Modeling method for learning-free artificial intelligence (AI) scale

Abstract

Provided in the present application is a modeling method for a learning-free artificial intelligence (AI) scale. The method comprises: by using a classification result of a plurality of weighed commodity photographs obtained by means of a neural network model and weighed commodity classification databases from a plurality of countries, associating a photograph of each weighed commodity with corresponding neural vector parameters to form a neural vector photograph, wherein each neural vector photograph is associated with at least one weighed commodity; and on the basis of weighed commodity information and the neural vector photographs, establishing a comparison table including multi-language naming information. The method comprises: using a word segmentation algorithm or an NLP model to associate weighed commodity information of a target scene with neural vector photographs; and on the basis of the similarities between a plurality of neural vector photographs and a weighing neural vector corresponding to a weighed commodity, an AI scale being capable of displaying a neural vector photograph and weighed commodity information according to the similarities. Thus, a new scale can be used immediately, and using intuitive photographs of weighed commodities of various countries as recognition results for display can accommodate environmental differences of different regions and has low capability requirements for personnel.

Description

Modeling method for learning-free artificial intelligence AI scale Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a modeling method for a learning-free artificial intelligence (AI) scale. Background Art

[0002] The development of the digital economy has changed people's clothing, food, housing and transportation habits. With the implementation of artificial intelligence (AI) technology in all aspects of life, retailers' weighing systems have also ushered in an intelligent and digital transformation. They have evolved from scale operators memorizing product codes and manually inputting them to AI scales that use AI models for automatic recognition.

[0003] During the initial use phase of a traditional AI scale, continuous learning over a period of weighing and selling goods is required to accurately identify and recommend product information using image recognition. If the person working on the scale has a low level of education and struggles to accurately identify product names, this will affect the AI ​​scale's learning efficiency. The categories, names, usage scenarios, and languages ​​of weighed goods vary across supermarkets, regions, and countries. The actual scale recognition and learning process is extremely time-consuming and labor-intensive, making it impossible for a new scale to immediately identify and use product photos. AI scales have low universality and compatibility across different regions and countries, and require a certain level of education for the person working on the scale.

[0004] Summary of the Invention

[0005] In order to solve the above problems, this application provides a modeling method and storage medium for a learning-free artificial intelligence AI scale, which can enable the new scale to be used and directly recommend photos of weighed goods. It can be compatible with the naming and language differences in different regions and countries, and has low cultural level and ability requirements for the scale operator.

[0006] In a first aspect, the present application provides a modeling method for a learning-free artificial intelligence (AI) scale, the method comprising:

[0007] S1. A computer device uses a neural network model to classify photos of a plurality of weighed commodities, and establishes an association database based on the classification results and a given weighed commodity classification database. The association database associates each photo of each weighed commodity with a corresponding set of neural vector parameters to form a neural vector photo, and each neural vector photo is associated with one or more weighed commodities;

[0008] The weighed commodity classification database is from multiple regions and / or countries, and the weighed commodity classification data of each region and / or country includes: weighed commodity classification rules used in the region or country and photos of the weighed commodities;

[0009] S2. The computer device establishes a comparison table based on the given weighed commodity information and the neural vector photos in the associated database, wherein the comparison table includes the commodity names of the weighed commodities, and the commodity names include language naming information of the multiple regions and / or countries.

[0010] In one possible implementation manner, the weighed commodities include but are not limited to fresh weighed commodities, and step S1 includes:

[0011] S11. Correlating photos of the multiple fresh weighed products in the fresh weighed product information according to national standard fresh weighed product classification rules for each region and / or country in a given fresh weighed product classification database, and determining a correspondence between the multiple fresh weighed products and the multiple fresh weighed product photos in the given fresh weighed product classification database;

[0012] S12. Establish the neural vector parameters of the fresh weighed commodity photo based on the corresponding relationship and construct a neural vector photo.

