Pet fresh food customized formula generation method based on atlas and multi-modal recognition

Abstract

The invention discloses a method for generating a pet fresh food customized formula based on atlas and multi-modal recognition, and relates to the technical field of pet fresh food, and the method comprises the steps: collecting food material information of different fresh food, extracting food material parts, constructing a knowledge base, constructing a unified node type for the knowledge base and a knowledge atlas, and generating a fresh food formula. The method has the advantages that an exclusive scheme can be generated according to the differentiated requirements of calcium supplement of puppies, phosphorus control of old dogs and the like, meanwhile, the content of nutrients such as protein and vitamins is adjusted by means of the index of the deviation degree of the single nutrition values, and the nutritional value deviation degree of the puppies and the old dogs is optimized. According to the method, the formula is ensured to meet the actual requirements of pets, the health risk caused by nutrient imbalance is reduced, when food materials or pet types are added subsequently, services can be expanded only by supplementing data, reusable and easily-expanded data assets are formed, long-term technical iteration is supported, and double improvement of precise customization and convenient experience is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pet fresh food, in particular to a pet fresh food customized formula generation method based on atlas and multi-modal recognition. BACKGROUND

[0002] With the improvement of people's living standards, the status of pets in the family is becoming more and more important, and pet owners are paying more and more attention to pet food, hoping to provide pets with healthier and more targeted fresh food. In the field of pet fresh food, different pets have many differences such as breed, age, weight, health status, and the demand for nutrition is different. At the same time, the types of fresh food ingredients are rich, and different ingredient parts, processing methods, cooking methods and cooking times will affect the nutritional ingredients, taste, etc. of the ingredients; The primary defect in the prior art in the field of pet fresh food customization is the lack of personalized matching capability. In the face of different needs such as calcium supplementation for puppies and phosphorus control for senior dogs, the prior art cannot achieve precise adaptation, often resulting in one-size-fits-all formula recommendations, which leads to problems such as excess or insufficient nutrition for some pets. Secondly, the lack of nutrition precision control is a key shortcoming, and the core goal of controllable nutrition cannot be achieved. Thirdly, the data scalability and assetization degree are low, and it is difficult to form reusable and easily expandable data assets like this solution. Therefore, we propose a pet fresh food customized formula generation method based on atlas and multi-modal recognition. SUMMARY

[0003] The purpose of the present application is to provide a pet fresh food customized formula generation method based on atlas and multi-modal recognition.

[0004] To achieve the above purpose, the present application provides the following technical scheme: a pet fresh food customized formula generation method based on atlas and multi-modal recognition, comprising the following steps: S100, collecting ingredient information of different fresh foods, extracting ingredient parts, and analyzing processing methods, cooking methods and cooking times of different ingredient parts; S200, constructing a knowledge base, arranging ingredient information into the knowledge base, constructing a knowledge graph in the knowledge base according to the ingredient information, and constructing a unified node type for the knowledge base and the knowledge graph; S300, collecting the weight change of the ingredient after cooking, analyzing the content of different nutrients in the ingredient after cooking, calculating the nutrition aggregation value of the fresh food formula according to the above, and recording it in the knowledge base; S400, analyzing the single nutrition value in the fresh food formula, converting it into a standard nutrition value, generating an optimized formula according to the standard nutrition value, analyzing the single nutrition value in the optimized formula, calculating the single nutrition value deviation of the optimized formula, and recording the optimized formula and the corresponding single nutrition value in the knowledge base; S500, establishing a target search engine, establishing an individual feature vector for the target search engine, retrieving a fresh food recipe from the knowledge base according to the individual feature vector, and obtaining the fresh food recipe.

[0005] As a further scheme of the present application: in the S100, a quadruple is constructed, wherein the format of the quadruple is food material-treatment method-cooking method-cooking parameter, and the cooking parameter includes cooking temperature and cooking time.

[0006] As a further scheme of the present application: in the S100, the mass and volume of the food material are unified to 100g of edible portion; The moisture basis is unified, the dry basis and the wet basis are unified as the wet basis, and the dry food and moisture description are unified as the water content conversion.

[0007] As a further scheme of the present application: in the S200, the node types include food material, part, nutrient, cooking method, functional label, and food material variety; And when the knowledge graph is constructed, the food material is regarded as the main framework, and the food material part is regarded as the first branch framework of the main framework; The cooking method, functional label and food material variety of the food material part are taken as the second branch framework; Finally, the nutrients possessed by the food material after being cooked by the corresponding cooking method are taken as the third branch framework, wherein the third branch framework is the branch framework of the cooking method.

