Intelligent control method and system for heat preservation chafing dish

Abstract

The application provides an intelligent control method and system of a heat preservation food warmer, the method comprising: receiving a dish image obtained by the heat preservation food warmer shooting dishes in the warmer through an image acquisition module, and a dish amount value obtained by weighing the dishes through a weighing module; identifying a dish type of the dishes in the warmer according to the dish image, and predicting a dish amount consumption trend of the dish type in a future period according to a consumption rate of the dish type and the dish amount value; when the dish amount consumption trend shows that the dish amount will drop to a critical threshold within a preset time window, generating a dish adding prompt information, and pushing the dish adding prompt information to a mobile terminal in a dish preparation area, so that the dish preparation area prepares dishes and adds dishes in a descending order of priority, to realize early prediction before dish supply interruption occurs, avoid long-term empty warmer phenomenon of the dishes before replenishment, and affect dining continuity and dining experience.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, and in particular to an intelligent control method and system for a food warmer. Background Technology

[0002] In centralized meal supply scenarios in the catering industry, insulated food ovens are the core equipment for keeping food warm and displaying it. The real-time monitoring and accurate prediction of the amount of food inside them are directly related to the continuity of meal supply and the rationality of meal preparation.

[0003] Current technologies primarily rely on manual inspections or simple weight threshold alarm mechanisms. Staff periodically check the remaining food in each warming oven and notify the preparation area to replenish it via phone or walkie-talkie when insufficient food is detected. However, both methods suffer from significant delays: firstly, manual inspections are typically spaced ten minutes or more apart, leading to delayed detection of low food levels and a high risk of food shortages, impacting the dining experience. Secondly, simple weight threshold alarm mechanisms are static, triggering only when the current food level drops to a fixed threshold. Therefore, by the time an alarm is triggered, the food may already be rapidly depleting. The time delay in preparing, cooking, and delivering the food to the warming ovens can easily result in prolonged empty oven periods before replenishment, affecting the continuity of the dining experience and overall dining experience. Summary of the Invention

[0004] This invention provides an intelligent control method and system for a food warmer, which can predict in advance when food supplies run out, thus avoiding a long period of empty ovens before food is replenished, which would affect the continuity of dining and the dining experience.

[0005] To solve the above problems, the present invention adopts the following technical solution: In a first aspect, the present invention provides an intelligent control method for a food warming oven, applied to a server. The food warming oven is equipped with an image acquisition module, a weighing module, and a wireless communication module. The weighing module is located at the bottom of the oven body, and the image acquisition module is located above the opening of the oven body and facing inwards. The food warming oven establishes a connection with the server through the wireless communication module. The intelligent control method includes: The system receives images of the food inside the insulated food oven after the image acquisition module captures the images, and the food quantity is obtained after the weighing module weighs the food. The type of dish in the oven is identified based on the image of the dish. Based on the consumption rate of the dish type and the amount of dish, the consumption trend of the dish type in the future period is predicted. When the trend of food consumption shows that the food quantity will drop to a critical threshold within a preset time window, a food replenishment prompt message is generated. The food replenishment prompt message includes the insulated stove icon, food type, priority, and suggested amount of food to be replenished. The message prompting you to add dishes is pushed to the mobile terminal in the food preparation area so that the food preparation area can prepare and add dishes in order of priority from high to low.

[0006] Preferably, predicting the consumption trend of the dish type in future time periods based on the consumption rate of the dish type and the dish quantity includes: The customer flow density of the dining environment is obtained, and the customer flow density is correlated with the consumption rate. The analysis yields the time lag correlation between food consumption and customer flow fluctuations. The time lag correlation characterizes the time delay and correlation strength between changes in customer flow density and changes in food consumption. Based on the time lag correlation, a consumption trend prediction model is constructed. The current trend of passenger flow density change and the value of the dish quantity are input into the consumption trend prediction model, and the consumption trend of the dish type in the future time period is output. The consumption trend includes the predicted value of the dish quantity for each future time window. The consumption trend prediction model is a long short-term memory neural network model.

[0007] Preferably, the step of constructing a consumption trend prediction model based on the time-lag correlation includes: The length of the input sequence of the long short-term memory neural network model is determined based on the delay time characteristics, so that the input sequence covers the complete delay response cycle; The weight initialization strategy for each gating unit in the long short-term memory neural network model is determined based on the correlation strength characteristics. The historical passenger flow density sequence, consumption rate sequence, and dish quantity value sequence are time-aligned according to the aforementioned delay time feature to construct a training sample set; The long short-term memory neural network model is trained using the training sample set, enabling the model to learn the nonlinear mapping relationship between changes in passenger flow density and changes in food consumption after a delay time, thus obtaining a trained consumption trend prediction model.

[0008] Preferably, the step of inputting the current customer flow density change trend and the dish quantity value into the consumption trend prediction model, and outputting the dish quantity consumption trend of the dish type in future time periods, includes: The current passenger flow density change trend is decomposed into three types of time series characteristics: instantaneous change rate, short-cycle fluctuation trend, and long-cycle evolution trend. The three types of time-series features are concatenated with the amount of food to construct a multi-dimensional input vector. The multi-dimensional input vector is then input into the input gate of the consumption trend prediction model. The input gate calculates the candidate memory state based on the multi-dimensional input vector at the current time and the hidden state at the previous time. The delay time feature corresponding to the time lag correlation is input into the forget gate of the consumption trend prediction model. The forget gate dynamically adjusts the retention ratio of historical memory state according to the delay time feature and the change range of current passenger flow density. The input gate and the forget gate work together to update the current memory unit state. The updated memory unit state is then input into the output gate of the consumption trend prediction model. The output gate calculates the current hidden state based on the current candidate memory state and the historical memory state. The hidden state at each time step is input into a fully connected layer. The fully connected layer converts the hidden state at each time step into the predicted quantity of food for the corresponding future time window through a non-linear mapping relationship. Based on the predicted quantity of food, the consumption trend of the food type in the future time period is constructed.

