A catering management system and method for a smart park
By analyzing the identification bias and similarity of dishes, and dynamically adjusting the dish group update strategy, the problem of decreased identification reliability caused by dish updates in the catering management system was solved, thereby improving the system's identification accuracy and efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HANGZHOU GREEN OLIVE NETWORK TECH CO LTD
- Filing Date
- 2026-04-09
- Publication Date
- 2026-07-21
Smart Images
Figure CN122435593A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of image recognition technology, and in particular relates to a catering management system and method for smart parks. Background Technology
[0002] To improve the efficiency of restaurant settlement, existing technical solutions utilize restaurant images and weight recognition results for automatic settlement, significantly improving efficiency and reliability. Specifically, patent application CN202511556738.0, "An Automatic Settlement Method for Restaurant Consumption Based on Image Recognition," uses nutritional database matching and image feature analysis for nutritional component analysis and health assessment. Through price calculation, nutritional statistics, intelligent recommendation, and settlement processing, it achieves intelligent settlement, quickly completing dish recognition, nutritional analysis, and dynamic pricing, providing a complete technical solution for smart catering services. However, it suffers from the following technical problems: Catering establishments often need to regularly update and upgrade their menus. This inevitably leads to situations where images of existing dishes are too similar to current ones, impacting the reliability of the recognition system. Therefore, it's crucial to determine how to leverage the similarity and correlation of image features among existing dishes to manage future updates. This approach aims to avoid both the technical problem of excessive recognition error due to overly similar existing dishes and the technical problem of increasing the similarity of existing dishes, thus lowering the reliability of later updated dishes.
[0003] Therefore, there is an urgent need for a catering management system and method for smart parks. Summary of the Invention
[0004] To achieve the objectives of this invention, the following technical solution is adopted: Specifically, this application provides a catering management method for smart parks, which includes: S1 uses the image recognition results of the checkout device to determine the correlation between the recognition deviation of the meal and different meals. Based on the correlation between the recognition deviation of the meal and other meals and the recognition processing data of the meal, it determines the update processing method of the meal group. Using the update processing method of the meal group and combining the meal data in different meal groups, it determines the recognition processing strategy for updating the meal. S2 uses the identification and processing strategy to evaluate the similarity between the updated dish and the images of dishes in the dish group, and uses the similarity evaluation and analysis results to determine the associated dish group of the updated dish and identify the risk type. S3 determines the update management method for updated meals based on the degree of overlap between the associated meal groups of different updated meals and the risk type of the updated meals.
[0005] The beneficial effects of this invention are as follows: The identification risk type of the updated dish is determined by the similarity between the images of the updated dish and the dishes in the dish group. Based on the association between the updated dish and the existing dish group, i.e., the degree of similarity, especially the number of associated dish groups, the identification risk of the updated dish after its addition is classified to determine which risk type it belongs to. This provides a basis for subsequent identification and processing strategies and lays the foundation for further management and control of future dish updates.
[0006] Based on the degree of overlap between different updated menu items and their associated menu item groups, as well as the risk type of the updated menu items, the update management method for updated menu items is determined. When a large number of updated menu items of the same risk type appear, the number of existing groups (i.e., interfering menu item groups) that these high-risk menu items point to is analyzed to determine the concentration of the risk source. This allows for a dynamic determination of which update management method to adopt for future updated menu items, in order to control the system similarity risk that is further aggravated by the addition of updated menu items. This is to prevent the overall identification confusion risk from being difficult to manage, and also to prevent the technical problem of future updated menu items being difficult to update effectively in the system due to the excessively high confusion risk.
[0007] Furthermore, the correlation between the identification deviation of the dish and different dishes is determined based on the number of times the dish is identified as other dishes.
[0008] Furthermore, the method for determining the update processing method for the meal group is as follows: S11 determines the average daily number of times the meal is identified and processed based on the identification and processing data of the meal; S12 Based on the identification deviation correlation between the dish and other dishes, the number of times the dish is identified as other dishes and the number of times other dishes are identified as the dish are taken as the number of deviation correlations between the dish and other dishes. S13 determines the update processing method for the meal group by using the number of deviation correlations between the meal and other meals and the average daily number of identification processing for the meal.
[0009] Furthermore, the method for determining the risk type of the updated meal is as follows: Based on the associated food group data of the updated food, determine the number of associated food groups of the updated food; The risk type of the updated meal is determined by the number of associated meal groups.
[0010] Secondly, this application provides a catering management system for smart parks, employing the aforementioned catering management method for smart parks, specifically including: Strategy determination module, risk type determination module, update management module; The strategy determination module is responsible for determining the identification and processing strategy for updating meals by utilizing the update processing method of the meal group and combining the meal data in different meal groups. The risk type determination module is responsible for using the similarity assessment and analysis results to determine the associated food group of the updated food and identify the risk type; The update management module is responsible for determining the update management method for the updated menu items.
[0011] Other features and advantages will be set forth in the following description, and the objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.
[0012] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description
[0013] The above and other features and advantages of the present invention will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings.
[0014] Figure 1 This is a flowchart of a catering management method for smart parks; Figure 2 This is a flowchart illustrating the method for determining the update processing method for food item groups; Figure 3 This is a flowchart illustrating the method for determining the updated identification and processing strategy for meals; Figure 4 This is a framework diagram of a catering management system for smart parks. Detailed Implementation
[0015] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.