[0013] In one possible implementation, the fresh weighed commodity information includes: commodity codes and / or commodity names of multiple fresh weighed commodities, wherein the commodity names include language naming information of multiple regions and / or countries; before step S1, the method further includes:

[0014] In the fresh weighed commodity classification database, the comparison table is pre-established based on the commodity codes, commodity names, and neural vector photos referring to the same fresh weighed commodity in the fresh weighed commodity classification rules used in multiple regions and / or countries; the fresh commodity classification rules include: the laws, regulations, and / or national standards used in the multiple regions and / or countries;

[0015] The neural vector photo is associated with fresh weighed commodity information of at least one region and / or country.

[0016] In a second aspect, the present application provides a modeling method for a learning-free artificial intelligence (AI) scale, the method comprising:

[0017] A. A computer device or AI scale obtains weighing commodity information in a target scene, and uses a given word segmentation algorithm or natural language processing (NLP) model to associate the weighing commodity information in the target scene with a neural vector photo, where the weighing commodity information in the target scene indicates multiple weighing commodities in the target scene, and the neural vector photo is pre-constructed using the modeling method for the learning-free AI scale described in the first aspect;

[0018] B. The AI ​​scale in the target scenario stores the multiple neural vector photos and a comparison table associated with the neural vector photos in a local hard disk, wherein the comparison table includes the names of the weighed products, and the product names include language naming information of the multiple regions and / or countries;

[0019] C. The AI ​​scale uses an image recognition model to process a photo of the item being weighed and generates a weighing neural vector;

[0020] D. The AI ​​scale calculates the similarity between the multiple neural vector photos and the weighing neural vector, and displays the neural vector photos ranked at the top target position in terms of similarity and the weighing commodity information associated with the neural vector photos and ranked at the top target position in terms of association.

[0021] In one possible implementation, the weighed commodities include but are not limited to fresh commodities, and step A includes:

[0022] Acquire information on a variety of fresh weighed commodities in the target scene, and use the given word segmentation algorithm or NLP model to associate the various fresh weighed commodities in the target scene with the neural vector photos, determine the correspondence between the commodity codes and / or commodity names of the various fresh weighed commodities in the target scene and the neural vector photos, and save them in the fresh weighed commodity classification database of the target scene.

[0023] In one possible implementation manner, after step A, the method further includes:

[0024] Based on manual review adjustment information and correction operations, the correspondence between the multiple fresh weighing commodities in the target scene and the multiple neural vector photos in the fresh weighing commodity classification database is updated.

[0025] In one possible implementation manner, step D includes:

[0026] The AI ​​scale calculates the similarity between the multiple neural vector photos and the weighing neural vectors, and calculates the correlation between the neural vector parameters of the neural vector photos and multiple fresh weighing commodity information;

[0027] Determine a comprehensive result of the similarity and the correlation, and display one or more neural vector photos sorted in the front target position and fresh weighing commodity information associated with the neural vector photos on a display screen in descending order of the comprehensive results.

[0028] In one possible implementation, the neural vector parameters include: standard weighing parameters for fresh weighing commodities.

[0029] In one possible implementation, the method further includes:

[0030] The AI ​​scale determines whether it is used for the first time, and executes steps B to D if it is used for the first time.

[0031] In a third aspect, a computing unit is provided, which includes a memory and a processor, wherein the memory stores at least one program, and the at least one program is executed by the processor to implement the modeling method of the learning-free artificial intelligence (AI) scale provided in the first aspect or the second aspect.

[0032] In a fourth aspect, a computer-readable storage medium is provided, in which at least one program is stored, and the at least one program is executed by a processor to implement the modeling method of the learning-free artificial intelligence (AI) scale provided in the first aspect or the second aspect.