[0009] Furthermore, after predicting the consumption trend of the dish type in future time periods based on the consumption rate of the dish type and the dish quantity value, the method further includes: When the current trend of passenger flow density shows that the passenger flow density will increase in the future, the critical threshold is raised so that the triggering time of the additional food prompt is brought forward. When the current trend of passenger flow density shows that the passenger flow density will decrease in the future, the critical threshold will be lowered to avoid triggering the additional food prompt message too early, which would cause food waste.

[0010] Furthermore, after receiving the image of the food inside the insulated food oven obtained by the image acquisition module, the system further includes: Edge detection is performed on the food image to extract the boundary contour lines of the food area; Calculate the centroid position of the region enclosed by the boundary contour lines; When the distance between the centroid position and the central axis of the oven is greater than a preset distance, it is determined that the distribution of dishes in the oven is biased, and a dish distribution correction prompt is triggered to the mobile terminal in the food preparation area.

[0011] Furthermore, after identifying the type of dish in the oven based on the dish image, the process further includes: Receives multi-point temperature data collected by the insulated food oven through a temperature sensor array; The maximum temperature difference of the dishes in the oven is calculated based on the multi-point temperature data, and the allowable temperature fluctuation range of the dish type is queried. When the maximum temperature difference is greater than the allowable temperature fluctuation range, the heating power of the insulated food oven is reduced in the oven area where the temperature is greater than the preset temperature value, and the heating power of the oven area where the temperature is less than the preset temperature value is increased, until the maximum temperature difference is less than or equal to the allowable temperature fluctuation range.

[0012] Preferably, the priority setting method includes: Calculate the remaining time for the amount of food in the oven to drop to the critical threshold based on the stated food consumption trend; The priority of the dish type is set according to the remaining time, and the priority is negatively correlated with the remaining time.

[0013] Furthermore, the intelligent control method for the heat-insulating oven also includes: Receive remote control commands sent by the mobile terminal, the remote control commands including commands to adjust the temperature value of the food warmer and commands to set the warming time; The remote control command is sent to the insulated food oven, so that the insulated food oven performs the corresponding operation according to the remote control command and feeds back the execution result, and forwards the execution result to the mobile terminal.

[0014] Secondly, the present invention also provides an intelligent control system for a food warming oven, the intelligent control system comprising a food warming oven, a server and a mobile terminal, the food warming oven being equipped with an image acquisition module, a weighing module and a wireless communication module, the weighing module being disposed at the bottom of the oven body, the image acquisition module being disposed above the opening of the oven body and facing inwards, the food warming oven establishing a connection with the server through the wireless communication module, the server also establishing a connection with the mobile terminal, the server being used to execute the intelligent control method for the food warming oven as described in any of the preceding claims.

[0015] Compared with the prior art, the technical solution of the present invention has at least the following advantages: This invention provides an intelligent control method and system for a food warming stove. It acquires images of the food using an image acquisition module deployed above the stove opening, and combines this with a weighing module located at the bottom of the stove to obtain the quantity of food, enabling real-time perception of the food's status within the stove. Based on the food images, it identifies the food type and predicts the future consumption trend of the food type based on its consumption rate and current quantity. This transforms meal preparation decisions from reactive to proactive, preventing prolonged periods of empty stoves before replenishment, which affects the continuity of dining and the dining experience. When the prediction indicates that the food quantity will drop to a critical threshold within a preset time window, it automatically generates a message containing the stove's identifier, food type, priority, and suggested replenishment quantity, and pushes this message to a mobile terminal in the preparation area. This allows staff to prepare and replenish food according to priority, effectively avoiding the risk of food shortages, improving preparation efficiency and resource utilization, and preventing food waste or deterioration in taste caused by excessive pre-preparation. Attached Figure Description

[0016] Figure 1 This is a flowchart of an embodiment of an intelligent control method for a food warmer according to the present invention; Figure 2 This is a flowchart illustrating another embodiment of the intelligent control method for a heat-insulating food stove according to the present invention; Figure 3 This is a structural block diagram of an embodiment of the intelligent control system for a heat-insulating food stove according to the present invention. Detailed Implementation

[0017] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0018] Please refer to Figure 1 As shown, this invention provides an intelligent control method for a food warmer, applied to a server. The food warmer is equipped with an image acquisition module, a weighing module, and a wireless communication module. The weighing module is located at the bottom of the furnace body, and the image acquisition module is located above the furnace opening and facing inwards. The food warmer establishes a connection with the server through the wireless communication module. The intelligent control method includes the following steps: S11. Receive the image of the food inside the oven obtained by the image acquisition module after the food is captured by the insulated food oven, and the food quantity value obtained by the weighing module after the food is weighed. S12. Based on the image of the dish, identify the type of dish in the oven, and based on the consumption rate of the dish type and the amount of dish, predict the consumption trend of the amount of dish type in the future period. S13. When the trend of food consumption shows that the food quantity will drop to a critical threshold within a preset time window, a food replenishment prompt message is generated. The food replenishment prompt message includes the insulated stove icon, food type, priority, and suggested food replenishment quantity. S14. Push the added dish prompt information to the mobile terminal in the food preparation area so that the food preparation area can prepare and add dishes in order of priority from high to low.