[0008] As a further scheme of the present application: in the S300, after the different nutrient contents in the cooked food material are obtained, the ratio between the different nutrient contents and the mass of the food material can be calculated to obtain the proportion of different kinds of nutrients The calculation formula of the proportion of different kinds of nutrients is as follows: ; Wherein is the mass of the food material, is the content of different kinds of nutrients.

[0009] As a further scheme of the present application: in the S300, when the nutrient aggregation value is calculated, the total value of different kinds of nutrients is calculated first: ; Wherein is the mass value of different food material varieties in the fresh food recipe, and S is the number of food material varieties; After the total value of different kinds of nutrients is calculated, it is added to obtain the nutrient aggregation value.

[0010] As a further scheme of the present application: the single-nutrient value deviation degree of the optimized recipe The calculation formula is as follows: ; These are standard nutrient values ​​for different types of nutrients. To optimize the nutrient values ​​of different types of nutrients in the formula.

[0011] As a further aspect of the present invention: In S500, after the individual feature vector is constructed, the individual feature vector is used to generate vector labels, and the vector labels are placed in the first frame of the corresponding ingredient part.

[0012] As a further aspect of the present invention: the individual feature vector includes pet breed, pet gender, whether it is spayed / neutered, the time of spaying / neutering, pet age, pet weight, and types of allergies. Fresh food recipes are retrieved based on the individual feature vector selected by the user. If no suitable fresh food recipe is found, a unique recipe is generated based on the individual feature vector and displayed.

[0013] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows: 1. This invention uses individual feature vectors as the core to deeply link pet profiles and personalized information, allowing fresh food recipes to break free from the limitations of general templates. By accurately matching knowledge base resources through a target search engine, it can generate exclusive solutions for differentiated needs such as calcium supplementation for puppies and phosphorus control for senior dogs. At the same time, by using single nutrient value deviation indicators, it can adjust the content of nutrients such as protein and vitamins to ensure that the recipe meets the actual needs of pets, reduce the health risks caused by nutritional imbalances, and standardize the data structure by unifying node types. When adding new ingredients or pet types in the future, only data needs to be added to expand the service, forming a reusable and easily expandable data asset to support long-term technology iteration, enhance users' trust in the safety of fresh food, and achieve a dual improvement in precise customization and convenient experience. 2. This invention achieves full-process traceability of food processing through a quadruple approach, ensuring unified measurement and eliminating formula errors caused by measurement differences. It also unifies moisture standards to enable horizontal comparability of nutrient content. The hierarchical framework design enables precise positioning and rapid querying, reducing knowledge retrieval costs. Unified node types ensure the scalability of the knowledge graph and adapt to industry knowledge iterations. Vector tags are associated with the first framework to achieve precise matching of individual food needs. Standardized vector tags enable quantifiable individual needs, providing a basis for subsequent formula optimization. 3、The pet food customization formula generation method based on the atlas and multi-modal recognition can fully consider the individual differences of different pets, meet the personalized nutritional needs of the pets, make the pets obtain the diet more suitable for their own conditions, obtain the nutritional aggregation value through relevant calculation, accurately and systematically quantitatively analyze the nutritional components of the pet food, facilitate the control of the nutritional balance degree, and can clearly optimize the difference between each nutrient in the formula and the standard nutritional value by using the single-nutrient value deviation formula, so that a scientific basis is provided for adjusting and optimizing the pet food formula, and the development of food more suitable for the nutritional needs of pets is facilitated. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 The method steps in the embodiments of the present application are schematically shown. DETAILED DESCRIPTION

[0015] The specific embodiments of the present application will be further described below in conjunction with the accompanying drawings, and it should be noted that the description of these embodiments is used to help understand the present application, but does not constitute a limitation on the present application.

