School canteen purchase cost actuarial system based on image recognition

Abstract

The invention provides a school canteen procurement cost actuarial system based on image recognition, and relates to the technical field of image recognition, and the system comprises a time-sharing image group recognition module which is used for determining the image collection frequency according to the dining busy degree, and carrying out the recognition of the characteristics of a crowd, and obtaining different people groups. And the group consumption actuarial module is used for establishing a consumption prediction model, inputting a personnel dining image for prediction to obtain personnel consumption predicted amount, and calculating predicted consumption cost in combination with the purchase cost parameter. And the consumption difference correlation analysis module is used for analyzing the comparison consumption by using a consumption difference multi-dimensional attribution method to obtain a difference reason and finding out a correlation difference relationship. And the cost optimization decision module is used for making a purchase decision according to the optimized consumption cost scheme. According to the method, the accuracy of crowd identification and the accuracy of food material consumption prediction are improved, the association relationship between the consumption and the difference reason is clearly compared, a targeted basis is provided for optimizing the purchase cost, and the cost accounting precision is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image recognition, and in particular to a school canteen procurement cost accounting system based on image recognition. BACKGROUND

[0002] Traditional school canteen procurement management has long been plagued by three major technical bottlenecks: first, crowd image acquisition and analysis efficiency is low, relying on fixed frequency acquisition or manual statistics, which cannot dynamically adapt to the flow changes of the canteen peak / flat period, resulting in missing or redundant image data of key scenes (such as the early morning rush), and fixed acquisition frequency is easy to cause waste of storage resources and feature recognition errors. Second, the crowd classification accuracy is insufficient, it is difficult to integrate multi-dimensional information such as facial features and behavior patterns, resulting in the recognition accuracy of sub-types such as "students- faculty" and "different grade groups", which cannot accurately capture the group consumption differences.

[0003] In the field of food consumption analysis, the crowd characteristics and real-time scene data are not deeply combined, resulting in high error rate of consumption prediction, especially in special scenes such as exam week and rainstorm, the error is even greater. In addition, the consumption deviation attribution lacks a systematic method, when the predicted value and the actual consumption differ, it is difficult to quantify the difference influencing factors from multiple dimensions such as dish categories, time periods, seasons, and it is difficult to identify the interactive effects of "out of stock + seasonal changes", resulting in insufficient accuracy of deviation cause identification, and unable to form an effective cost optimization closed loop.

[0004] With the advancement of smart canteen construction, the application of image recognition and machine learning technology provides a possibility to break through the above bottlenecks, but the existing solutions still have technical gaps: on the one hand, the image acquisition frequency and the adaptation mechanism of crowd dynamic characteristics are missing, which cannot optimize data efficiency while ensuring recognition accuracy; on the other hand, the consumption prediction model and the deviation attribution method fail to fully integrate multi-modal image features and statistical analysis methods, resulting in insufficient scientificity and real-time of cost accounting.

[0005] Therefore, it is urgent to build a school canteen procurement cost accounting system based on image recognition to realize accurate control of canteen procurement cost. SUMMARY

[0006] The present application provides a school canteen procurement cost accounting system based on image recognition to solve the defects of low efficiency in collecting crowd images, inaccurate classification, inaccurate food consumption analysis, and difficulty in finding the reasons for the difference between the predicted value and the actual value in the prior art.

[0007] The application provides a school canteen procurement cost accounting system based on image recognition, comprising: a time-sharing image recognition module, which is used for determining an image acquisition frequency according to the dining busy degree of the canteen at different time periods, and obtaining personnel dining images and consumption images by image acquisition of the canteen area to identify the characteristics of different personnel groups.

[0008] A group consumption accounting module is used for analyzing the consumption rules of different personnel groups, establishing a consumption prediction model, inputting the personnel dining images to obtain personnel consumption prediction values, and calculating the predicted consumption cost in combination with the procurement cost parameters.

[0009] A consumption difference correlation analysis module is used for comparing the personnel consumption prediction values with the actual consumption values to obtain comparative consumption values, using a consumption difference multi-dimensional attribution method to analyze the difference reasons, and finding the correlation difference relationship between the comparative consumption values and the difference reasons by using a consumption difference correlation method.

[0010] A cost optimization decision module is used for updating the predicted consumption cost according to the correlation difference relationship to obtain an optimized consumption cost scheme, and formulating procurement decisions according to the optimized consumption cost scheme.