[0019] The image acquisition module includes a camera device that takes pictures of the food inside the oven at fixed time intervals to acquire image data containing the visual features of the food. The weighing module collects the total weight of the oven and the food in real time, and obtains the net weight of the food as the quantity value after tare processing. Both types of data are synchronously uploaded to the server via a wireless communication module using a preset transmission protocol.

[0020] After receiving the dish image and quantity, the server inputs the dish image into a pre-trained dish recognition model, extracts the color, texture, and shape features of the dish, compares and matches these features with the sample features in the dish type database, and outputs the matching dish type recognition result, such as identifying it as a specific dish category like braised pork, stir-fried vegetables, or rice.

[0021] Based on the consumption rate and quantity of the identified dish type, the server predicts the consumption trend of that dish type over a future period. The consumption rate can be calculated based on historical weight data changes of that dish type during the current dining period, reflecting the average consumption speed per unit time. The server inputs the current quantity and consumption rate into a pre-trained consumption trend prediction model, which then uses this model to calculate the quantity change curve of that dish type over a preset future time period, thus obtaining the consumption trend.

[0022] The critical threshold can be preset according to the type of dish, representing the minimum amount of food needed to maintain normal supply. The preset time window can be set according to the average preparation time to ensure that the food is not exhausted before preparation is completed. The food replenishment prompt information includes a warming stove icon to locate the specific stove; the type of dish to indicate which dish needs to be replenished; the priority, which can be dynamically calculated based on the urgency of the food quantity dropping to the critical threshold; and the suggested amount of food to replenish, which can be calculated based on the current consumption trend and the capacity of the warming stove.

[0023] The server pushes food replenishment notifications to mobile terminals in the food preparation area. The mobile terminals display these notifications in a list format, automatically sorting them by priority from highest to lowest. Staff then prepare and replenish food according to this sorting, prioritizing tasks with higher urgency. It's important to understand that food preparation can include preparing and cooking ingredients, or directly cooking pre-prepared ingredients. Replenishing food involves transporting the cooked dishes to the designated food warming oven.

[0024] This invention provides an intelligent control method for a food warming stove. It acquires images of the food using an image acquisition module deployed above the stove opening, and combines this with a weighing module located at the bottom of the stove to obtain the quantity of food, enabling real-time perception of the food's status within the stove. Based on the food images, the method identifies the food type and predicts the future consumption trend based on the consumption rate of that type and the current quantity. This transforms meal preparation decisions from reactive responses to proactive predictions, preventing prolonged periods of empty stoves before replenishment, which affects the continuity of dining and the dining experience. When the prediction indicates that the food quantity will drop to a critical threshold within a preset time window, the method automatically generates a message containing the stove's identifier, the food type, priority, and suggested replenishment quantity, and pushes this message to a mobile terminal in the preparation area. This allows staff to prepare and replenish food according to priority, effectively avoiding the risk of food shortages, improving preparation efficiency and resource utilization, and preventing food waste or deterioration in taste caused by over-preparation.

[0025] In one embodiment, please refer to Figure 2 As shown, predicting the consumption trend of a dish type in future time periods based on the consumption rate of the dish type and the dish quantity includes: S121. Obtain the customer flow density of the dining environment, perform a correlation analysis between the customer flow density and the consumption rate, and analyze the time lag correlation between food consumption and customer flow fluctuation. The time lag correlation characterizes the delay time feature and correlation strength feature between the change in customer flow density and the change in food consumption. S122. Construct a consumption trend prediction model based on the time lag correlation, input the current passenger flow density change trend and the dish quantity value into the consumption trend prediction model, and output the dish quantity consumption trend of the dish type in the future time period. The dish quantity consumption trend includes the predicted dish quantity value for each future time window. The consumption trend prediction model is a long short-term memory neural network model.

[0026] Customer flow density is statistically analyzed in real time using video capture devices or infrared sensors deployed in the dining area, reflecting the current distribution of diners per unit area. The video capture devices or infrared sensors send customer flow density data to the server at fixed time intervals, forming a time-series sequence of customer flow density. The server receives this time-series sequence and performs a sliding window correlation calculation between it and the food consumption time-series sequence to determine the time delay characteristic of the transmission from changes in customer flow density to changes in food consumption—that is, how long it takes for food consumption to significantly increase after an increase in customer flow. Simultaneously, the correlation coefficient between the two sequences is calculated to obtain the correlation strength characteristic, which characterizes the degree of impact of customer flow fluctuations on food consumption. For example, some food types may have an immediate consumption response, with customer flow growth and food consumption changing almost synchronously; while other food types may have a delayed consumption response, requiring a period of time for food consumption to increase significantly after a customer flow increase.

[0027] The server builds a prediction framework based on a long short-term memory neural network model. This model includes structures such as input gates, forget gates, and output gates, which can effectively capture long-term dependencies in time-series data. The server uses the historical passenger flow density change trend as an external input variable and the corresponding historical food quantity value as a state variable. Combined with the delay time parameter determined by the time lag correlation, the server trains a consumption trend prediction model.