[0016] Example 1 like Figure 1As shown, this application provides a catering management method for smart parks, specifically including: S1 uses the image recognition results of the checkout device to determine the correlation between the recognition deviation of the meal and different meals. Based on the correlation between the recognition deviation of the meal and other meals and the recognition processing data of the meal, it determines the update processing method of the meal group. Using the update processing method of the meal group and combining the meal data in different meal groups, it determines the recognition processing strategy for updating the meal. S2 uses the identification and processing strategy to evaluate the similarity between the updated dish and the images of dishes in the dish group, and uses the similarity evaluation and analysis results to determine the associated dish group of the updated dish and identify the risk type. S3 determines the update management method for updated meals based on the degree of overlap between the associated meal groups of different updated meals and the risk type of the updated meals.
[0017] Furthermore, the image recognition result of the checkout device is determined based on the historical recognition data of the checkout device for different food items.
[0018] Furthermore, the correlation between the identification deviation of the dish and different dishes is determined based on the number of times the dish is identified as other dishes.
[0019] Specifically, such as Figure 2 As shown, the method for determining the update processing method of the meal group is as follows: Based on the recognition frequency of different dishes and their correlation with recognition deviations with other dishes, the update processing method for dish groups is dynamically determined to optimize the accuracy and efficiency of the image recognition system.
[0020] Its core decision-making logic is a hierarchical decision tree based on identification frequency and the strength of deviation association. This logic first performs a first-level triage based on the average daily number of times a dish is identified. High-frequency dishes are processed using the most lenient "basic update processing method," meaning that groups are merged as long as a deviation association exists, because high-frequency dishes have a higher risk of identification deviation. For low-frequency dishes, the number of associated dishes and the strength of the deviation association are further analyzed. If there are too many associated dishes, it indicates that the dish is easily confused with multiple dishes, and the basic update method is still required to improve the reliability of identification and handling of confusion risks. If the number of associated dishes is limited, strongly associated "deviation-related dishes" are selected, and their proportion among all associated dishes is evaluated. When the proportion of strongly associated dishes is high, the most refined third-level decision is entered, comprehensively considering identification frequency and the number of strong associations to calculate a "risk impact value," ultimately deciding whether to use the basic update method or a more stringent "other update processing method" (merging only strongly associated dishes). This logic aims to balance the sensitivity and accuracy of identification, avoiding system performance degradation due to excessive or insufficient merging.
[0021] S11 determines the average daily number of times the meal is identified and processed based on the identification and processing data of the meal; If the average daily number of identification processing times for the dish is greater than the preset identification processing time threshold in the above steps, then the update processing method for the dish group is determined to be the basic update processing method. That is, as long as there is a deviation correlation number between other dishes and any dish in the dish group to which the dish belongs, then the other dishes are assigned to the dish group of the dish.
[0022] Recognition and processing data: refers to the historical recognition records of food items in the image recognition system, including information such as the name of the food item recognized each time, the recognition result, and the recognition time.
[0023] Average daily number of recognition processing times: This refers to the average number of times a particular food item is recognized and processed by an image recognition system within a unit of time (usually one day). It reflects the frequency of appearance and popularity of this food item in the cafeteria or restaurant.
[0024] Preset identification processing threshold: A pre-defined number of times used to distinguish between high-frequency and low-frequency dishes. High-frequency dishes usually refer to popular dishes, such as scrambled eggs with tomatoes and braised pork.
[0025] Basic update processing method: A most lenient item grouping update rule, which means that as long as there is any identification discrepancy between two items (i.e., A is identified as B, or B is identified as A), they are classified into the same item group. This method has the lowest merging threshold and the fastest group expansion.
[0026] Deviation association count: This refers to the cumulative number of times two different dishes are incorrectly identified. For example, the number of times scrambled eggs with tomatoes is identified as scrambled eggs with tomatoes, plus the number of times scrambled eggs with tomatoes is identified as scrambled eggs with tomatoes, equals the deviation association count between the two.
[0027] High-frequency food items appear frequently in the identification system, resulting in a large amount of data. Therefore, the risk of identification confusion is relatively high. Thus, the most lenient approach is used to construct groups of easily confused food items, thereby ensuring the efficiency and comprehensiveness of the identification and handling of food item confusion risks.
[0028] It should also be noted that if the average daily number of identification processing times for the food item is not greater than the preset identification processing time threshold, then proceed to step S12.
[0029] S12 Based on the identification deviation correlation between the dish and other dishes, the number of times the dish is identified as other dishes and the number of times other dishes are identified as the dish are taken as the number of deviation correlations between the dish and other dishes. The above steps include the following: S121 Other dishes that have a number of deviation associations with the dish are considered as associated dishes. If the number of associated dishes is greater than a preset threshold for the number of associated dishes, the update processing method for the dish group is determined to be the basic update processing method. That is, as long as other dishes have a number of deviation associations with any dish in the dish group to which the dish belongs, the other dishes are assigned to the dish group.
[0030] S122 If the number of associated dishes is not greater than a preset threshold for the number of associated dishes, other dishes whose number of deviation associations with the dish is greater than a preset threshold for the number of associations are regarded as deviation associated dishes. When the proportion of the number of deviation associated dishes of the dish in the associated dishes does not meet the requirements, the update processing method of the dish group of the dish is determined to be the basic update processing method. That is, as long as other dishes have a number of deviation associations with any dish in the dish group to which the dish belongs, the other dishes are assigned to the dish group of the dish. If the number of associated deviation dishes of the dish meets the requirements, then proceed to step S13.
[0031] Related dishes: These refer to all other dishes that have had at least one identification discrepancy with the target dish (whether the target was identified as it, or it was identified as the target). This is the set of "suspects" of the target dish.
[0032] Preset threshold for the number of associated dishes: A pre-defined integer threshold used to determine whether the target dish is confused with too many other dishes. If there are too many associated dishes, it means that the dish is easily confused with many other dishes, possibly due to their similarity in appearance, name, or ingredients.
[0033] Deviation Association Frequency Threshold: A pre-set frequency value used to filter out dishes that have a "strong association" with the target dish. Only dishes with a deviation association frequency exceeding this threshold are considered truly easily confused "deviation-related dishes".