[0016] In addition, the technical features involved in each of the embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0017] Embodiment one: The pet fresh food customization formula generation method based on the atlas and multi-modal recognition can fully consider the individual differences of different pets, meet the personalized nutritional needs of the pets, make the pets obtain the diet more suitable for their own conditions, obtain the nutritional aggregation value through relevant calculation, accurately and systematically quantitatively analyze the nutritional components of the pet food, facilitate the control of the nutritional balance degree, and can clearly optimize the difference between each nutrient in the formula and the standard nutritional value by using the single-nutrient value deviation formula, so that a scientific basis is provided for adjusting and optimizing the pet food formula, and the development of food more suitable for the nutritional needs of pets is facilitated; Therefore, in order to effectively solve the above problems, the present application provides a pet fresh food customization formula generation method based on the atlas and multi-modal recognition, as shown in the accompanying drawings of the specification Figure 1 The method comprises the following steps: S100, collecting information of different fresh food materials, extracting material parts, and analyzing processing methods, cooking methods and cooking times of different material parts; S200, constructing a knowledge base, arranging the material information in the knowledge base, constructing a knowledge graph in the knowledge base according to the material information, and constructing a unified node type for the knowledge base and the knowledge graph; S300, collect the weight change of the food material after cooking, analyze the different nutrient content in the food material after cooking, calculate the nutrition aggregation value of the fresh food formula according to the same, and record it in the knowledge base; S400, analyze the single nutrient value in the fresh food formula, convert it into a standard nutrient value, generate an optimized formula according to the standard nutrient value, analyze the single nutrient value in the optimized formula, calculate the single nutrient value deviation of the optimized formula, and record the optimized formula and the corresponding single nutrient value in the knowledge base; S500, establish a target search engine, establish an individual feature vector for the target search engine, retrieve the formula from the knowledge base according to the individual feature vector, and obtain the fresh food formula; Specifically, and also extract rules from authoritative standards, set the lower limit and upper limit of essential nutrients and the proportion relationship; An illegal list is established in the optimized formula, when the single nutrient value deviation is greater than X, the corresponding optimized formula is regarded as an illegal formula, which is then included in the illegal list, and the user has the right to edit the specific value of X, X is a natural number; Collect food material information of different fresh foods, extract food material parts, analyze different food material parts processing methods, cooking methods and cooking times, build a knowledge base, arrange food material information in the knowledge base, build a knowledge graph according to the food material information in the knowledge base, build a unified node type for the knowledge base and the knowledge graph, collect the weight change of the food material after cooking, analyze the different nutrient content in the food material after cooking, calculate the nutrition aggregation value of the fresh food formula according to the same, and record it in the knowledge base, analyze the single nutrient value in the fresh food formula, convert it into a standard nutrient value, generate an optimized formula according to the standard nutrient value, analyze the single nutrient value in the optimized formula, calculate the single nutrient value deviation of the optimized formula, and record the optimized formula and the corresponding single nutrient value in the knowledge base, establish a target search engine, establish an individual feature vector for the target search engine, retrieve the formula from the knowledge base according to the individual feature vector, and obtain the fresh food formula.

[0018] Example two: In S100, a quadruple is constructed, and the format of the quadruple is food material-processing method-cooking method-cooking parameter, and the cooking parameter includes cooking temperature and cooking time; In S100, the food material quality and volume are unified to 100g edible part; Unify the moisture reference, unify the dry basis and wet basis to wet basis, and unify the dry food and moisture description to water content conversion; In S200, the node types include food material, part, nutrient, cooking method, functional label and food material variety; When building the knowledge graph, the food material is regarded as the main framework, and the food material part is regarded as the first branch framework of the main framework; The cooking method of the food material part, the function label, and the food material variety are taken as the second branch framework; Finally, the nutrients possessed by the food material after being cooked by the corresponding cooking method are taken as the third branch framework, wherein the third branch framework is the branch framework of the cooking method; In S300, after the different nutrient contents in the cooked food material are obtained, the ratio between the different nutrient contents and the food material quality can be calculated to obtain the proportion of different kinds of nutrients The calculation formula of the proportion of different kinds of nutrients is as follows: ; wherein is the food material quality, is the content of different kinds of nutrients; Specifically, a sector diagram is drawn, and the sector diagram is perfected according to the content of different nutrients in different food materials to obtain a nutrient radar chart. Meanwhile, the user can also record the state change of the pet, and can also feed back the state change of the pet to the recipe generation method. Example four-tuple: Salmon, cut, steam, small fire for 20 minutes, wherein the small fire can also be expressed as 80℃-90℃; The four-tuple is constructed, wherein the format of the four-tuple is food material-processing method-cooking method-cooking parameter, the cooking parameter includes cooking temperature and cooking time, the food material quality and volume are unified to every 100g edible part, the water content is unified as a reference, the dry basis and wet basis are unified as a wet basis, the dry food and water content description are unified as water content conversion, the node types include food material, part, nutrient, cooking method, function label, and food material variety, and when the knowledge graph is constructed, the food material is taken as the main framework, the food material part is taken as the first branch framework of the main framework, the cooking method, function label, and food material variety of the food material part are taken as the second branch framework, and finally the nutrients possessed by the food material after being cooked by the corresponding cooking method are taken as the third branch framework, wherein the third branch framework is the branch framework of the cooking method. After the different nutrient contents in the cooked food material are obtained, the ratio between the different nutrient contents and the food material quality can be calculated to obtain the proportion of different kinds of nutrients .