[0028] The server inputs the current trend of customer flow density and the amount of food consumed into a pre-trained consumption trend prediction model, and outputs the consumption trend of food types in future time periods. This consumption trend prediction model can calculate the predicted amount of food consumed in each future time window based on the input trend of rising or falling customer flow density and the time-lag response patterns obtained from historical training, forming a complete consumption trend curve.

[0029] In this embodiment, by introducing the dining environment's customer flow density as an external correlation variable and analyzing its time-lag correlation with food consumption, a precise quantitative characterization of the dynamic lag effect of customer flow fluctuations on food consumption is achieved. Furthermore, by constructing a consumption trend prediction model using a long short-term memory neural network model, and fully utilizing its gating mechanism's ability to model time-series dependencies, the model integrates the customer flow density change trend with the current food consumption value for prediction. This enables the food consumption trend prediction to proactively respond to changes in customer flow, significantly improving prediction accuracy and lead time, effectively avoiding prediction failures caused by sudden changes in customer flow, and further ensuring the stability and timeliness of food supply.

[0030] In one embodiment, constructing a consumption trend prediction model based on the time lag correlation includes: The length of the input sequence of the long short-term memory neural network model is determined based on the delay time characteristics, so that the input sequence covers the complete delay response cycle; The weight initialization strategy for each gating unit in the long short-term memory neural network model is determined based on the correlation strength characteristics. The historical passenger flow density sequence, consumption rate sequence, and dish quantity value sequence are time-aligned according to the aforementioned delay time feature to construct a training sample set; The long short-term memory neural network model is trained using the training sample set, enabling the model to learn the nonlinear mapping relationship between changes in passenger flow density and changes in food consumption after a delay time, thus obtaining a trained consumption trend prediction model.

[0031] The delay time feature characterizes the response time required for changes in passenger flow density to be transmitted to changes in food consumption. The server sets the length of the input sequence to cover the entire delay response period, ensuring that the model input contains information about the entire process from the start of passenger flow fluctuations to the response of food consumption. This avoids the model being unable to capture the complete time delay dependency due to an input sequence that is too short, or introducing irrelevant noise that interferes with the prediction accuracy due to an input sequence that is too long.

[0032] Furthermore, the correlation strength feature reflects the degree of influence between customer flow density and food consumption. For food types with high correlation strength, the server adopts an initialization strategy that increases the weight of the input gate to enhance the model's sensitivity to the input variable of customer flow density. For food types with low correlation strength, the server adopts an initialization strategy that increases the weight of the forget gate to make the model focus more on maintaining historical food consumption information, weaken the interference of weakly correlated external inputs, and achieve adaptive matching between the model initialization parameters and food characteristics.

[0033] The server aligns historical passenger flow density, consumption rate, and food quantity sequences according to time delay features to construct a training sample set. Specifically, the server shifts the passenger flow density sequence forward using the time delay feature as an offset, aligning the shifted passenger flow density data points with the corresponding delayed food quantity consumption change data points on the time axis. The time-aligned passenger flow density, consumption rate, and food quantity sequences are then sliced ​​according to a preset time window, generating multiple sets of input-output paired data to construct the training sample set. The input features include the aligned passenger flow density, consumption rate, and current food quantity value, while the output label is the food quantity value for a future time period.

[0034] The server trains a long short-term memory neural network model using a training sample set, enabling the model to learn the nonlinear mapping relationship between changes in passenger flow density and changes in food consumption after a delay. Specifically, during training, the model calculates predicted food consumption values ​​through forward propagation, updates the weight parameters of the gating units through backpropagation, and iteratively optimizes until the prediction error converges to a preset range, resulting in a trained consumption trend prediction model. This model can output the future food consumption trend, taking into account the delay response effect, based on the input current passenger flow density change trend and food consumption value.

[0035] In this embodiment, by determining the length of the input sequence based on the delay time characteristics, the model input covers the entire delay response cycle, effectively ensuring the complete modeling of time-delay dependencies. Secondly, by setting the weight initialization strategy of the gating unit based on the association strength characteristics, adaptive matching between model parameters and dish type characteristics is achieved, improving the model's adaptability to scenarios with different association strengths. Furthermore, by aligning the time of multi-source time-series data according to the delay time characteristics, training samples conforming to the physical transmission mechanism are constructed, making the model learning process more aligned with actual business logic. Finally, the consumption trend prediction model obtained through training can accurately depict the nonlinear mapping relationship between changes in customer flow density and changes in dish consumption after the delay time, significantly improving the predictive model's adaptability and prediction accuracy in complex dynamic environments.

[0036] In one embodiment, inputting the current customer flow density change trend and the dish quantity value into the consumption trend prediction model, and outputting the dish quantity consumption trend of the dish type in future time periods, includes: The current passenger flow density change trend is decomposed into three types of time series characteristics: instantaneous change rate, short-cycle fluctuation trend, and long-cycle evolution trend. The three types of time-series features are concatenated with the amount of food to construct a multi-dimensional input vector. The multi-dimensional input vector is then input into the input gate of the consumption trend prediction model. The input gate calculates the candidate memory state based on the multi-dimensional input vector at the current time and the hidden state at the previous time. The delay time feature corresponding to the time lag correlation is input into the forget gate of the consumption trend prediction model. The forget gate dynamically adjusts the retention ratio of historical memory state according to the delay time feature and the change range of current passenger flow density. The input gate and the forget gate work together to update the current memory unit state. The updated memory unit state is then input into the output gate of the consumption trend prediction model. The output gate calculates the current hidden state based on the current candidate memory state and the historical memory state. The hidden state at each time step is input into a fully connected layer. The fully connected layer converts the hidden state at each time step into the predicted quantity of food for the corresponding future time window through a non-linear mapping relationship. Based on the predicted quantity of food, the consumption trend of the food type in the future time period is constructed.