[0033] The technical solution provided by this application includes at least the following technical effects:

[0034] The technical solution provided by this application pre-classifies and associates weighed commodity information using a neural network model. A one-to-one database associating photos with neural vector parameters is established based on the classification results and the weighed commodity classification database. Each set of neural vector parameters corresponds to a neural vector photo of a weighed commodity. Each neural vector photo of the fresh weighed commodity is associated with one or more weighed commodities, and each weighed commodity information includes naming information in various languages. A computer device or AI scale pre-classifies the weighed commodities in the target scene with neural vector photos using a word segmentation algorithm or a natural language processing (NLP) model. The AI ​​scale stores data associating multiple neural vector photos with the weighed commodities in the target scene on its local hard disk. When identifying the currently weighed commodity, the scale compares the weighing neural vector of the currently weighed commodity with the multiple pre-stored neural vector photos for similarity, directly displaying the neural vector photo closest to the weighed commodity and one or more in-store weighed commodity information associated with it. The above solution allows the new scale to be used right out of the box, using intuitive photos of goods weighed according to national standards of various countries as the display of recognition results, eliminating the need to learn how to weigh goods on the scale. It is compatible with naming and language differences in different regions, and does not require scale operators to memorize and check product codes and product names, and has low cultural and ability requirements for scale operators. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] FIG1 is a flow chart of a method for modeling a learning-free artificial intelligence (AI) scale according to an embodiment of the present application;

[0036] FIG2 is a flow chart of another method for modeling a learning-free artificial intelligence (AI) scale according to an embodiment of the present application;

[0037] FIG3 is an effect diagram of a product identification and recommendation result provided by an embodiment of the present application;

[0038] FIG4 is a flow chart of another method for modeling a learning-free artificial intelligence (AI) scale according to an embodiment of the present application;

[0039] FIG5 is a schematic diagram of the hardware structure of a computing unit provided in an embodiment of the present application.

[0040] DETAILED DESCRIPTION

[0041] To further illustrate each embodiment, the present application provides drawings. These drawings are part of the disclosure of the present application, which are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, a person of ordinary skill in the art should be able to understand other possible implementations and the advantages of the present application. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components. The term "at least one" in the present application means one or more, and the term "multiple" in the present application means two or more, for example, a plurality of species means two or more species.

[0042] The present application is now further described in conjunction with the accompanying drawings and specific implementation methods. The AI ​​scale in this application is also the abbreviation of artificial intelligence AI scale.

[0043] The modeling method of the learning-free AI scale provided in the embodiment of the present application can be applied to AI scales with computing capabilities.

[0044] FIG1 is a flow chart of a method for modeling a learning-free artificial intelligence (AI) scale provided in an embodiment of the present application. Referring to FIG1 , the method includes steps S1 and S2.

[0045] S1. The computer device uses a neural network model to classify photos of a plurality of weighing commodities, and establishes an associated database according to the classification results and a given weighing commodity classification database.

[0046] Each photo of each weighed commodity in the association database is associated with a corresponding set of neural vector parameters, forming a neural vector photo. Each neural vector photo is associated with one or more weighed commodities. For example, if the neural vector photo is "apple," the weighed commodities associated with the photo include Fuji apples, Huaniu apples, Aksu apples, etc.; for another example, if the neural vector photo is "red apple," the weighed commodities associated with the photo include Red Fuji apples, Gansu red apples, Huaniu red apples, Aksu red apples, etc. In one possible embodiment, the neural vector parameters include: standard weighing parameters for fresh weighed commodities.

[0047] The weighed commodity classification database is from multiple regions and / or countries, and the weighed commodity classification data of each region and / or country includes: weighed commodity classification rules used in the region or country and photos of the weighed commodities.

[0048] Specifically, weighed goods include fresh produce, fruits and vegetables, dry goods, and snack foods sold in supermarkets. The weighed goods photos can be derived from national laws and regulations governing fresh produce and fruits, as well as national standards. They can also be derived from historically collected fresh produce classification data from various regions, countries, and supermarkets. In this embodiment, a neural network model is used to classify, summarize, and perform feature processing on the fresh produce classification database, ultimately representing each weighed fresh produce photo as neural vector parameters.

[0049] The national standard is, for example, a commodity classification standard prescribed by a designated region, country or unified standard.

[0050] In one possible implementation, weighing commodities includes but is not limited to weighing fresh commodities, and step S1 includes:

[0051] S11. Correlating photos of multiple fresh weighing products in the fresh weighing product information according to national standard fresh weighing product classification rules for each region and / or country in a given fresh weighing product classification database, and determining a correspondence between the multiple fresh weighing products and the multiple fresh weighing product photos in the given fresh weighing product classification database;

[0052] S12. Based on the corresponding relationship, establish the neural vector parameters of the fresh weighing product photo and construct the neural vector photo.