[0019] Embodiment three: In S300, when the nutrient aggregation value is calculated, the total value of different kinds of nutrients is first calculated : ; wherein is the quality value of different food material varieties in the fresh food recipe, and S is the number of food material varieties; ​After the total value of different kinds of nutrients is calculated, the total value is added to obtain the nutritional aggregation value; Optimized formula single-nutrient value deviation The calculation formula is as follows: ; The standard nutritional value of different nutrients, The nutritional value of different nutrients in the optimized formula; In S500, after the individual characteristic vector is constructed, the individual characteristic vector generates a vector label, and the vector label is arranged in the first frame of the corresponding food material part; The individual characteristic vector includes pet breed, pet gender, whether spayed, spaying time, pet age, pet weight, and allergy type. Fresh food recipes are retrieved according to the individual characteristic vector selected by the user. When no suitable fresh food recipe is retrieved, a unique recipe is generated according to the individual characteristic vector, and the unique recipe is displayed; Specifically, the taste experience and cost weight are updated slightly based on the context strategy gradient, and the upper limit of the food material that is frequently rejected is lowered; When calculating the nutritional aggregation value, the total value of different kinds of nutrients is calculated first After the total value of different kinds of nutrients is calculated, the total value is added to obtain the nutritional aggregation value, after the individual characteristic vector is constructed, the individual characteristic vector generates a vector label, and the vector label is arranged in the first frame of the corresponding food material part, the individual characteristic vector includes pet breed, pet gender, whether spayed, spaying time, pet age, pet weight, and allergy type. Fresh food recipes are retrieved according to the individual characteristic vector selected by the user. When no suitable fresh food recipe is retrieved, a unique recipe is generated according to the individual characteristic vector, and the unique recipe is displayed.

[0020] The working steps of a pet fresh food customized formula generation method based on atlas and multi-modal recognition are as follows: Step one, collect food material information of different fresh foods, extract food material parts, analyze processing methods, cooking methods, and cooking times of different food material parts, and construct four-tuples, wherein the format of the four-tuples is food material-processing method-cooking method-cooking parameter, and the cooking parameter includes cooking temperature and cooking time; Unify the food material quality and volume to 100g of edible part, unify the moisture basis, unify the dry basis and wet basis to wet basis, and unify the dry food and moisture description to water content conversion; Step two, construct a knowledge base, arrange the food material information in the knowledge base, construct a knowledge graph in the knowledge base according to the food material information, and construct a unified node type for the knowledge base and the knowledge graph; The node types include food materials, parts, nutrients, cooking methods, function labels, and food material varieties, and when the knowledge graph is constructed, the food materials are regarded as a main framework, and the food material parts are regarded as a first branch framework of the main framework; Step three, the cooking methods, function labels, and food material varieties of the food material parts are taken as a second branch framework; Finally, the nutrients of the food materials after being cooked by the corresponding cooking methods are taken as a third branch framework, and the third branch framework is a branch framework of the cooking method; Step four, the weight change of the food materials after cooking is collected, the different nutrient contents in the cooked food materials are analyzed, and the nutrient aggregation value of the fresh food formula is calculated accordingly; After the different nutrient contents in the cooked food materials are obtained, the ratio between the different nutrient contents and the food material quality can be calculated to obtain the proportion of different types of nutrients . When the nutrient aggregation value is calculated, the total value of different types of nutrients is calculated first . After the total value of different types of nutrients is calculated, it is added to obtain the nutrient aggregation value; Step five, the single nutrient value in the fresh food formula is analyzed, which is converted into a standard nutrient value, and an optimized formula is generated according to the standard nutrient value, the single nutrient value in the optimized formula is analyzed, and the single nutrient value deviation of the optimized formula is calculated; Step six, a target search engine is established, an individual feature vector is established for the target search engine, and a fresh food formula is generated; After the individual feature vector is constructed, the individual feature vector is generated into a vector label, and the vector label is arranged in the first branch framework of the corresponding food material part; The individual feature vector includes pet species, pet gender, whether spayed, spaying time, pet age, pet weight, and allergy types, a fresh food formula is generated according to the individual feature vector selected by the user, and thus the entire workflow is completed.