[0011] The application provides a school canteen procurement cost accounting system based on image recognition, comprising: defining a dining busy degree index according to the dining population density, the queue length and different dining time distribution, and dividing the canteen into multiple areas according to the function and time period to obtain corresponding initial sampling frequencies.

[0012] The dining population density, the queue length and the dining time are monitored in real time, and whether the current dining busy degree index reaches a preset threshold value is evaluated according to a preset time interval, and if yes, the initial sampling frequency of the corresponding area is adjusted to obtain an image sampling frequency.

[0013] The application provides a school canteen procurement cost accounting system based on image recognition, comprising: analyzing the personnel dining images and consumption images according to a target detection algorithm to identify the dining images of the personnel area in the images and the facial features.

[0014] The clothing, shape and behavior of the personnel are extracted from the dining images as personnel features, and the payment method, consumption amount and consumption frequency of the personnel are identified as consumption behavior features according to the facial features.

[0015] The comprehensive feature vectors are constructed in combination with the personnel features, the facial features and the consumption behavior features, and the common features and behavior patterns are obtained by clustering analysis according to the consumption behavior and the dining time.

[0016] The personnel in the images are divided into different personnel groups according to the common features and the behavior patterns.

[0017] The application provides a school canteen procurement cost accounting system based on image recognition, comprising:

[0018] Dining data of different personnel groups in the canteen is collected, outliers and missing values are removed, and key features affecting food consumption are extracted therefrom.

[0019] Statistical indicators of different personnel groups at different time periods and for different dishes are calculated based on the dining data for analysis to obtain food consumption.

[0020] The correlation between the key features and the food consumption is analyzed, food consumption features that affect the food consumption to a preset range are found, and personnel groups are clustered to obtain food consumption rules.

[0021] An LSTM model is selected according to the food consumption rules and target requirements.

[0022] The personnel groups and the key features are taken as feature inputs, and the actual food consumption is taken as an output, and a prediction error index is calculated according to the mean absolute error.

[0023] The LSTM model is parameter-adjusted according to the prediction error index to obtain a consumption prediction model.

[0024] The application provides a school canteen procurement cost accounting system based on image recognition, comprising: image size adjustment, normalization and grayscale operation are performed on personnel dining images to meet the input requirements of the consumption prediction model, and the corresponding personnel consumption prediction amount is predicted and output.

[0025] The purchase cost parameters of each food material are obtained by combining the current market situation and the actual operation situation of the canteen.

[0026] The cost of each food material is calculated according to the personnel consumption prediction amount and the purchase cost parameters to obtain the predicted consumption cost.

[0027] The application provides a school canteen procurement cost accounting system based on image recognition, comprising:

[0028] The personnel consumption prediction amount and the actual consumption amount are aligned in granularity, and the influence of school holidays and temporary activity dates is eliminated, and the consumption caliber is unified and the dish mapping is calibrated.

[0029] The difference rate is calculated according to the dish category and the price interval, and the deviation category is identified in combination with the taste preference, the portion standard and the dish sales change.

[0030] According to the comparison of the personnel consumption prediction amount and the actual consumption amount of the morning, noon and evening meals, the fluctuation period is identified, and the consumption deviation trend within a week is combined with the people flow image of the period to analyze the consistency with the activity rules within a week to obtain a period analysis result.

[0031] The consumption deviation factor is obtained by comparing the current quarter consumption deviation with the historical same period, and the seasonal analysis result is obtained by combining the weather data analysis extreme climate on the consumption deviation factor.

[0032] The difference reason is obtained according to the structure change of the personnel group and the influence of external events on the deviation category, period analysis combination and seasonal analysis result.

[0033] The present application provides a school canteen procurement cost actuarial system based on image recognition, comprising:

[0034] The comparative consumption and the difference reason are aligned according to the time dimension to form a deviation time-influence factor record library, each row represents an analysis unit, and each column represents a difference reason.

[0035] According to the record library, the statistical indicators of the comparative consumption corresponding to each difference reason are calculated, and a scatter plot between each difference reason is drawn, and the trend and relationship are analyzed to obtain deviation causal data.

[0036] The deviation causal data and each difference reason are analyzed using variance analysis method to obtain associated variables, and the associated difference relationship reaching the preset strength with the comparative consumption is found out according to the relationship direction and strength.

[0037] The present application provides a school canteen procurement cost actuarial system based on image recognition, comprising:

[0038] The original hypothesis is established, the between-group sum of squares and the within-group sum of squares of each difference reason are calculated according to the deviation time-influence factor record library, and the mean square is calculated to obtain test data.