[0034] Deviation-related dishes: These are dishes that have a deviation-related number of times they are associated with the target dish greater than a preset threshold; they are strongly associated objects of confusion.
[0035] The proportion does not meet the requirements: This means that the number of dishes with biased associations accounts for a relatively high percentage of all associated dishes. This implies that although the target dish has been misidentified once or twice with many other dishes, the vast majority of misidentifications are common and form stable, strong associations.
[0036] For low-frequency dishes, further analysis of their confusion patterns is needed. If it has been misidentified with too many dishes (too many associated dishes), it indicates that its own characteristics are not obvious and it is easily confused with multiple dishes. In this case, a lenient basic update approach should still be used to cover all potential risks. If the number of associated dishes is limited, then repeatedly occurring strong associations (deviation-related dishes) should be selected. If the proportion of strong associations is high, it indicates that most misidentifications are common events, and the basic update approach can accommodate these errors and avoid omissions. Only when the proportion of strong associations is low is it necessary to further determine the confusion risk that the dish brings to the updated dish, thus requiring more refined decision-making.
[0037] S13 determines the update processing method for the meal group by using the number of deviation correlations between the meal and other meals and the average daily number of identification processing for the meal.
[0038] Furthermore, the risk impact value of the identification deviation of the meal is determined based on the average daily number of identification processing times of the meal and the number of meals associated with the deviation of the meal. It is then determined whether the risk impact value of the identification deviation of the meal is greater than a preset impact threshold. If so, the update processing method for the meal group of the meal is determined to be the basic update processing method, that is, as long as other meals have a number of deviation associations with any meal in the meal group to which the meal belongs, the other meals are assigned to the meal group of the meal. If not, the update processing method for the meal group of the meal is determined to be another update processing method, that is, as long as the number of deviation associations between other meals and any meal in the meal group to which the meal belongs is greater than a preset association number threshold, the other meals are assigned to the meal group of the meal.
[0039] Risk Impact Value of Misidentification: A comprehensive quantitative indicator used to assess the systemic risk or impact that may result from misidentification of a dish. It is typically determined by the average daily number of misidentifications of the dish (representing the breadth of impact) and the number of dishes associated with the misidentification (representing the concentration of confusion). For example, it can be defined as the ratio of the average daily number of misidentifications to the preset number of misidentifications, and the average percentage of the number of dishes associated with the misidentification among the associated dishes.
[0040] Preset impact threshold: A pre-set value used to determine whether the risk impact of the dish is high enough to require a more conservative management strategy.
[0041] Other update processing methods: A more stringent rule for updating food groupings is to only group two foods into the same food group if the number of deviation associations between them exceeds a preset association threshold. This method has a higher merging threshold and is more cautious in expanding groups.
[0042] This is the most refined layer of the decision-making logic. When a dish has a few strongly related objects and its daily recognition frequency is acceptable, its "risk" needs to be comprehensively assessed. If the risk impact value is high (for example, although the recognition frequency is high, the large number of strongly related objects means that the risk of confusion is high), the basic update processing method is still used, erring on the side of incorrect merging rather than letting it go unchecked. If the risk impact value is low, a more precise "other update processing method" can be used, merging only those strongly related dishes that are truly frequently confused, thereby reducing the difficulty of data analysis and processing of the group.
[0043] A large industrial park's cafeteria has introduced an intelligent image recognition payment system. To improve recognition accuracy, the system needs to regularly update and manage the food item groups. This article uses "Sweet and Sour Pork" as an example to demonstrate the process of determining its group update handling method.
[0044] Step 1: Frequency-based traffic splitting (S11): The system statistics show that the average daily number of times "sweet and sour pork" is identified and processed in the past month is 30.
[0045] 30 times < the preset threshold of 50 recognition processing times, therefore proceed to step S12 for further analysis.
[0046] Step 2: Analyze the correlation strength between associated dishes and deviations (S12) S121: The system query of historical identification data found that "Sweet and Sour Pork" has been misidentified with the following dishes: It was mistakenly identified as "Pineapple Sweet and Sour Pork" 4 times; It was mistakenly identified as "Yu Xiang Rou Si" (fish-flavored shredded pork) twice. It was misidentified as "shredded pork with sweet bean sauce" twice. One instance of misidentification with "Kung Pao Chicken"; It was misidentified as "braised pork ribs" once; Therefore, there are 5 related dishes for "Sweet and Sour Pork" (Pineapple Sweet and Sour Pork, Fish-Flavored Shredded Pork, Beijing-Style Shredded Pork, Kung Pao Chicken, and Braised Pork Ribs).
[0047] The number of associated dishes is 5, which is equal to the preset threshold of 5. Therefore, it is not greater than the threshold and the basic update processing method is not triggered. Then proceed to S122.
[0048] S122: Based on the preset association number threshold of 3 times, select dishes that are misidentified with "sweet and sour pork" more than 3 times, namely "pineapple sweet and sour pork" (4 times), as the biased association dishes.
[0049] The number of dishes associated with the deviation is 1, which accounts for 1 / 5 = 20% of all 5 associated dishes.
[0050] The percentage is no greater than the preset "proportion meets the requirements" standard of 20%, therefore the requirements are met, and proceed to step S13.
[0051] Step 3: Calculate the risk impact value and make a final decision (S13): Calculate the risk impact value of identification bias for "Sweet and Sour Pork": Risk impact value = (average number of daily identifications / basic number of identifications + proportion of deviation-related meals) / 2 = (30 / 100 + 0.2) / 2 = 0.25.
[0052] 0.25> The preset impact threshold of 0.2 indicates that although the number of dishes associated with the deviation is small, the overall risk is high due to the high number of daily identifications.