[0053] In one possible implementation, the fresh produce weighing information includes: product codes and / or product names of multiple fresh produce weighing products, where the product names include language naming information for multiple regions and / or countries. Exemplarily, the weighing product information can be provided to the computer device in the form of a product name list.

[0054] In the embodiments of the present application, a neural network model is used to classify, summarize, and perform feature processing on a database of fresh produce classifications. Finally, images of each fresh produce classification are presented as neural vector images. The neural vector images also include standard weighing parameters for the fresh produce classifications. These parameters, such as the standard weight range and the stable weighing speed, can be included in the neural vector images as supplementary features for determining the classification of fresh produce classifications.

[0055] Taking into account that different supermarkets, regions and countries have different categories and actual names for the same weighing commodity, before executing step S1, the embodiment of the present application pre-associates the weighing commodity classification data of each region and country in the weighing commodity classification database.

[0056] In this example, before step S1, the modeling method of the learning-free artificial intelligence AI scale further includes:

[0057] In the fresh weighing commodity classification database, a comparison table is established in advance based on the commodity codes, commodity names and various neural vector photos referring to the same fresh weighing commodity in the fresh weighing commodity classification rules used in multiple regions and / or countries; the fresh commodity classification rules include: laws, regulations and / or national standards used in multiple regions and / or countries; wherein the neural vector photos are associated with fresh weighing commodity information of at least one region and / or country.

[0058] Based on this, product names and codes used in different scenarios (e.g., fresh produce supermarkets) can be automatically associated with existing names in fresh produce classification databases across regions and countries. Specifically, this association can be accomplished through word-to-word comparison, enabling one-to-many cross-regional name association.

[0059] In the embodiments of the present application, the computer device can be any electronic device with computing capabilities, such as a personal computer, a PDA, a tablet computer and other terminal computer devices, or a server, a server cluster consisting of multiple physical servers or a distributed file system, or a cloud server cluster that provides cloud storage and cloud services, cloud databases, cloud computing, cloud functions, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), big data and artificial intelligence platforms and other basic cloud computing services. This application does not limit this.

[0060] S2. The computer device establishes a comparison table based on the given weighed commodity information and the neural vector photos in the associated database.

[0061] The comparison table contains the product names of the weighed goods, which can be found in multiple languages ​​across regions and / or countries. For example, the product name "Snake Apple" can be called "Red Delicious Apple," "Red Delicious Apple," or "Apple" in different languages ​​across countries.

[0062] In an embodiment of the present application, the various fresh weighing commodities involved in the target scene are classified, and according to the fresh weighing commodity information involved in the target scene, the fresh weighing commodity information and the corresponding neural vector photos and the comparison table associated with the neural vector photos are stored in the local hard disk of the AI ​​scale.

[0063] In this embodiment of the present application, a computer device or AI scale pre-generates neural vector image associations for various weighed product information, establishing a correlation database and comparison table. The AI ​​scale then stores the weighed product information and the neural vector image association results for the target scene on its hard drive, allowing the AI ​​scale to be used immediately in the target scene.

[0064] Among them, the target scene can be scenes such as shopping malls, supermarkets, indoor / outdoor markets in different regions and countries that require the use of weighing tools. Since the languages ​​used and the classification rules adopted in different regions and countries will lead to different naming / coding of the same fresh weighing commodity, the present application proposes a modeling method for a learning-free artificial intelligence AI scale that can overcome the differences in the names of fresh weighing commodities in different regions. Figure 2 is a flow chart of another modeling method for a learning-free artificial intelligence AI scale provided in an embodiment of the present application. Referring to Figure 2, the method includes steps A to D.

[0065] A. The computer device or AI scale obtains the weighing product information of the target scene and uses a given word segmentation algorithm or natural language processing (NLP) model to associate the weighing product information in the target scene with the neural vector photo.