[0039] Under the original hypothesis, the significance level is determined according to the test data, the corresponding critical value is found out by combining the F distribution table, and each difference reason is judged to obtain the corresponding P value.

[0040] If the test data is less than the critical value, the original hypothesis is accepted, and it is considered that the difference reason has no significant influence on the comparative consumption.

[0041] It is judged whether the P value is greater than the significance level, if yes, it is considered that there is a significant association between the current difference reason and the comparative consumption, and the associated variable is output, otherwise the next difference reason is judged.

[0042] The present application provides a school canteen procurement cost actuarial system based on image recognition, comprising:

[0043] The associated difference relationship is input into the consumption prediction model as a new variable for training, the optimized predicted consumption is output again, and the predicted consumption cost is recalculated by combining the procurement cost parameter to obtain the current consumption cost.

[0044] According to the current consumption cost, the food material purchase quantity is adjusted, and the menu structure and dishes are updated in combination with the difference reasons, the food material classification consumption data is obtained by grading management according to the food material consumption fluctuation, and the optimized consumption cost scheme is obtained in combination with the supplier cooperation purchase information.

[0045] The application provides a school cafeteria purchase cost actuarial system based on image recognition, comprising:

[0046] According to the optimized predicted consumption quantity, the required food materials in different time periods and seasons are arranged to obtain the current food material requirement, and the safe inventory quantity of each type of food material is determined according to the food material classification consumption data.

[0047] According to the current food material requirement and the safe inventory quantity, the purchase quantity of each type of food material in a preset time period is calculated, the best purchase time is determined in combination with the food material consumption speed and the purchase cycle, and the supplier is selected according to the supply stability, the food material quality and the price rationality.

[0048] The purchase decision is made in combination with the purchase quantity, the best purchase time and the determined supplier.

[0049] The school cafeteria purchase cost actuarial system based on image recognition provided by the application reduces manual intervention by dynamically adjusting the image acquisition frequency, automatically identifying dishes and consumption images in combination with a target detection algorithm, integrates consumption images, purchase cost parameters and external environment data, establishes full-link data association, realizes cross-department data sharing and dynamic adjustment through a deviation time-influence factor record library, finds out the difference reasons between the actual consumption quantity and the predicted quantity, and clearly compares the association relationship between the consumption quantity and the difference reasons, thereby providing a targeted basis for optimizing the purchase cost, improving the cost accounting precision, updating the predicted consumption cost according to the associated difference relationship, making an optimized consumption cost scheme, determining the best purchase time and the purchase quantity, selecting a suitable supplier, and improving the purchase efficiency and the cost benefit. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can also be obtained by those skilled in the art without any creative effort.

[0051] Figure 1 is a process schematic diagram of the school cafeteria purchase cost actuarial system based on image recognition provided by the embodiments of the application. DETAILED DESCRIPTION

[0052] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be described clearly and completely below in conjunction with the drawings in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the protection scope of the present application.

[0053] As shown in Figure 1 The school canteen procurement cost actuarial system based on image recognition provided by the embodiment of the present application comprises:

[0054] The time-sharing image recognition module is used for determining the image acquisition frequency according to the dining busy degree of the canteen at different time periods, and obtaining the personnel dining image and consumption image by image acquisition of the canteen area, and obtaining different personnel groups by crowd feature recognition. The step of obtaining the personnel dining image and consumption image by image acquisition of the canteen area can be to divide the canteen into multiple areas according to function (such as entrance, meal selling area, dining area) and time period (morning, noon and evening) and give an initial sampling frequency (such as 10 fps), then monitor the number density, queue length and dining time of each area in real time, evaluate whether the current busy degree reaches the threshold (such as the queue length exceeds 5 meters) according to the preset time interval, if it reaches, adjust the initial sampling frequency of the corresponding area (such as increase to 30 fps), finally through the camera deployed in each area of the canteen, collect the image containing the personnel dining scene, facial features, clothing, body shape, behavior and consumption process (such as payment method, dish selection) according to the adjusted sampling frequency, provide data support for subsequent crowd feature recognition and consumption behavior analysis.

[0055] The step of determining the image acquisition frequency comprises:

[0056] The dining busy degree index is defined according to the number density of diners, queue length and different dining time distribution, and the canteen is divided into multiple areas according to function and time period to obtain the initial sampling frequency of the corresponding area.