[0053] Therefore, a basic update processing method was adopted for "Sweet and Sour Pork": if any other dish has any deviation correlation with this dish or any dish in its group, then that other dish will be classified into the same dish group.
[0054] Based on the above decision, the system uses the base update method for updating "Sweet and Sour Pork". This means that in the future, when updating food groups, any food that has been misidentified once with "Sweet and Sour Pork" (such as Shredded Pork with Garlic Sauce, Shredded Pork with Peking Sauce, etc.), as well as any food that has been misidentified once with "Pineapple Sweet and Sour Pork" (which may be in the same group due to a confirmed strong association), will be included in the same large food group.
[0055] Specifically, such as Figure 3 As shown, the method for determining the updated meal identification and processing strategy is as follows: Based on the proportion of groups in the food category that use the basic update processing method, we dynamically determine which food image features should be compared and analyzed when food identification needs to be updated, so as to reasonably allocate computing resources while ensuring timely detection of potential confusion risks.
[0056] Its core decision-making logic is a tiered coverage logic based on the degree of risk exposure. The core judgment criterion for this logic is the proportion of groups using the basic update method. A high proportion of groups using the basic update method means that there are a large number of groups in the system that use lenient merging rules. These groups may contain many dissimilar items due to the low merging threshold, resulting in a higher risk of interference. Therefore, the most comprehensive strategy of "comparing with all items in all groups" is needed to thoroughly investigate the risks. Conversely, if the proportion of groups using the basic update method is not high, it means that most groups in the system use stricter merging rules, the relationships within the groups are clear, and the risk of interference is low. In this case, a more refined range selection can be made based on the group size and update method to reduce the computational load.
[0057] S21 uses the update processing method of the food group to determine the food group with the basic update processing method in the food group; The above steps include the following: S211 Obtain the number of the food group. If the number of the food group is less than the preset food group number threshold, then determine that the identification and processing strategy for the updated food is to compare and analyze the image features of all the food in all the food groups to determine whether there are related food groups. If the number of the food group is not less than the preset food group number threshold, then proceed to step S212. S212 Based on the food group data of the basic update processing method in the food group, determine the proportion of the food group in the basic update processing method and use it as the basic update proportion. Determine whether the basic update proportion is greater than the preset update proportion threshold. If so, determine that the identification and processing strategy of the updated food is to compare and analyze the image features of all food in all food groups to determine whether there are related food groups. If not, proceed to step S22.
[0058] Basic update processing method: This refers to the most lenient food grouping update rule, which assigns two foods to the same food group if there is any identification discrepancy between them. This method has a high group update frequency and can reflect the latest confusion relationships in a timely manner, but it may also be too sensitive and merge some foods with large differences in characteristics together, resulting in more interfering items and dissimilar foods within the group.
[0059] Preset threshold for the number of food groups: A pre-defined integer used to determine the overall size of food groups in the current system. When the number of groups is small, the computational cost of a comprehensive comparison is low, and this method can be used preferentially.
[0060] Basic Update Ratio: This refers to the proportion of all food groups that use the "basic update processing method." This ratio reflects the number of groups in the system that may have significant interference.
[0061] Preset update ratio threshold: A pre-defined ratio value used to determine if there are too many basic update method groups. If the basic update ratio is greater than this threshold, it indicates that there are too many basic update groups, and there may be a lot of interference within these groups. A global full comparison is needed to comprehensively investigate the risks. If the basic update ratio is less than or equal to the threshold, it indicates that there are not many basic update groups, and the interference risk is relatively controllable. The next step can be to conduct more refined analysis.
[0062] This step is the first level of traffic splitting, based on the total number of groups and the proportion of basic update groups. When the total number of groups is small, the computational load is not large regardless of the comparison method used, so the most comprehensive comparison is directly selected to ensure no omissions. When the number of groups is large, the proportion of basic update groups is further analyzed: if the basic update proportion is high, it indicates that there are a large number of groups in the system that may contain interference, and interference needs to be eliminated through a global full comparison; if the basic update proportion is not high, it indicates that the interference risk is controllable, and we can proceed to the next step to adopt differentiated comparison strategies for different groups.
[0063] Assuming a cafeteria's food identification system has 10 food groups (with a preset group size threshold of 15), then 10 < 15, so a full comparison strategy is directly adopted. If the number of groups is 20, exceeding the threshold, proceed to the next step. At this point, the basic update ratio is calculated. If 12 of the 20 groups are basic update method groups, the ratio is 60%, exceeding the preset update ratio threshold of 30%, indicating too many basic update groups and a high risk of interference; therefore, a global full comparison is adopted. If there are only 5 basic update groups, the ratio is 25% ≤ 30%, indicating few basic update groups and a controllable risk of interference; proceed to S22 for further analysis.
[0064] S22 determines the quantity of food items in the food item group based on the updated data of food items in different food item groups; In the above steps, the number of dishes in the dish group is obtained, and it is determined whether the number of dishes in the dish group is greater than a preset threshold for the number of dishes. If so, the identification and processing strategy for updating the dishes is determined to be to compare and analyze the image features of all dishes in the dish group to determine whether there is a related dish group. If not, proceed to step S23.
[0065] Number of dishes: refers to the total number of dishes included in a specific group of dishes.
[0066] Preset item quantity threshold: A pre-defined integer used to determine if a group is too large. If the number of items in a group exceeds this threshold, it indicates that the group is large, has complex internal relationships, and is highly likely to cause confusion. In this case, it is necessary to compare the group with all other items in the group to fully capture potential confusion relationships.
[0067] This step assesses the size of a single group. We proceed to this step when there are few basic update groups and the risk of interference is manageable, to evaluate the size of each group. Larger groups, regardless of their update methods, have a higher risk of internal confusion; therefore, a comprehensive comparison of all items within the group is necessary.