[0037] The instantaneous rate of change can be obtained by calculating the difference in passenger flow density between adjacent time points, reflecting the immediate increase or decrease in passenger flow; the short-term fluctuation trend can be extracted by moving average filtering, which reveals the local oscillation pattern of passenger flow after filtering out instantaneous noise; the long-term evolution trend can be obtained by low-pass filtering, characterizing the overall trend of passenger flow over a longer period. These three types of time-series features characterize the dynamic characteristics of passenger flow density at different time scales, providing hierarchical input information for the model.

[0038] The channel splicing operation stacks the instantaneous rate of change, short-cycle fluctuation trend, long-cycle evolution trend, and current food quantity value on the feature dimension to form a comprehensive representation that includes multi-scale dynamic information of customer flow and the current state of the food. The server inputs this multi-dimensional input vector into the input gate of the consumption trend prediction model. The input gate calculates the candidate memory state through the activation function based on the multi-dimensional input vector at the current time and the hidden state at the previous time. The candidate memory state contains the latent feature representation of the current input information after nonlinear transformation.

[0039] The server inputs the delay time feature corresponding to the time lag correlation into the forget gate of the consumption trend prediction model. This delay time feature serves as prior knowledge to guide the information filtering behavior of the forget gate. The forget gate dynamically adjusts the retention ratio of historical memory states based on the delay time feature and the magnitude of changes in current passenger flow density. Specifically, when the magnitude of changes in current passenger flow density exceeds a preset threshold and the delay time feature indicates that the response period is about to begin, the forget gate reduces the retention ratio of historical memory states, making the model pay more attention to the new information currently input; conversely, it increases the retention ratio of historical memory states to maintain the model's stable memory of long-term patterns.

[0040] The candidate memory state output by the input gate is weighted and fused with the historical memory state filtered by the forget gate to form an updated memory unit state. This state contains both historical accumulated information and current observation information. The server inputs the updated memory unit state into the output gate of the consumption trend prediction model. The output gate calculates the hidden state at the current time through an activation function based on the candidate memory state at the current time and the historical memory state adjusted by the forget gate. This hidden state is the effective output of the memory unit state after being filtered by the output gate.

[0041] The server inputs the hidden states at each time step into a fully connected layer. The fully connected layer converts the hidden states at each time step into the predicted quantities of food for the corresponding future time window through a non-linear mapping relationship. This non-linear mapping relationship is learned through the training process and can decode the high-dimensional hidden states into interpretable food quantity values. Finally, the server arranges the predicted food quantity values ​​for each future time window in chronological order to construct a complete trend of food quantity consumption for each dish type in the future period.

[0042] In this embodiment, by decomposing the passenger flow density change trend into a multi-scale time window, multi-level feature extraction of instantaneous fluctuations, short-cycle oscillations, and long-cycle evolution is achieved, enriching the information dimension of the model input. Furthermore, by concatenating multi-scale time-series features with food quantity values, a comprehensive input representation integrating the external environment and internal state is constructed, enhancing the model's ability to perceive complex scenarios. Further, by inputting delayed time features into the forget gate, time-delay prior knowledge guides the gating behavior, enabling the model to dynamically adjust its memory update strategy based on the magnitude of passenger flow changes and the delayed response cycle, improving the model's adaptability to time-delay effects. Moreover, through the synergistic effect of the input gate, forget gate, and output gate, a fine balance is achieved between preserving historical patterns and integrating current information. Finally, the nonlinear mapping of the fully connected layer outputs the predicted food quantity values ​​for each future time window, significantly improving the accuracy and reliability of food consumption trend prediction.

[0043] In one embodiment, after predicting the consumption trend of the dish type in future time periods based on the consumption rate of the dish type and the dish quantity value, the method further includes: When the current trend of passenger flow density shows that the passenger flow density will increase in the future, the critical threshold is raised so that the triggering time of the additional food prompt is brought forward. When the current trend of passenger flow density shows that the passenger flow density will decrease in the future, the critical threshold will be lowered to avoid triggering the additional food prompt message too early, which would cause food waste.

[0044] Specifically, the server analyzes the directional characteristics of long-term evolution trends. When it determines that the customer flow density is about to enter an increase phase, it calculates the threshold adjustment amount based on the increase rate, making the critical threshold higher than the original set value. The increased critical threshold makes it easier for the food consumption trend to meet the condition of falling to the critical threshold within a preset time window, thereby triggering the generation of additional food prompts in advance and reserving more response time for food preparation operations to cope with the upcoming peak customer flow demand.

[0045] When the server determines that customer flow density is about to decline, it calculates the threshold reduction based on the decline rate, making the critical threshold lower than the original set value. The lowered critical threshold makes it more difficult for the food consumption trend to meet the trigger conditions, delaying the generation of food replenishment prompts. This avoids triggering food preparation instructions too early during periods of reduced customer flow, preventing food stockpiling and energy waste caused by excessive food preparation.