[0057] The number density of diners, queue length and dining time are monitored in real time, and whether the current dining busy degree index reaches the preset threshold is evaluated according to the preset time interval, and if it does, the initial sampling frequency of the corresponding area is adjusted to obtain the image sampling frequency.

[0058] The step of obtaining different personnel groups comprises:

[0059] According to the target detection algorithm, the personnel dining image and consumption image are analyzed, and the dining image and face features of the personnel region in the image are recognized. First, the collected image is preprocessed by using a target detection algorithm such as YOLOv8 or FasterR-CNN: the image size is adjusted (such as unified to 640x480 pixels), normalized (pixel value is reduced to [0, 1]), and grayed to reduce the amount of calculation, and at the same time, Gaussian filtering is used to remove noise. Then, the personnel region is detected: each person in the image is pixel-level segmented to generate a mask containing head, torso, hand and other parts, and the dining behavior region (such as holding a tray, sitting, eating and other actions) is accurately located; the region of interest (ROI) is set for the food selling window, payment terminal and other key areas, and the personnel face, hand (such as card swiping, code scanning payment) and food contact behavior are detected first to improve the recognition efficiency of the consumption image.

[0060] The clothing, body shape and behavior of the personnel are extracted from the dining image as personnel features, and the payment method, consumption amount and consumption frequency of the personnel are recognized as consumption behavior features according to the face features.

[0061] By associating the face features with big data, the payment method (such as campus card payment, face payment, code scanning payment) of the personnel can be obtained through the existing canteen consumption system.

[0062] In the canteen scene, there are situations of "similar clothing (such as uniform school uniforms) and similar behavior (such as queuing for food)", which are difficult to accurately distinguish individuals only by clothing and behavior. Through the face features, the specific personnel can be accurately located.

[0063] From the clothing, the school uniforms are distinguished from the formal or casual clothes worn by the faculty, and the food consumption rules of people with different clothing are different. Classification from the behavior features can include the crowd of fast food, packed and taken away, and the crowd of dining and multi-person dining. The crowd of fast food, packed and taken away prefers easy-to-carry dishes such as rice balls and braised dishes; during the examination period, the proportion of this "fast food" crowd increases, and the main food procurement quantity needs to be increased to meet the demand. The other type is the crowd of dining and multi-person dining, who prefer hot dishes and soups, and have more demand for food variety, with an average of 4-5 kinds of dishes per table, and increased consumption of soups.

[0064] Most of the people using face payment are high-frequency diners with stable consumption frequency. Most of the people using code scanning payment are temporary diners with unstable consumption frequency, and they prefer special dishes such as new dish sets, with large consumption fluctuations; when the proportion of code scanning payment users exceeds 20%, 10% of special dish materials need to be additionally reserved to avoid out-of-stock.

[0065] The comprehensive feature vector is constructed by combining the personnel features, face features and consumption behavior features, and the common features and behavior patterns are obtained by clustering analysis according to the consumption behavior and dining time.

[0066] According to the common characteristics and behavior patterns, the personnel in the image are divided into different personnel groups.

[0067] The group consumption actuarial module is used to analyze the food consumption rules of different personnel groups, establish a consumption prediction model, input the personnel dining image to obtain the personnel consumption prediction, and calculate the predicted consumption cost combined with the procurement cost parameter.

[0068] The steps of establishing the consumption prediction model include:

[0069] Collect dining data of different personnel groups in the canteen, remove outliers and missing values, and extract key features affecting food consumption from them. Dining data can include the number of diners, dish selection, consumption amount, dining time, etc. Key features can include dining time, dish type, season, etc. For seasonal features, a year can be divided into four categories: spring, summer, autumn and winter.

[0070] According to the dining data, the statistical indicators of different personnel groups in different time periods and different dishes are calculated and analyzed to obtain the food consumption.

[0071] Analyze the correlation between key features and food consumption, find out the food consumption features that affect food consumption to a preset range, and perform cluster analysis on personnel groups to obtain food consumption rules.

[0072] According to the food consumption rules and target requirements, select an LSTM model, which is expressed by the formula:

[0073]

[0074]

[0075] In the formula, is the hidden state, is the memory cell, is the weight matrix, is the hidden state at the previous moment, is the cell state at the previous moment, is the bias term, is the prediction output at the moment.