[0068] In systems where interference risks are manageable, further focus should be placed on large, high-risk groups, prioritizing the accuracy of identification in these areas.
[0069] Continuing with the previous example, after the system proceeds to S22, for each food group (e.g., the "Sweet and Sour Series Group"), the number of food items within it is counted as 12. If the preset threshold for the number of food items is 10, then 12 > 10. Therefore, for this group, the identification and processing strategy for new food items is determined to be: only compare and analyze the image features with all 12 food items in the "Sweet and Sour Series Group". If the number of food items in the group is 8, which is less than or equal to 10, then proceed to S23 for further analysis.
[0070] S23 determines the identification and processing strategy for the updated meal based on the basic update processing method of the meal group and the number of meals in different meal groups.
[0071] Specifically, the meal group that is compared and analyzed with the image features of all meals in the meal group is taken as a reliable comparison meal group. It is determined whether the proportion of the reliable comparison meal group in the meal group is greater than a preset reliable proportion threshold. If so, the identification and processing strategy for updating the meal is determined to be to compare and analyze the image features with the preset proportion of meals in the meal group to determine whether there is a related meal group. If not, proceed to the next step.
[0072] If the update processing method for the food group is determined to be a basic processing method, then the identification processing strategy for the updated food is to compare and analyze the image features of the food in the food group with a preset proportion of the food to determine whether it belongs to an associated food group. If it does, then the identification processing strategy for the updated food is to compare and analyze the image features of all the food in the food group to determine whether it belongs to an associated food group.
[0073] Reliable comparison food groups: These are groups that were determined in S22 to have a number of food items exceeding a preset threshold, thus requiring comparison with all food items within the group. These groups are considered key areas where risk has been fully covered.
[0074] Preset Reliability Ratio Threshold: A pre-defined ratio value used to determine whether the proportion of reliable comparison food groups in the total group is too high. If the proportion is high, it means that most groups have covered the risks through comprehensive comparison, and a more lenient strategy can be adopted for the remaining smaller groups.
[0075] Preset proportion of meals: refers to a portion of meals randomly selected from the group or selected according to certain rules for comparison. It is a sampling strategy that reduces the amount of computation while ensuring a certain coverage.
[0076] Basic processing method: This refers to the fact that the group itself was formed using the "basic update processing method", which means that the association between the dishes in the group is loose and there may be a lot of interference.
[0077] This is the finest layer in the decision-making logic, specifically handling groups that are neither large nor belong to the basic update groups requiring global full-scale updates. First, it assesses the prevalence of groups already subjected to full comparison in the system (reliable comparison groups). If the proportion of reliable comparison groups is high, it indicates that most risks have been covered, and even if the remaining groups are compared using a preset ratio, there will be no significant risk omissions. Therefore, a preset ratio comparison is uniformly applied to all remaining groups. If the proportion of reliable comparison groups is low, individual judgment is required for each remaining group: for groups using basic update methods, due to their lenient merging rules, there may be more interference; to ensure identification reliability, full comparison is still necessary. For groups using non-basic update methods, due to their strict merging rules, clear internal relationships, and low interference risk, a preset ratio comparison can be used to save resources.
[0078] Assume there are 20 food item groups in the system. Based on S22, 8 groups have a food item quantity exceeding a threshold, thus becoming "reliable comparison food item groups." The proportion of reliable comparison groups = 8 / 20 = 40%. If the preset reliable proportion threshold is 50%, then 40% < 50%, which does not meet the condition for uniformly using a proportion comparison. Therefore, individualized judgment needs to be performed for each unreliable comparison group (12 small and medium-sized groups). Taking a group with 8 food items as an example: if this group belongs to the basic processing method, then a full comparison within the group is used; if it belongs to a non-basic processing method, then a preset proportion comparison is used, such as 30%.
[0079] Furthermore, the associated meal group of the updated meal is a meal group that contains meals whose image features are similar to the updated meal to meet the requirements.
[0080] Furthermore, the method for determining the risk type of the updated meal is as follows: Based on the association between the updated menu items and existing menu item groups, especially the number of associated menu item groups, the risk of the updated menu items is classified to determine which risk type they belong to, thus providing a basis for subsequent identification and handling strategies.
[0081] Its core decision-making logic is a risk grading logic based on the breadth of association. The more image feature similarities an updated dish has with existing dish groups, the more easily it will be confused with multiple groups of dishes during the identification process, and the higher the integration difficulty and potential systemic risk. Therefore, by quantifying the number of associated dish groups and mapping them to preset risk level ranges, updated dishes can be divided into three risk types: high risk, medium risk, and low risk.
[0082] Based on the associated food group data of the updated food, determine the number of associated food groups of the updated food; Updated dishes: refers to dishes newly added to the identification system, or existing dishes that need to have their identification features updated.
[0083] Related food group: This refers to a group of food items that, in image feature comparison, have image features that are similar to the updated food item to a preset requirement. In other words, as long as at least one food item in an existing group is sufficiently similar to the updated food item in image features, that group is called a related food group of the updated food item.
[0084] Similarity meets the requirement: This means that the similarity score calculated by the image feature comparison algorithm exceeds a pre-set similarity threshold. This threshold determines the strictness of the association judgment.
[0085] Identifying related food groups is fundamental to risk stratification. Only by first identifying which existing groups are "related" to the updated food can the breadth of this relationship and its potential impact be further assessed.