[0021] Although the present application is disclosed in the preferred embodiments as above, it is not intended to limit the present application, and any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, any modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application, which does not deviate from the technical solution of the present application, falls within the protection scope defined by the claims of the present application.

Claims

1. A pet fresh food customized formula generation method based on atlas and multi-modal recognition, characterized in that, The method comprises the following steps: S100, collecting information of different fresh food materials, extracting material parts, analyzing processing methods, cooking methods and cooking times of different material parts; S200, constructing a knowledge base, arranging material information in the knowledge base, and constructing a knowledge graph in the knowledge base according to the material information, and constructing a unified node type for the knowledge base and the knowledge graph; S300, collecting weight changes of the material after cooking, analyzing different nutrient contents in the material after cooking, calculating the nutrient aggregation value of the fresh food formula, and recording it in the knowledge base; S400, analyzing single nutrient values in the fresh food formula, converting them into standard nutrient values, generating an optimized formula according to the standard nutrient values, analyzing single nutrient values in the optimized formula, calculating the single nutrient value deviation of the optimized formula, and recording the optimized formula and the corresponding single nutrient values in the knowledge base; S500, establishing a target search engine, establishing an individual feature vector for the target search engine, retrieving a formula from the knowledge base according to the individual feature vector, and obtaining a fresh food formula.

2. The pet fresh food customized formula generation method based on atlas and multi-modal recognition according to claim 1, characterized in that: In S100, a quadruple is constructed, wherein the format of the quadruple is material-processing method-cooking method-cooking parameter, and the cooking parameter includes cooking temperature and cooking time. 3.The pet fresh food customized formula generation method based on atlas and multi-modal recognition according to claim 2, characterized in that: In S100, the material quality and volume are unified to 100g of edible part; The moisture basis is unified, the dry basis and wet basis are marked as wet basis, and the dry food and moisture description are converted into water content conversion.

4. The pet fresh food customized formula generation method based on atlas and multi-modal recognition according to claim 1, characterized in that: In S200, the node types include materials, parts, nutrients, cooking methods, function labels and material varieties; When constructing the knowledge graph, the material is regarded as the main framework, and the material part is regarded as the first branch framework of the main framework; The cooking method, function label and material variety of the material part are regarded as the second branch framework; Finally, the nutrients possessed by the material after being cooked by the corresponding cooking method are regarded as the third branch framework, wherein the third branch framework is the branch framework of the cooking method.

5. The pet fresh food customized formula generation method based on atlas and multi-modal recognition according to claim 1, characterized in that: In the S300, after the different nutrient contents in the cooked food material are obtained, the ratio between the different nutrient contents and the mass of the food material can be calculated to obtain the proportion of different types of nutrients The proportion of different types of nutrients The calculation formula is as follows: ； wherein is the mass of the foodstuff, is the content of the different kinds of nutrients.

6. The pet fresh food customized formula generation method based on atlas and multi-modal recognition according to claim 5, characterized in that: In the S300, the total value of the different types of nutrients is calculated first when calculating the nutritional polymerization value : ； wherein S is the number of food material categories; and S is the number of food material categories; and After calculating the total value of different types of nutrients, they are added to obtain the nutrient aggregation value.

7. The pet fresh food customized formula generation method based on atlas and multi-modal recognition according to claim 6, characterized in that: The optimized formula has a single nutrient numerical deviation degree The calculation formula is as follows: ； standard nutrient values for different nutrient categories, nutrient values for optimizing the nutrient categories in the formula. 8.The pet fresh food customized formula generation method based on atlas and multi-modal recognition according to claim 3, characterized in that: In S500, after the individual feature vector is constructed, the individual feature vector is generated into a vector label, and the vector label is arranged in the first branch framework of the corresponding material part. 9.The pet fresh food customized formula generation method based on atlas and multi-modal recognition according to claim 8, characterized in that: The individual feature vector includes pet variety, pet gender, whether spayed, spaying time, pet age, pet weight and allergy type, and the fresh food formula is retrieved according to the individual feature vector selected by the user. When no suitable fresh food formula is retrieved, a unique formula is generated according to the individual feature vector, and it is displayed.

Images