[0076] Take the personnel group and key features as feature input, and the actual food consumption as output. According to the mean absolute error, the prediction error index is calculated, which is expressed by the formula:

[0077]

[0078] In the formula, is the prediction error index, is the total number of samples, is the true value of the first sample, is the predicted value of the first sample, is the true value of the first sample. is the predicted value of the first sample.

[0079] According to the prediction error index, the LSTM model is adjusted to obtain a consumption prediction model.

[0080] The step of calculating the predicted consumption cost includes:

[0081] The personnel dining image is subjected to image size adjustment, normalization and grayscale operation to meet the input requirements of the consumption prediction model, and the predicted output is subjected to personnel consumption prediction.

[0082] The purchase cost parameters of each food material are obtained by combining the current market situation and the actual operation of the canteen.

[0083] The cost of each food material is calculated according to the personnel consumption prediction and the purchase cost parameters to obtain the predicted consumption cost, which is expressed as:

[0084]

[0085] In the formula, is the predicted consumption cost, is the consumption prediction, is the purchase cost parameter, is the loss rate.

[0086] The consumption difference correlation analysis module is used to compare the personnel consumption prediction with the actual consumption to obtain a comparison consumption, and the consumption difference multi-dimensional attribution method is used to analyze the difference reasons and find the correlation difference relationship between the comparison consumption and the difference reasons through the consumption difference correlation method.

[0087] The step of analyzing the difference reasons includes:

[0088] The personnel consumption prediction and the actual consumption are aligned in granularity, and the influence of school holidays and temporary activity dates is eliminated, and the consumption caliber is unified and the dish mapping is calibrated.

[0089] Consumption caliber unification: clearly define the "consumption" (such as actual consumption = purchase quantity + initial inventory - final inventory - loss quantity), ensure that the calculation logic of the prediction model and the actual statistics are consistent, and avoid comparison failure due to definition deviation.

[0090] Dish mapping calibration: through the "dish standardization dictionary", the predicted "stewed pork" is unified with the actual statistical "stewed five-flavor pork" and other aliases, and the name ambiguity (such as "tomato fried eggs" in a canteen may be recorded as "tomato fried eggs") is eliminated.

[0091] Calculate the difference rate of dishes by category and price range, and identify the deviation category by combining taste preference, portion standard, and dish sales changes.

[0092] According to the comparison of personnel consumption prediction and actual consumption in early, mid, and late meals, identify the fluctuation period, and combine the personnel flow image statistics of the consumption deviation trend within a week to analyze the consistency with the weekly activity rule to obtain the period analysis result.

[0093] Compare the current quarter consumption deviation with the historical same period to obtain the consumption deviation factor, and analyze the extreme climate impact on the consumption deviation factor to obtain the seasonal analysis result.

[0094] According to the influence of personnel group structure changes and external events on the deviation category, period analysis combination, and seasonal analysis result, obtain the difference reason. The structure change of the personnel group can be a sudden increase in the proportion of junior high school students, whose meat preference is lower than that of high school students, leading to a decrease in pork consumption, or a dish taste problem (such as too salty braised pork leading to actual sales lower than prediction). External events can be temporary school notice for physical examination and cafeteria launch of special price package, etc. The difference reason can be inconsistent statistical caliber, food material unit conversion error, image recognition error, supply side problem, demand side change, processing and service link loss, etc.

[0095] The steps to obtain the associated difference relationship include:

[0096] Align the comparison consumption and difference reasons according to the time dimension to form the deviation time-influence factor record library, each row representing an analysis unit, and each column representing a difference reason.

[0097] Calculate the statistical indicators of the comparison consumption corresponding to each difference reason according to the record library, and draw a scatter plot between each difference reason to analyze the trend and relationship to obtain the deviation cause data.

[0098] Use variance analysis method to analyze the deviation cause data and each difference reason to obtain the associated variable.

[0099] The steps to analyze the associated variable include:

[0100] Establish the null hypothesis, which usually assumes that each difference reason has no significant impact on the comparison consumption, i.e. the mean of each group data is equal. According to the deviation time-influence factor record library, calculate the inter-group sum of squares and intra-group sum of squares of each difference reason, and calculate the mean square to obtain the test data, which is expressed as:

[0101]

[0102] In the formula, is the inter-group square, is the number of groups, is the sum of squares within groups, is the total number of all samples, is the test data.

[0103] The formula for calculating the sum of squares between groups and the sum of squares within groups is:

[0104]

[0105] In the formula, is the number of samples in the th group, is the sample mean of the th group, is the overall mean of all samples, is the th sample value of the th group.