[0086] A cafeteria is preparing to add a new dish, "Sweet and Sour Pork with Pineapple." The system compares its image features with those of dishes in all existing dish groups and finds that it has an 85% similarity to "Sweet and Sour Pork" in the "Sweet and Sour Series Group" and a 78% similarity to "Pineapple Fried Rice" in the "Fruit-Infused Dishes Group," both exceeding the preset similarity threshold of 70%. Therefore, both the "Sweet and Sour Series Group" and the "Fruit-Infused Dishes Group" are identified as related dish groups for "Sweet and Sour Pork with Pineapple."
[0087] The risk type of the updated meal is determined by the number of associated meal groups.
[0088] Specifically, the risk types for the updated meals include three types: Type 1, Type 2, and Type 3. Type 1 risk is more severe than Type 2 risk, and Type 2 risk is more severe than Type 3 risk.
[0089] Number of associated food groups: This refers to the total number of existing food groups that meet the similarity requirements and are associated with the updated food. This number reflects the "breadth of confusion" of the updated food in the existing system.
[0090] Risk type identification: This involves classifying the risks that updating a menu item may pose to the identification system. A higher risk type means the item is more likely to be confused with multiple existing groups, making integration more difficult and potentially requiring special handling of the identification strategy.
[0091] Risk Type 1: High Risk Level. This typically corresponds to a large number of associated food groups, indicating that the updated food is significantly similar to multiple existing groups, making cross-group misidentification highly likely.
[0092] Category II Risk: Medium Risk Level. This corresponds to a moderate number of associated food groups, indicating that the updated food items are similar to some existing groups, but the risk is manageable.
[0093] Three risk types: Low risk level. This corresponds to a small or even zero number of associated food groups, indicating that the updated food has distinct characteristics, is not easily confused with existing groups, and can be easily integrated into the system.
[0094] Quantity Range: The quantity range of the associated food group is divided into several consecutive intervals, each interval corresponding to a risk type. For example, a quantity ≥ 5 corresponds to risk type 1, 2-4 corresponds to risk type 2, and 0-1 corresponds to risk type 3.
[0095] The number of associated food groups is an intuitive and effective indicator of risk quantification. A larger number indicates more "common" characteristics of the updated food, blurring its boundaries with existing categories. This makes it more likely that the food will be incorrectly classified into multiple different groups during future identification, posing challenges to the system's stability and accuracy. By dividing the number into intervals, continuous numerical values can be transformed into discrete risk levels, facilitating standardized processing strategies in the future.
[0096] Specific example: The system has preset risk level classification rules: ≥4 associated food groups are classified as risk type 1, 2-3 as risk type 2, and 0-1 as risk type 3.
[0097] For "Pineapple Sweet and Sour Pork", the number of associated food groups is 2 (sweet and sour series group, fruit-infused dishes group), which falls into the range of 2-3, and therefore it is identified as a second-class risk type.
[0098] If another new dish, "Classic Braised Pork," is only associated with the "Braised Series Group," and the number of associated groups is 1, then it is identified as a type of risk.
[0099] If we assume that an "innovative fusion dish" is similar to five other groups, namely "sweet and sour series", "spicy series", "stir-fried series", and "cold dish series", and the number of associated groups is five, then it is identified as a type of risk that requires special attention and handling.
[0100] Furthermore, the risk type of the updated meal is determined based on the risk type corresponding to the quantity range of the associated meal group.
[0101] Specifically, the method for determining the update management method for the updated menu items is as follows: When a large number of new dishes of the same risk type appear, the degree of concentration of the risk source can be determined by analyzing the number of existing groups (i.e., interfering dish groups) that these high-risk dishes all point to. This allows for a dynamic determination of what update management methods to adopt for future new dishes in order to control the system similarity risk that is further aggravated by the addition of new dishes.
[0102] Its core decision-making logic is a tiered control logic based on the concentration of risk sources. This logic first assesses the similarity of the existing system by analyzing the number of risk types in historically updated menu items—if there are too many risk types, it indicates that the existing systems are already highly similar. At this point, further analysis of the existing groups associated with these risk types is needed: if they are concentrated in a few groups (few interfering food groups), it indicates that the risk sources are concentrated, and a refined control strategy can be adopted; if they are widely concentrated in a large number of groups (many interfering food groups), it indicates that the risk sources are dispersed, and the entire system faces impact, requiring the strictest control strategy. Through this tiered approach, it can precisely protect key groups with concentrated risks while also implementing comprehensive restrictions when risks spread widely.
[0103] S41 determines the proportion of updated meals with different identified risk types based on the identified risk type of the updated meal; Furthermore, the above steps include the following: S411 Based on the proportion of different risk types of updated meals in the updated meals, determine a type of risk type of updated meals, and determine whether there is a type of risk type of updated meals. If yes, proceed to step S412. If no, it means that the risk of confusion between meals is small at this time. Therefore, it is determined that the update management method for future updated meals is uncontrolled, that is, no special restrictions are imposed on new meals. S412 determines whether the number of updated meals of the first type of risk is greater than the preset updated meal quantity threshold. If so, the update management method for future updated meals is determined to be the preset management method. That is, if the number of associated meal groups of the updated meal is greater than the preset associated group quantity threshold, the updated meal cannot be updated, thereby avoiding the occurrence of excessive overall interference risk, which in turn leads to excessive risk of meal identification deviation and excessive interference with subsequent updated meals. If not, proceed to step S413. S413 Based on the quantity of the updated dishes and the proportion of updated dishes of different risk types, determine the interference risk coefficient of the dishes, and determine whether the interference risk coefficient of the dishes is greater than the preset interference risk coefficient threshold. If yes, proceed to step S42. If no, determine that the update management method for future updated dishes is other management methods. That is, if the number of associated dish groups of the updated dishes is greater than the preset associated group number threshold, and the number of dishes in any associated dish group is greater than the preset number of dishes, then the updated dishes cannot be updated. Update dishes: Adding new dishes to the recognition system may make existing dishes that are already similar even more similar.