[0046] In this embodiment, a dynamic adjustment mechanism for critical thresholds is introduced to achieve a linkage response between the triggering conditions for adding dishes and changes in the customer flow environment. When customer flow is rising, the critical threshold is raised, effectively advancing the triggering time of adding dishes and avoiding the risk of supply shortages during peak periods. When customer flow is declining, the critical threshold is lowered, reasonably delaying the triggering time of adding dishes, suppressing excessive food preparation during low-demand periods, and preventing food stockpiling and energy waste caused by excessive food preparation. This dynamic adjustment mechanism transforms the decision-making process for adding dishes from a fixed parameter mode to an environment-adaptive mode, significantly improving the flexibility and accuracy of food preparation resource allocation, and reducing operating costs and food waste while ensuring a continuous supply of dishes.

[0047] In one embodiment, after receiving the image of the food inside the insulated food oven obtained by the image acquisition module, the method further includes: Edge detection is performed on the food image to extract the boundary contour lines of the food area; Calculate the centroid position of the region enclosed by the boundary contour lines; When the distance between the centroid position and the central axis of the oven is greater than a preset distance, it is determined that the distribution of dishes in the oven is biased, and a dish distribution correction prompt is triggered to the mobile terminal in the food preparation area.

[0048] The server can use a gradient-based edge detection algorithm to analyze the pixel grayscale changes in the dish image, identify the boundary position between the dish and the inner wall of the oven or between the dish and an empty area, and form a continuous boundary contour line by connecting the boundary points. This contour line completely depicts the space occupied by the dish inside the oven.

[0049] The server uses the area inside the boundary outline as the effective area for each dish. It then calculates a weighted average of the coordinates of each pixel within this area to obtain the centroid position, which represents the center of the dish's quality distribution. The centroid position reflects the concentrated distribution point of the dishes within the oven space and is a key indicator for judging the evenness of the dish distribution.

[0050] The server compares the centroid position with the central axis of the oven and calculates the spatial distance between them. When the centroid position deviates from the central axis of the oven by a preset distance, it is determined that the food distribution inside the oven is biased. This bias indicates that the food is piled up on one side or in a localized area, which may lead to uneven heat preservation, localized overheating, or inconvenience in retrieving the food. The server triggers a food distribution correction prompt, which includes the oven's logo and the bias location information, and pushes it to the mobile terminal in the food preparation area, prompting the staff to stir or re-level the food inside the oven.

[0051] In this embodiment, by calculating the centroid position of the area enclosed by the boundary contour line and comparing it with the deviation of the central axis of the furnace body, the automatic identification of the food offset shape is realized. In addition, by triggering the food distribution correction prompt and pushing it to the mobile terminal, the food preparation staff can be notified and deal with the distribution abnormality in a timely manner, effectively avoiding problems such as local heat preservation failure, quality deterioration and difficulty in retrieval caused by food accumulation and offset, thus improving the operation and maintenance quality of the insulated food furnace and the dining experience.

[0052] In one embodiment, after identifying the type of dish in the oven based on the dish image, the method further includes: Receives multi-point temperature data collected by the insulated food oven through a temperature sensor array; The maximum temperature difference of the dishes in the oven is calculated based on the multi-point temperature data, and the allowable temperature fluctuation range of the dish type is queried. When the maximum temperature difference is greater than the allowable temperature fluctuation range, the heating power of the insulated food oven is reduced in the oven area where the temperature is greater than the preset temperature value, and the heating power of the oven area where the temperature is less than the preset temperature value is increased, until the maximum temperature difference is less than or equal to the allowable temperature fluctuation range. Temperature sensor arrays are deployed in a grid pattern on the bottom or side walls of the oven, covering the main areas where the food is placed inside. Each sensor collects the temperature value at its location at fixed time intervals, forming a multi-point temperature dataset that characterizes the temperature field distribution inside the oven, and then uploads it to the server.

[0053] The server iterates through multiple temperature data points, extracts the highest and lowest temperature values, and calculates the difference between them as the maximum temperature difference. At the same time, based on the identified dish type, the server searches the dish temperature control parameter library to obtain the allowable temperature fluctuation range corresponding to that dish type. The allowable temperature fluctuation range is preset according to the physical characteristics and quality maintenance requirements of the dish.

[0054] When the maximum temperature difference exceeds the allowable temperature fluctuation range, the server determines that the temperature distribution inside the oven is uneven and exceeds the tolerance of the dishes, and generates a zoned temperature control command. Specifically, the server identifies areas inside the oven where the temperature is higher than a preset temperature value, which can be an ideal heat preservation temperature set via a mobile terminal in the food preparation area. Different types of dishes have different ideal heat preservation temperatures. Subsequently, the server sends a control command to the oven to reduce the heating power of that area to suppress local overheating. At the same time, it identifies areas inside the oven where the temperature is lower than the preset temperature value and sends a control command to increase the heating power of that area to enhance local heating capacity. The oven can independently adjust each heating zone according to the commands, realizing dynamic redistribution of heat.

[0055] The server continuously receives temperature data from multiple points, iteratively calculates the maximum temperature difference, and continues until the maximum temperature difference is less than or equal to the allowable temperature fluctuation range. Once the server determines that the temperature distribution within the oven meets the requirements for maintaining food quality, it terminates the zoned temperature control adjustment and maintains the current heating power configuration. By identifying overheated and underheated areas and implementing targeted power adjustments, the server achieves dynamic heat redistribution and balanced temperature control. Furthermore, the oven can adaptively adjust its temperature control strategy according to the type of dish, effectively suppressing food quality deterioration caused by localized overheating or insulation failure caused by localized underheating. This reduces energy consumption while ensuring food taste and safety, and improves the intelligent management level of the oven.