[0106] Under the null hypothesis, the significance level is determined according to the test data, the corresponding critical value is found in the F distribution table, and each difference reason is judged to obtain the corresponding P value. The judgment steps include: if the test data is greater than the critical value, the null hypothesis is rejected, and it is considered that the difference reason has a significant impact on the comparison consumption. Output as P value.

[0107] If the test data is less than the critical value, the null hypothesis is accepted, and it is considered that the difference reason has no significant impact on the comparison consumption.

[0108] Determine whether the P value is greater than the significance level. If yes, it is considered that there is a significant correlation between the current difference reason and the comparison consumption, and output as the correlation variable. Otherwise, continue to judge the next difference reason.

[0109] And according to the relationship direction and strength, find the associated difference relationship that reaches the preset strength with the comparison consumption.

[0110] The cost optimization decision module is used to update the predicted consumption cost according to the associated difference relationship to obtain an optimized consumption cost scheme, and to make procurement decisions according to the optimized consumption cost scheme.

[0111] The steps to update the optimized consumption cost scheme include:

[0112] The associated difference relationship is input as a new variable into the consumption prediction model for training, and the optimized predicted consumption is output again. The predicted consumption cost is recalculated by combining the procurement cost parameters to obtain the current consumption cost.

[0113] According to the current consumption cost, the food material purchase quantity is adjusted, and the menu structure and dishes are updated in combination with the difference reasons, the food material classification consumption data is obtained by grading management according to the food material consumption volatility, and the optimized consumption cost scheme is obtained in combination with the supplier cooperation purchase information.

[0114] The steps of making the purchase decision include:

[0115] According to the optimized predicted consumption quantity, the current food material requirement quantity required in different time periods and seasons is arranged, and the safe inventory quantity of each type of food material is determined according to the food material classification consumption data. The optimized predicted consumption quantity is split according to different time periods such as breakfast, lunch, dinner and night snack, in combination with historical same period seasonal data (such as 15% increase of food material consumption of cold dishes in summer, 20% increase of food material consumption of hot soup in winter), to generate a food material demand matrix including time period and seasonal dimensions. At the same time, based on the food material consumption volatility (such as 25% standard deviation of weekly consumption of leafy vegetables, only 8% of root vegetables), the food material is divided into three categories of high volatility, medium volatility and low volatility, and the safe inventory quantity is set respectively - the safe inventory of high volatility food material is set as 20% of the predicted consumption quantity, the safe inventory of medium volatility food material is set as 10%, and the safe inventory of low volatility food material is set as 5%, to cope with unexpected demand or delayed supply.

[0116] According to the current food material requirement quantity and the safe inventory quantity, the purchase quantity of each food material in the preset time period is calculated, the best purchase time is determined in combination with the food material consumption speed and the purchase cycle, and the supplier is selected according to the supply stability, the food material quality and the price rationality. The purchase quantity calculation formula is: purchase quantity = current food material requirement quantity + safe inventory quantity - existing inventory quantity - in-transit inventory quantity. For the consumption speed, the daily average consumption quantity of each food material is calculated through historical data (such as daily average consumption of pork 50 kg), in combination with the purchase cycle (leafy vegetables are purchased daily, frozen products are purchased weekly), the economic order quantity (EOQ) model is used to optimize the purchase batch quantity, and excessive inventory is avoided. The determination of the best purchase time needs to consider the shelf life of the food material (such as shelf life of fresh milk 7 days, which needs to be consumed within 5 days after arrival) and the transportation time (1 day for local suppliers, 3 days for out-of-town suppliers), for example, for food material with a shelf life of 10 days, the purchase time is calculated by "consumption speed x (purchase cycle + 3 days of shelf life safety days)". For the selection of suppliers, a three-dimensional evaluation system is established: supply stability (on-time delivery rate ≥ 95% in the past 3 months), food material quality (100% of agricultural residue detection qualified rate), and price rationality (average price not higher than market wholesale price by 5%), the weights (stability 40%, quality 35%, price 25%) are assigned by the analytic hierarchy process (AHP), the comprehensive score of the supplier is calculated and sorted.