[0104] Risk type identification: Risk levels are categorized based on the number of associations between the updated food and existing groups. A risk type indicates that there are similarities in characteristics with multiple existing groups.
[0105] One type of risky updated menu items: The large number of such menu items in history directly reflects the high similarity of the existing menu system itself. It is precisely because existing menu items are easily confused that a large number of updated menu items are also likely to be similar to multiple groups.
[0106] Preset update item quantity threshold: Used to determine if the quantity of a certain type of risky item has increased sufficiently to prove that the existing system is too similar. If the threshold is exceeded, it indicates that the existing system is in a high-risk state, and the strictest measures must be taken to prevent further deterioration.
[0107] Preset management method: The strictest update control rules, that is, for any updated food item, if the number of its associated food item groups exceeds the preset threshold for the number of associated food item groups, the update of that food item is prohibited.
[0108] Preset threshold for the number of associated groups: Used to measure the breadth of confusion that a single updated dish may cause.
[0109] Interference Risk Coefficient: A comprehensive quantitative indicator used to assess the overall risk level of this batch of new food items.
[0110] Preset interference risk coefficient threshold: Used to determine whether the overall risk is high enough to require further analysis of the risk sources at the group level.
[0111] Other management methods: A moderately strict update control rule, which prohibits updates if the number of associated food groups exceeds a preset threshold for the number of associated groups, and at least one of the associated groups has an existing number of food items exceeding a preset value.
[0112] This step starts by analyzing the risk distribution of new dishes and then infers the similarity of the existing system. If there are no risky dishes, the existing system is acceptable and requires no control. If there is one risky dish, the quantity is considered: too many dishes indicate that the existing systems are already highly similar, requiring the strictest pre-set management methods; a small quantity but a high interference risk coefficient requires further analysis of the risk source.
[0113] Assume there are 10 updated menu items in the past, of which 2 are of the risky type. The preset threshold for the number of updated menu items is 3. Since 2 < 3, the preset management method is not used. The interference risk coefficient is calculated as the number of updated menu items in the past (10) divided by the preset number of menu items (10), and then multiplied by the proportion of the risky type of menu items (0.2), which equals 0.2. This is greater than the preset threshold of 0.1, so proceed to step S42 for further analysis.
[0114] S42 determines that a food group belongs to an updated food group based on the degree of overlap between the associated food groups of different updated food items; The above steps include the following: The updated meals belonging to the associated meal groups are considered as interfering meals. It is determined whether there are meal groups with a number of interfering meals greater than a preset threshold for the number of interfering meals. If yes, proceed to step S43. If no, it is determined that the update management method for the updated meals is another management method. That is, if the number of associated meal groups of the updated meal is greater than a preset threshold for the number of associated groups, and the number of meals in any associated meal group is greater than a preset value for the number of meals, the updated meal cannot be updated. The degree of overlap between related food groups: This refers to whether different updated food items have historically been associated with the same existing groups. The reason why a large number of updated food items have historically become a category of risky food items is precisely because they all point to those groups—these groups themselves are the source of risk.
[0115] Disruptive Items: For a specific group, all updated items that are associated with that group are considered "disruptive items" for that group, meaning they become disruptive items after being interfered with. The common target of these items makes that group a potential "disruptive item group".
[0116] Preset threshold for the number of interfering dishes: Used to determine whether a group is considered an "interfering dish group," meaning a group whose dishes have been excessively associated with each other in historical updates. Exceeding the threshold indicates that the group is a key source causing a large number of risky dishes to appear.
[0117] Interfering food groups: These refer to existing groups that have been associated with multiple updated food items (exceeding a preset threshold for the number of interfering food items) in the past. These groups are the root cause of new food items becoming a type of risk—it is the existence of these groups that makes updated food items prone to being similar to multiple groups.
[0118] This step aims to identify the "source clusters" that cause a large number of risky food items. If such interfering food clusters exist, it indicates that the risk sources are concentrated, and further analysis of the number of these sources is needed. If no interfering food clusters exist, it means that the risk of the new food item is not caused by a few groups, but rather by scattered and sporadic occurrences. In this case, other moderately stringent management methods can be used for control.
[0119] Continuing the example above, after entering S42, the number of updated dishes associated with each existing dish group is counted. It is found that group X is associated with 5 updated dishes, group Y with 4, and all other groups with no more than 1. If the preset threshold for the number of interfering dishes is 2, then groups X and Y both have more than 2 interfering dishes, thus becoming interfering dish groups, and proceed to S43. If the number of interfering dishes in all groups is ≤2, then other management methods are determined. If the number of associated dish groups for future updated dishes is greater than 2, and the number of dishes in any one of the associated dish groups is greater than 10, then update control processing is required.
[0120] S43 uses the proportion of updated meals with different risk identification types and the number of updated meals belonging to related meal groups to determine the update management method for the updated meals.
[0121] Furthermore, groups of dishes with a number of interfering dishes exceeding a preset threshold for the number of interfering dishes are designated as groups of dishes with interference risk. It is determined whether the number of these groups exceeds a preset threshold for the number of risky groups. If so, the update management method for the updated dishes is defined as a preset management method. This means that if the number of associated groups of the updated dish exceeds a preset threshold for the number of associated groups, the updated dish cannot be updated, thus avoiding an excessively high overall interference risk, which could lead to excessively high identification errors and excessive interference with subsequent updated dishes. If not, the update management method is defined as a strict management method. This means that if the number of associated groups of the updated dish exceeds a preset threshold for the number of associated groups, and the number of dishes in any associated group exceeds a preset value for the number of dishes, or if there is an interfering risky group among the associated groups, the updated dish cannot be updated.