[0066] The weighing commodity information in the target scene indicates a plurality of weighing commodities in the target scene, and the neural vector photo is pre-constructed using the method introduced in the above steps S1 to S2.

[0067] In one possible implementation, weighing commodities includes but is not limited to weighing fresh commodities, and step A includes:

[0068] Acquire information on a variety of fresh weighing products in the target scene, and use a given word segmentation algorithm or NLP model to associate the various fresh weighing products in the target scene with neural vector photos, determine the correspondence between the product codes and / or product names of the various fresh weighing products in the target scene and the neural vector photos, and save them in the fresh weighing product classification database of the target scene.

[0069] For example, a product name in the fresh produce category is "Red Fuji Apple," and the multiple keywords obtained through word segmentation are, for example, "apple," "Red Fuji," "red," and "Fuji." By matching these keywords against the fresh produce classification database, the corresponding fresh produce category can be determined, thereby identifying the fresh produce photos involved in the target scene.

[0070] In one possible implementation, after step A, the modeling method for a learning-free artificial intelligence (AI) scale further includes:

[0071] Based on manual review, adjustment information, and corrections, the correspondences between the various fresh produce items in the target scene and the various neural vector images in the fresh produce classification database are updated. To improve the accuracy of the established correspondences, the correspondences between the various fresh produce items and the various fresh produce images in the fresh produce classification database can be further updated based on manual adjustments, enabling precise manual corrections.

[0072] B. The AI ​​scale in the target scene stores multiple neural vector photos and a comparison table associated with the neural vector photos in the local hard disk.

[0073] The comparison table includes the product names of the weighed products, and the product names include language naming information of multiple regions and / or countries.

[0074] In an embodiment of the present application, an AI scale is used in a target scenario. Specifically, the AI ​​scale detects that fresh produce is placed on the scale platform through a weight sensor and / or camera, and then collects photos of the fresh produce.

[0075] For example, there are three hundred neural vector photos summarized by step S2, and there are, for example, one thousand specific fresh weighing commodities in the target scene. Each neural vector photo stored by the AI ​​scale corresponds to a fresh weighing commodity photo, and each fresh weighing commodity photo actually corresponds to multiple similar fresh weighing commodities in the store. For example, if the neural vector photo is a photo of a Red Fuji apple, its associated commodities may be Shandong Red Fuji, Shaanxi Red Fuji, Aksu apple, etc. When the AI ​​scale receives multiple neural vector photos, it also receives the comparison table established in step S21. In addition, this comparison table can be manually adjusted and corrected.

[0076] C. The AI ​​scale uses an image recognition model to process photos of the item being weighed and generate a neural vector for the weight.

[0077] The AI ​​scale runs a pre-trained image recognition model. It can be used directly in the target scenario, using the configured image recognition model to perform neural vector photo feature extraction on collected photos of fresh produce being weighed.

[0078] In the technical solution of the present application, the AI ​​scale determines whether it is the initial use. If it is the initial use, the AI ​​scale uses the neural vector photos stored in the local hard disk of the AI ​​scale in step S3 to match and recommend photos of fresh weighing products.

[0079] Among them, the weighing neural vector photo includes: the collected photo of the current fresh weighing product and the real-time weighing parameters of the current fresh weighing product, for example, the real-time weighing parameters include the weighing AD value, the weighing AD change rate, etc.

[0080] Optionally, the image recognition model is, for example, a convolutional neural network model or a YOLO target detection model, etc., which is not limited in this application.

[0081] D. The AI ​​scale calculates the similarity between multiple neural vector photos and weighing neural vectors, and displays the neural vector photos ranked at the top of the similarity list and the weighing product information associated with the neural vector photos and ranked at the top of the similarity list.

[0082] In one possible implementation, step D includes:

[0083] D1 and the AI ​​scale calculate the similarity between multiple neural vector photos and weighing neural vectors, and calculate the correlation between the neural vector parameters of the neural vector photos and the information of multiple fresh weighing products.

[0084] Among them, similarity is used to identify and classify products from the perspective of image feature matching, and association is used to identify and classify products from the perspective of text matching.