[0117] Make procurement decisions based on procurement quantity, optimal procurement time, and identified suppliers. Generate detailed procurement orders with clear food material name, specification, procurement quantity, unit price, and delivery time; develop time-based procurement plans (e.g., breakfast food materials arrive by 6 a.m. daily, lunch food materials arrive by 10 a.m.); establish an emergency contact mechanism for suppliers, automatically triggering the procurement process for the backup supplier (the supplier ranked second in the score) when the main supplier fails to deliver on time. At the same time, synchronize procurement decisions to the inventory management system, update inventory status in real time, and notify relevant personnel through SMS or APP to execute procurement, ensuring that food supply and consumption prediction are highly matched, reducing the risk of stockout and inventory costs.

[0118] The school cafeteria procurement cost accounting system based on image recognition provided by the present example dynamically adjusts the image acquisition frequency according to the dining busy degree of the cafeteria at different times, achieving efficient and accurate image acquisition. A consumption prediction model is established, combined with population feature analysis, effectively improving the accuracy of food consumption prediction. Combined with procurement cost parameters and predicted consumption, cost calculation is performed, making cost estimation more close to actual demand. Through accurate prediction of consumption and optimization of cost calculation, food waste and excessive procurement are reduced, effectively reducing the operating cost of the cafeteria. By analyzing the food consumption patterns of different populations, the menu design is optimized to better meet the taste preferences of teachers and students, improving the quality of cafeteria service and satisfaction. Integrating image recognition, data analysis, and prediction models, the intelligent and automated management of cafeteria procurement costs is achieved, improving the informatization level and management level of the cafeteria.

[0119] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be implemented by means of software plus the necessary general hardware platform, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in terms of contribution to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0120] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An image recognition-based school cafeteria procurement cost accounting system, characterized by, The method comprises the following steps: The time-sharing image recognition module is used to determine the image acquisition frequency according to the dining busy degree of the canteen at different time periods, and to acquire personnel dining images and consumption images by image acquisition of the canteen area to recognize the characteristics of different personnel groups; The group consumption accounting module is used to analyze the food consumption rules of different personnel groups, establish a consumption prediction model, input the personnel dining images to obtain personnel consumption prediction, and calculate the predicted consumption cost by combining the procurement cost parameters; The consumption difference correlation analysis module is used to compare the personnel consumption prediction with the actual consumption to obtain a comparison consumption, use the consumption difference multi-dimensional attribution method to analyze the difference reasons, and find the correlation difference relationship between the comparison consumption and the difference reasons by the consumption difference correlation method; The cost optimization decision module is used to update the predicted consumption cost according to the correlation difference relationship to obtain an optimized consumption cost scheme, and make procurement decisions according to the optimized consumption cost scheme.

2. The image recognition-based school cafeteria procurement cost accounting system according to claim 1, characterized by, The step of determining the image acquisition frequency comprises: According to the dining number density, the queue length and the different dining time distribution, define the dining busy degree index, and divide the canteen into multiple areas according to the function and time period to obtain the corresponding initial sampling frequency; Real-time monitoring of the dining number density, the queue length and the dining time is carried out, and whether the current dining busy degree index reaches the preset threshold is evaluated according to the preset time interval. If yes, the initial sampling frequency of the corresponding area is adjusted to obtain the image sampling frequency.

3. The image recognition-based school cafeteria procurement cost accounting system according to claim 1, characterized by, The step of obtaining different personnel groups comprises: According to the target detection algorithm, the personnel dining images and the consumption images are analyzed to recognize the dining images of the personnel area in the image, and the facial features; From the dining images, the clothing, body shape and behavior of the personnel are extracted as personnel features, and the payment method, consumption amount and consumption frequency of the personnel are recognized as consumption behavior features according to the facial features; The comprehensive feature vector is constructed by combining the personnel features, the facial features and the consumption behavior features, and the common features and behavior patterns are obtained by clustering analysis according to the consumption behavior and the dining time; According to the common features and the behavior patterns, the personnel in the image are divided into different personnel groups.

4. The image recognition-based school cafeteria procurement cost accounting system according to claim 1, characterized by, The step of establishing the consumption prediction model comprises: Collecting dining data of different personnel groups in the canteen, removing outliers and missing values, and extracting key features affecting food consumption from the data; According to the dining data, statistical indicators of different personnel groups at different time periods and different dishes are calculated to analyze the food consumption; The correlation between the key features and the food consumption is analyzed to find out the food consumption features that affect the food consumption within a preset range, and the personnel groups are clustered to obtain the food consumption rules; According to the food consumption rules and target requirements, an LSTM model is selected; The personnel groups and the key features are used as feature inputs, and the actual food consumption is used as output. The prediction error index is calculated according to the mean absolute error; According to the prediction error index, the parameters of the LSTM model are adjusted to obtain the consumption prediction model.