[0122] Interference risk food group: refers to existing groups that are associated with multiple updated food items (exceeding the preset threshold for the number of interference food items), which are the sources of risk that cause a large number of a certain type of risk food item to appear.
[0123] Preset risk group quantity threshold: Used to determine if the number of food groups causing interference risks is excessive. If the number exceeds this threshold, it indicates that the risk sources are widely dispersed, and many groups are the root cause of the problem. In this case, the strictest preset management method must be used for broad restriction. If the number is small, it indicates that the risk sources are concentrated, and a more refined and strict management method can be used.
[0124] Strict Management Approach: A refined update control rule stipulates that for any updated menu item, if the number of associated menu item groups exceeds a preset threshold, and (either at least one associated group has an existing menu item count exceeding the threshold, or at least one associated group is itself a high-risk menu item group), then the update is prohibited. This rule is suitable for situations where the sources of risk are concentrated, and can precisely protect the key groups that cause problems.
[0125] The steps involve a final assessment of the number of food groups at risk of interference. If there are too many source groups, the problem is widespread and requires comprehensive restrictions using pre-defined management methods. If the number of source groups is limited, the risk is concentrated in a few groups, allowing for stricter management methods to precisely restrict food items associated with those source groups, while allowing other food items to be updated normally. This approach controls risk while avoiding excessive restrictions.
[0126] Continuing the example, there are two groups of food items at risk of interference (groups X and Y). The preset threshold for the number of risk groups is 3; since 2 < 3, a strict management method is adopted. The rule is: if the number of associated groups for a food item is greater than 3, and (there exists an associated group with more than 10 existing food items or associated with group X or Y), then updating is prohibited. In this way, food items associated with group X or Y, even if their association count is small, may be prohibited, thus protecting these two source groups from further interference.
[0127] Example 2 Secondly, such as Figure 4 As shown, this application provides a catering management system for smart parks, employing the aforementioned catering management method for smart parks, specifically including: Strategy determination module, risk type determination module, update management module; The strategy determination module is responsible for determining the identification and processing strategy for updating meals by utilizing the update processing method of the meal group and combining the meal data in different meal groups. The risk type determination module is responsible for using the similarity assessment and analysis results to determine the associated food group of the updated food and identify the risk type; The update management module is responsible for determining the update management method for the updated menu items.
[0128] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.
[0129] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.
[0130] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.
Claims
1. A catering management method for smart parks, characterized in that, Specifically, it includes: Using the image recognition results of the checkout device, the correlation between the recognition deviations of different dishes is determined. Based on the correlation between the recognition deviations of the dishes and other dishes, as well as the recognition processing data of the dishes, the update processing method for the dish groups is determined. Using the update processing method for the dish groups, and in combination with the dish data in different dish groups, the recognition processing strategy for updating the dishes is determined. The similarity between the updated dish and the dishes in the dish group is evaluated using the identification and processing strategy. The analysis results of the similarity evaluation are used to determine the associated dish group of the updated dish and identify the risk type. The method for managing the update of meals is determined based on the degree of overlap between the associated meal groups of different updated meals and the risk type of the updated meals.
2. The catering management method for smart parks as described in claim 1, characterized in that, The correlation between the identification deviation of a dish and other dishes is determined based on the number of times the dish is identified as other dishes.
3. The catering management method for smart parks as described in claim 1, characterized in that, The method for determining the update processing method of the meal group is as follows: Based on the identification and processing data of the food items, determine the average daily number of identification and processing times for the food items; Based on the identification deviation correlation between the dish and other dishes, the number of times the dish is identified as other dishes and the number of times other dishes are identified as the dish are taken as the deviation correlation number between the dish and other dishes. The update processing method for the meal group is determined by using the number of deviation correlations between the meal and other meals and the average number of daily identification and processing times of the meal.
4. The catering management method for smart parks as described in claim 3, characterized in that, If the average daily number of identification processing times for the food item exceeds a preset threshold for the number of identification processing times, then the update processing method for the food item group is determined to be the basic update processing method.
5. The catering management method for smart parks as described in claim 1, characterized in that, The associated meal group of the updated meal is a group of meals that have image features that meet the requirements of similarity to the updated meal.
6. The catering management method for smart parks as described in claim 1, characterized in that, The method for determining the risk type of the updated meal items is as follows: Based on the associated food group data of the updated food, determine the number of associated food groups of the updated food; The risk type of the updated meal is determined by the number of associated meal groups.
7. The catering management method for smart parks as described in claim 6, characterized in that, The risk types identified in the updated meals include three types: Type 1 risk, Type 2 risk, and Type 3 risk. Type 1 risk is greater than Type 2 risk, and Type 2 risk is greater than Type 3 risk.
8. The catering management method for smart parks as described in claim 7, characterized in that, The risk type of the updated food item is determined based on the risk type corresponding to the quantity range of the associated food item group.
9. The catering management method for smart parks as described in claim 1, characterized in that, The method for determining the update management method for the updated menu items is as follows: Based on the identified risk type of the updated meal, determine the proportion of updated meals with different identified risk types in the updated meal; Based on the degree of overlap between the related food groups of different updated food items, determine which food group belongs to the updated food items of the related food group; The update management method for the updated meals is determined by using the proportion of updated meals with different risk identification types and the number of updated meals belonging to related meal groups.
10. A catering management system for smart parks, employing the catering management method for smart parks as described in any one of claims 1-9, characterized in that, Specifically, it includes: Strategy determination module, risk type determination module, update management module; The strategy determination module is responsible for determining the identification and processing strategy for updating meals by utilizing the update processing method of the meal group and combining the meal data in different meal groups. The risk type determination module is responsible for using the similarity assessment and analysis results to determine the associated food group of the updated food and identify the risk type; The update management module is responsible for determining the update management method for the updated menu items.