of the candidate spoiled food ingredients, a target spoiled food ingredient is determined from the candidate spoiled food ingredients, so as to generate a spoilage warning information of the target spoiled food ingredient and push it to the user terminal that matches the food storage device.

9. A food ingredient management device, characterized in that, Applied to a food storage device, the device includes: The first acquisition module is used to acquire the identification information of the target food stored in the food storage device; The second acquisition module is used to acquire multi-dimensional status information related to the target ingredient based on the identification information. The multi-dimensional status information includes at least two of the following: attribute information of the target ingredient, first environmental data of the environment in which the target ingredient is located, ingredient status data of the target ingredient, operating status data of the ingredient storage device, and second environmental data of the environment in which the ingredient storage device is located. The determination module is used to input the multidimensional state information into a pre-trained shelf-life determination model and output the shelf-life information of the target food ingredient. The operation module is used to determine the target management operation corresponding to the target ingredient based on the shelf life information, and execute the target management operation to manage the target ingredient.

10. A food ingredient management system, characterized in that, include: A food storage device, a sensing module and a management module installed in the food storage device; The sensing module is used to collect multidimensional sensing data of the food stored in the food storage device and send the multidimensional sensing data to the management module. The management module is used to: acquire the identification information of the target food stored in the food storage device, process the multidimensional sensing data, and obtain multidimensional status information related to the target food from the processing result based on the identification information. The multidimensional status information includes at least two of the following: attribute information of the target food, first environmental data of the environment in which the target food is located, food status data of the target food, operating status data of the food storage device, and second environmental data of the environment in which the food storage device is located. The multidimensional state information is input into a pre-trained shelf-life determination model, which outputs the shelf-life information of the target food ingredient. Based on the shelf life information, determine the target management operation corresponding to the target ingredient, and execute the target management operation to manage the target ingredient.

Images