[0085] D2. Determine the combined results of similarity and association, and display one or more neural vector photos and corresponding fresh produce weighing information in the top target position on the display screen in descending order of the combined results. For example, the top target position is the first five.

[0086] In this embodiment of the present application, the "weighing neural vector photo" generated by image recognition is compared with the "neural vector photos" corresponding to multiple pre-stored photos of fresh produce weighing products to determine the most similar "neural vector photos." These photos are then displayed on the display in order of similarity. This similarity is equivalent to the confidence level that the currently weighed fresh produce product is the product corresponding to the corresponding neural vector photo.

[0087] In one possible implementation, a confidence threshold is set. If the similarity between each neural vector image is less than the confidence threshold, it indicates that the fresh produce category to which the current fresh produce product belongs has not been successfully identified. In this exemplary embodiment, manual search methods can be performed, including: phonetic search (pinyin initials or English initials), product code (fresh produce code) search, price search, and voice search.

[0088] The key point of the technical solution provided by this application is that the confidence level of the matching results of each major category of fresh weighing goods is displayed in sequence by displaying photos of fresh weighing goods. Compared with displaying text, the intuitiveness of photos can effectively accommodate differences in different regions, countries, and languages, and the cultural level and ability requirements of the weighing personnel are very low.

[0089] In the embodiments of the present application, the photos of fresh weighing products are universal and general images for different regions. For example, even in different countries and regions, there is a consensus on the basic image features of apples. Therefore, the image of apples can be used as a photo of fresh weighing products. Apples actually include hundreds of varieties such as Red Fuji, Huaniu, Aksu, etc., and the names of these hundreds of varieties are different in different countries and regions. This application takes into account the difficulties brought about by the name differences in the cross-regional use of AI scale products, and adopts photos of fresh weighing products as the presentation standard for fresh weighing product identification. No matter what region it is in, no matter what fresh weighing product name is used in this region, intuitive neural vector photos are used to recommend to the scale personnel as the recommendation of the closest classification of fresh weighing products.

[0090] In some embodiments, the AI ​​scale displays a photo of the fresh produce item and provides a re-identification control. In response to triggering the re-identification control, the AI ​​scale re-executes steps B through D, re-identifying and recommending the fresh produce item being weighed.

[0091] Among them, various neural vector photos are used to display the standard features of the corresponding fresh weighing products. For example, the neural vector photo is an apple, which can display images of the unique features of apples. For example, the unique features of Aksu apples are reflected in the cross-section of the core, and the unique features of Huaniu apples are reflected in the fruit shape and skin color.

[0092] For ease of understanding, the present application embodiment provides a rendering of the product identification and recommendation results. Referring to Figure 3, the AI ​​scale can directly identify the multiple recognition results in Figure 3 without learning, and displays the corresponding neural vector photos in descending order of similarity, and displays the specific product name and product code (fresh weighing product information) associated with the neural vector photo. Referring to Figure 3, the photos and fresh weighing product information displayed in descending order of similarity and association are "Red Apple", "Red Fuji", "Gansu Red Apple", "Premium Red Apple", and "Aksu Apple".

[0093] In one possible implementation, the AI ​​scale can operate offline. The AI ​​scale can also access multiple neural vector images and corresponding comparison tables stored by other AI scales in the target scene via the internet. The AI ​​scale can also broadcast the parameters of its image recognition model to other AI scales for synchronization. Therefore, the AI ​​scale can also utilize historical image recognition models learned from other smart electronic scales to identify photos of the fresh produce currently being weighed, directly determining the recognition result for the fresh produce currently being weighed.

[0094] To facilitate understanding of the logical connections between the various possible implementations described above, an embodiment of the present application provides a flowchart illustrating another method for modeling a learning-free artificial intelligence (AI) scale. Referring to Figure 4 , the process begins by obtaining product information (category information for fresh produce weighing) from the store (target scenario). Using a word segmentation algorithm or NLP model, the in-store products are associated with corresponding neural vector photos. Manual review and correction can also be performed. The AI ​​scale stores the associated results (neural vector photos) on its local hard drive or transmits the associated data to the scale, enabling the scale to directly identify product photos without learning. Once the AI ​​scale detects the currently weighed product, it displays the neural vector photo closest to the product and the in-store product information associated with the neural vector photo (fresh produce name, product code, product price, in-store product photo, etc.). Alternatively, if the corresponding neural vector photo or associated in-store product information is not found, a manual search for the corresponding product can be performed using audio search, fresh produce code (item number or product code), price, or voice. The weighing staff can browse the image and click the correct product hot key to complete the weighing sale.