[0056] Preferably, the priority setting method may specifically include: Calculate the remaining time for the amount of food in the oven to drop to the critical threshold based on the stated food consumption trend; The priority of the dish type is set according to the remaining time, and the priority is negatively correlated with the remaining time.

[0057] The trend of vegetable consumption includes a sequence of predicted vegetable quantities for each future time window. The server traverses this sequence to locate the time window corresponding to the first drop in the predicted vegetable quantity to the critical threshold. The time span from the current moment to this time window is the remaining duration, which represents the duration for which the vegetables can maintain a normal supply.

[0058] The server establishes a mapping relationship between remaining time and priority, making the priority negatively correlated with the remaining time. Specifically, the shorter the remaining time, the higher the risk of food running out, and the higher the corresponding priority; conversely, the longer the remaining time, the more abundant the food supply, and the lower the corresponding priority. The server assigns this priority value to the priority field in the food replenishment prompt information, enabling mobile terminals in the food preparation area to sort and display the food replenishment tasks of each insulated food stove according to priority, prioritizing the replenishment of food that is about to run out, effectively avoiding the resource scheduling chaos problem in multi-insulated food stove scenarios, and significantly improving food preparation response efficiency and food supply continuity.

[0059] In one embodiment, the intelligent control method for the insulated food oven further includes: Receive remote control commands sent by the mobile terminal, the remote control commands including commands to adjust the temperature value of the food warmer and commands to set the warming time; The remote control command is sent to the insulated food oven, so that the insulated food oven performs the corresponding operation according to the remote control command and feeds back the execution result, and forwards the execution result to the mobile terminal.

[0060] Food preparation staff can use a mobile terminal's human-computer interaction interface to input temperature adjustment commands to modify the heating temperature of the keep-warm oven, or input keep-warm duration setting commands to adjust the duration of automatic keep-warming, based on the current status of the dishes or unexpected needs. The mobile terminal then encrypts and uploads the remote control commands, which include the target keep-warm oven's identifier and control parameters, to the server.

[0061] The server verifies the validity of remote control commands. If the verification is successful, it forwards the command to the corresponding insulated food heater. The insulated food heater receives the remote control command via its wireless communication module, parses the command type and parameter content, and drives the temperature control module to adjust the temperature or the timing module to set the heat preservation time. After execution, the insulated food heater sends the execution result, including the current temperature, remaining heat preservation time, and execution status code, back to the server.

[0062] The server receives the execution result from the insulated food heater and forwards it to the mobile terminal that initiated the control. The mobile terminal parses the execution result and displays the real-time status after the control on the human-computer interaction interface.

[0063] In this embodiment, a two-way communication channel between the server and the mobile terminal enables remote control of the insulated food oven by the food preparation personnel. Secondly, by supporting the flexible issuance of temperature adjustment commands and insulated time setting commands, the insulated parameters can be dynamically optimized according to the characteristics of the dishes or on-site needs. Furthermore, the feedback and forwarding mechanism for execution results ensures the traceability and visibility of remote operations, significantly improving the management flexibility and maintenance efficiency of the insulated food oven and reducing the cost of manual inspections.

[0064] Please refer to Figure 3 As shown, an embodiment of the present invention also provides an intelligent control system for a food warming stove. The intelligent control system includes a food warming stove 100, a server 200, and a mobile terminal 300. The food warming stove 100 is equipped with an image acquisition module, a weighing module, and a wireless communication module. The weighing module is located at the bottom of the stove body, and the image acquisition module is located above the opening of the stove body and facing inward. The food warming stove 100 establishes a connection with the server 200 through the wireless communication module. The server 200 also establishes a communication connection with the mobile terminal 300. The server 200 is used to execute the intelligent control method of the food warming stove.

[0065] Regarding the system in the above embodiments, the specific manner in which each module performs its operations has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0066] In one embodiment, the present invention also provides a storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the aforementioned intelligent control method for a food warmer. The storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device, etc.

[0067] 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. This computer program can be stored in a storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0068] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0069] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. An intelligent control method of a thermal pot, characterized in that, Applied to a server, the insulated food oven is equipped with an image acquisition module, a weighing module, and a wireless communication module. The weighing module is located at the bottom of the oven body, and the image acquisition module is located above the opening of the oven body and faces inward. The insulated food oven establishes a connection with the server through the wireless communication module. The intelligent control method includes: The system receives images of the food inside the insulated food oven after the image acquisition module captures the images, and the food quantity is obtained after the weighing module weighs the food. The type of dish in the oven is identified based on the image of the dish. Based on the consumption rate of the dish type and the amount of dish, the consumption trend of the dish type in the future period is predicted. When the trend of food consumption shows that the food quantity will drop to a critical threshold within a preset time window, a food replenishment prompt message is generated. The food replenishment prompt message includes the insulated stove icon, food type, priority, and suggested amount of food to be replenished. The message prompting you to add dishes is pushed to the mobile terminal in the food preparation area so that the food preparation area can prepare and add dishes in order of priority from high to low.