5. The image recognition-based school cafeteria procurement cost accounting system according to claim 1, characterized by, The step of calculating the predicted consumption cost comprises: The personnel dining image is subjected to image size adjustment, normalization and grayscale operation to meet the input requirements of the consumption quantity prediction model, and the personnel consumption prediction quantity corresponding to the prediction output is obtained; The purchase cost parameters are obtained by acquiring the purchase price and loss rate of each food material and combining the current market situation and the actual operation situation of the canteen; The cost of each food material is calculated according to the personnel consumption prediction quantity and the purchase cost parameters to obtain the predicted consumption cost.

6. The image recognition-based school cafeteria procurement cost accounting system according to claim 1, characterized by, The step of analyzing the difference causes comprises: The personnel consumption prediction quantity and the actual consumption quantity are aligned in granularity, the influence of school holidays and temporary activity dates is eliminated, the consumption caliber is unified and the dish mapping is calibrated; The difference rate is calculated according to dish categories and price intervals, and the deviation categories are identified by combining taste preferences, portion standards and dish sales changes; According to the personnel consumption prediction quantity and the actual consumption quantity of the early, midday and evening meals, the fluctuation period is identified, the consumption deviation trend within a week is calculated by combining the people flow image of the period, and the consistency with the activity law within a week is analyzed to obtain the period analysis result; The consumption deviation factors are obtained by comparing and analyzing the current quarter consumption deviation with the historical same period, and the seasonal analysis result is obtained by combining the weather data to analyze the extreme climate on the consumption deviation factors; The difference causes are obtained according to the influence of the structure change of the personnel group and external events on the deviation categories, the period analysis combination and the seasonal analysis result.

7. The image recognition-based school cafeteria procurement cost accounting system according to claim 6, characterized by, The step of obtaining the associated difference relationship comprises: The comparison consumption quantity and the difference causes are aligned according to the time dimension to form a deviation time-impact factor record library, each row represents an analysis unit, and each column represents a difference cause; Statistical indicators of the comparison consumption quantity corresponding to each difference cause are calculated according to the record library, and a scatter plot between each difference cause is drawn to analyze the trend and relationship to obtain deviation causal data; The associated variables are obtained by using the analysis of variance method to analyze the deviation causal data and each difference cause, and the associated difference relationship reaching a preset strength with the comparison consumption quantity is found out according to the relationship direction and strength.

8. The image recognition-based school cafeteria procurement cost accounting system according to claim 7, characterized by, The step of analyzing the associated variables comprises: An original hypothesis is established, the inter-group sum of squares and the intra-group sum of squares of each difference cause are calculated according to the deviation time-impact factor record library, and the mean square is calculated to obtain test data; Under the original hypothesis, the significance level is determined according to the test data, the corresponding critical value is found out by combining the F distribution table, and each difference cause is judged to obtain the corresponding P value; It is judged whether the P value is greater than the significance level, yes, it is considered that the current difference cause and the comparison consumption quantity have significant association and is output as the associated variable, otherwise the next difference cause is judged.

9. The image recognition based school cafeteria procurement cost accounting system according to claim 1, characterized in that, The step of updating the optimized consumption cost scheme comprises: The associated difference relationship is input into the consumption quantity prediction model as a new variable for training, the optimized prediction consumption quantity is re-output, and the current consumption cost is obtained by recalculating the predicted consumption cost in combination with the purchase cost parameters. According to the current consumption cost, the food material purchase quantity is adjusted, and the menu structure and dishes are updated in combination with the difference reason, food material classification consumption data is obtained by grading management according to food material consumption fluctuation, and the optimized consumption cost scheme is obtained in combination with the supplier's purchase information.

10. The image recognition based school cafeteria procurement cost actuarial system according to claim 1, wherein, The step of making the purchase decision comprises: According to the optimized predicted consumption quantity, the current food material requirement is obtained by arranging food materials required in different time periods and seasons, and the safety stock quantity of each type of food material is determined according to the food material classification consumption data; According to the current food material requirement and the safety stock quantity, the purchase quantity of each type of food material in a preset time period is calculated, the best purchase time is determined in combination with the food material consumption speed and the purchase cycle, and the supplier is selected according to the supply stability, the food material quality and the price rationality; The purchase decision is made in combination with the purchase quantity, the best purchase time and the determined supplier.

Images