[0095] The technical solution provided by the embodiments of the present application enables a new scale to be used immediately, and uses intuitive photos of the weighed goods as the display of the identification results. It is compatible with naming and language differences in different regions, and does not require the weighing personnel to memorize and check the product codes and product names. The cultural level and ability requirements for the weighing personnel are low. Considering the difficulty brought about by name differences in the cross-regional use of AI scale products, photos of fresh weighed goods are used as the presentation standard for fresh weighed goods identification. This allows the weighing personnel to use intuitive photos of fresh weighed goods as the closest classification recommendation for fresh weighed goods, regardless of the region and the name of the fresh weighed goods used in the region.

[0096] The present application provides a computing unit that can be used to implement a modeling method for a learning-free artificial intelligence (AI) scale. Figure 5 is a schematic diagram of the hardware structure of a computing unit provided in an embodiment of the present application. As shown in Figure 5, the computing unit includes a processor 501, a memory 502, a bus 503, and a computer program (for example, a pre-trained image recognition model) stored in the memory 502 and executable on the processor 501. The processor 501 includes one or more processing cores, and the memory 502 is connected to the processor 501 via the bus 503. The memory 502 is used to store program instructions. When the processor executes the computer program, all or part of the steps in the embodiment of the modeling method for the learning-free artificial intelligence (AI) scale provided in the present application are implemented.

[0097] Furthermore, as an executable solution, a computer unit may include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the aforementioned computer unit configuration is merely an example and does not limit the computer unit. A computer unit may include more or fewer components than those described above, or a combination of certain components, or different components. For example, a computer unit may also include input / output devices, network access devices, buses, etc., although this is not limited in the present embodiment.

[0098] Furthermore, as an executable solution, the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor. The processor is the control center of the computer unit and connects the various parts of the entire computer unit using various interfaces and lines.

[0099] The memory can be used to store computer programs and / or modules. The processor implements the various functions of the computer unit by running or executing the computer programs and / or modules stored in the memory and accessing the data stored in the memory. The memory can primarily include a program storage area and a data storage area. The program storage area can store the operating system and at least one application required for a function; the data storage area can store data generated based on the use of the mobile phone. Furthermore, the memory can include high-speed random access memory (RAM) and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0100] The present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the above method in the embodiment of the present application are implemented.

[0101] If the modules / units integrated into the computer unit are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the present application can also implement all or part of the process steps in the above-mentioned method embodiments by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include any entity or device capable of carrying the computer program code, recording medium, USB flash drive, removable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased based on the requirements of legislation and patent practice in a jurisdiction.

[0102] Although the present application has been specifically shown and described in conjunction with preferred embodiments, it should be understood by those skilled in the art that various changes in form and details may be made to the present application without departing from the spirit and scope of the present application as defined by the appended claims, and all such changes are within the scope of protection of the present application.

Claims

1. A modeling method for a no-learning artificial intelligence AI scale, characterized in that The method includes: S1. The computer device classifies photos of various weighed commodities using a neural network model, and establishes an association database based on the classification results and a given weighed commodity classification database. In the association database, each photo of a weighed commodity is associated with a corresponding set of neural vector parameters one by one to form a neural vector photo, and one or more weighed commodities are associated with each neural vector photo. Among them, the weighed commodity classification database comes from multiple regions and / or countries. The weighed commodity classification data of each region and / or country includes: the weighed commodity classification rules used in the region or country and photos of weighed commodities. S2. The computer device establishes a comparison table based on the given weighed commodity information and the neural vector photos in the association database. Among them, the comparison table includes the commodity names of the weighed commodities, and the commodity naming includes the language naming information of the multiple regions and / or countries.