2. The method according to claim 1, characterized in that, The step of predicting the consumption trend of the dish type in future time periods based on the consumption rate of the dish type and the dish quantity includes: The customer flow density of the dining environment is obtained, and the customer flow density is correlated with the consumption rate. The analysis yields the time lag correlation between food consumption and customer flow fluctuations. The time lag correlation characterizes the time delay and correlation strength between changes in customer flow density and changes in food consumption. Based on the time lag correlation, a consumption trend prediction model is constructed. The current trend of passenger flow density change and the value of the dish quantity are input into the consumption trend prediction model, and the consumption trend of the dish type in the future time period is output. The consumption trend includes the predicted value of the dish quantity for each future time window. The consumption trend prediction model is a long short-term memory neural network model.

3. The method according to claim 2, characterized in that, The step of constructing a consumption trend prediction model based on the time lag correlation includes: The length of the input sequence of the long short-term memory neural network model is determined based on the delay time characteristics, so that the input sequence covers the complete delay response cycle; The weight initialization strategy for each gating unit in the long short-term memory neural network model is determined based on the correlation strength characteristics. The historical passenger flow density sequence, consumption rate sequence, and dish quantity value sequence are time-aligned according to the aforementioned delay time feature to construct a training sample set; The long short-term memory neural network model is trained using the training sample set, enabling the model to learn the nonlinear mapping relationship between changes in passenger flow density and changes in food consumption after a delay time, thus obtaining a trained consumption trend prediction model.

4. The method according to claim 2, characterized in that, The step of inputting the current trend of passenger flow density change and the value of the dish quantity into the consumption trend prediction model, and outputting the consumption trend of the dish type in future time periods, includes: The current passenger flow density change trend is decomposed into three types of time series characteristics: instantaneous change rate, short-cycle fluctuation trend, and long-cycle evolution trend. The three types of time-series features are concatenated with the amount of food to construct a multi-dimensional input vector. The multi-dimensional input vector is then input into the input gate of the consumption trend prediction model. The input gate calculates the candidate memory state based on the multi-dimensional input vector at the current time and the hidden state at the previous time. The delay time feature corresponding to the time lag correlation is input into the forget gate of the consumption trend prediction model. The forget gate dynamically adjusts the retention ratio of historical memory state according to the delay time feature and the change range of current passenger flow density. The input gate and the forget gate work together to update the current memory unit state. The updated memory unit state is then input into the output gate of the consumption trend prediction model. The output gate calculates the current hidden state based on the current candidate memory state and the historical memory state. The hidden state at each time step is input into a fully connected layer. The fully connected layer converts the hidden state at each time step into the predicted quantity of food for the corresponding future time window through a non-linear mapping relationship. Based on the predicted quantity of food, the consumption trend of the food type in the future time period is constructed.

5. The method according to claim 2, characterized in that, After predicting the consumption trend of the dish type in the future period based on the consumption rate of the dish type and the dish quantity, the method further includes: When the current trend of passenger flow density shows that the passenger flow density will increase in the future, the critical threshold is raised so that the triggering time of the additional food prompt is brought forward. When the current trend of passenger flow density shows that the passenger flow density will decrease in the future, the critical threshold will be lowered to avoid triggering the additional food prompt message too early, which would cause food waste.

6. The method according to claim 1, characterized in that, After the image acquisition module captures images of the food inside the insulated food oven, the receiving oven also includes: Edge detection is performed on the food image to extract the boundary contour lines of the food area; Calculate the centroid position of the region enclosed by the boundary contour lines; When the distance between the centroid position and the central axis of the oven is greater than a preset distance, it is determined that the distribution of dishes in the oven is biased, and a dish distribution correction prompt is triggered to the mobile terminal in the food preparation area.

7. The method according to claim 1, characterized in that, After identifying the type of dish in the oven based on the dish image, the process further includes: Receives multi-point temperature data collected by the insulated food oven through a temperature sensor array; The maximum temperature difference of the dishes in the oven is calculated based on the multi-point temperature data, and the allowable temperature fluctuation range of the dish type is queried. When the maximum temperature difference is greater than the allowable temperature fluctuation range, the heating power of the insulated food oven is reduced in the oven area where the temperature is greater than the preset temperature value, and the heating power of the oven area where the temperature is less than the preset temperature value is increased, until the maximum temperature difference is less than or equal to the allowable temperature fluctuation range.

8. The method according to claim 1, characterized in that, The priority setting methods include: Calculate the remaining time for the amount of food in the oven to drop to the critical threshold based on the stated food consumption trend; The priority of the dish type is set according to the remaining time, and the priority is negatively correlated with the remaining time.

9. The method according to claim 1, characterized in that, Also includes: Receive remote control commands sent by the mobile terminal, the remote control commands including commands to adjust the temperature value of the food warmer and commands to set the warming time; The remote control command is sent to the insulated food oven, so that the insulated food oven performs the corresponding operation according to the remote control command and feeds back the execution result, and forwards the execution result to the mobile terminal.

10. An intelligent control system for a food warmer, characterized in that, The intelligent control system includes a food warmer, a server, and a mobile terminal. The food warmer is equipped with an image acquisition module, a weighing module, and a wireless communication module. The weighing module is located at the bottom of the furnace body, and the image acquisition module is located above the furnace body opening and facing inward. The food warmer establishes a connection with the server through the wireless communication module, and the server also establishes a connection with the mobile terminal. The server is used to execute the intelligent control method of the food warmer according to any one of claims 1 to 9.

Images