2. The AI scale modeling method according to claim 1, wherein The weighed commodities include but are not limited to fresh weighed commodities. The step S1 includes: S11. According to the national standard fresh weighed commodity classification rules of each region and / or country in the given fresh weighed commodity classification database, perform photo association on the various fresh weighed commodities in the fresh weighed commodity information to determine the corresponding relationship between the various fresh weighed commodities and the photos of the various fresh weighed commodities in the given fresh weighed commodity classification database. S12. According to the corresponding relationship, establish each of the neural vector parameters of the fresh weighed commodity photos and construct neural vector photos.

3. The AI scale modeling method according to claim 2, wherein The fresh weighed commodity information includes: the commodity codes and / or commodity names of the various fresh weighed commodities, and the commodity names include the language naming information of multiple regions and / or countries. Before the step S1, the method further includes: In the fresh weighed commodity classification database, a comparison table is established in advance according to the commodity codes, commodity names, and the respective neural vector photos that represent the same fresh weighed commodity in the fresh weighed commodity classification rules used in multiple regions and / or countries. The fresh commodity classification rules include: laws, regulations, and / or national standards used in the multiple regions and / or countries. At least one region and / or country's fresh weighed commodity information is associated with the neural vector photo.

4. A modeling method for a no-learning artificial intelligence AI scale, characterized in that, The method includes: A. The computer device or the AI scale obtains the weighed commodity information of the target scene, and associates the weighed commodity information in the target scene with the neural vector photos by using a given word segmentation algorithm or a natural language processing NLP model. The weighed commodity information in the target scene indicates various weighed commodities in the target scene, and the neural vector photos are pre-constructed by using the method described in any one of claims 1 to 3. B. The AI scale in the target scene stores the multiple neural vector photos and the comparison table associated with the neural vector photos in its local hard disk. The comparison table includes the commodity names of the weighed commodities, and the commodity naming includes the language naming information of the multiple regions and / or countries. C. The AI scale uses an image recognition model to process the photo of the currently weighed commodity to generate a weighed neural vector. D. The AI scale calculates the similarity between the multiple neural vector photos and the weighing neural vector, and displays the neural vector photos with the highest similarity rankings and the weighing commodity information associated with the neural vector photos and having the highest association rankings.

5. The modeling method of the learning-free artificial intelligence AI scale according to claim 4, characterized in that, The weighing commodities include, but are not limited to, fresh weighing commodities, and step A includes: Obtaining information on various fresh weighing commodities in the target scenario, and using the given word segmentation algorithm or NLP model to associate the various fresh weighing commodities in the target scenario with the neural vector photos, determining the corresponding relationships between the commodity codes and / or commodity names of the various fresh weighing commodities in the target scenario and the neural vector photos, and storing them in the fresh weighing commodity classification database of the target scenario.

6. The modeling method of the learning-free artificial intelligence AI scale according to claim 5, characterized in that After step A, the method further includes: Updating the corresponding relationships between the various fresh weighing commodities in the target scenario and the various neural vector photos in the fresh weighing commodity classification database according to the manual review adjustment information and rectification operations.

7. The modeling method of the learning-free artificial intelligence AI scale according to claim 4, characterized in that, Step D includes: The AI scale calculates the similarity between the multiple neural vector photos and the weighing neural vector respectively, and calculates the association degree between the neural vector parameters of the neural vector photos and the information on various fresh weighing commodities; Determining the comprehensive result of the similarity and the association degree, and displaying, on the display screen in descending order of the comprehensive result, one or more of the neural vector photos with the highest rankings and the fresh weighing commodity information associated with the neural vector photos.

8. The modeling method of the non-learning artificial intelligence AI scale according to claim 4, characterized in that, The neural vector parameters include: standard weighing parameters of fresh weighing commodities.

9. The modeling method of the learning-free artificial intelligence AI scale according to claim 4, characterized in that, The method further includes: The AI scale determines whether it is in the initial use stage, and performs steps B to D when the AI scale is in the initial use stage.

Images