Customized product data management system and method based on cloud platform

Abstract

The invention relates to the technical field of data management, and particularly discloses a customized product data management system and method based on a cloud platform, and the method comprises the steps: firstly uploading existing product design drawings to the cloud platform, carrying out the image feature extraction of each existing product design drawing through a deep neural network disposed on the cloud platform, and carrying out the image feature extraction; according to the method, key visual elements and design intentions in all existing product design drawings are captured according to the user customization requirements, and on the basis, efficient semantic query matching analysis is further performed on the product customization requirements input by the user and all the existing product design drawings, so that the existing product design drawings matched with the user customization requirements are accurately retrieved. In this way, designers can be helped to quickly position the design drawings meeting specific requirements, and the efficiency and quality of customized product development are remarkably improved.

Description

Technical Field

[0001] This application relates to the technical field of data management, and more specifically, to a customized product data management system and method based on a cloud platform. Background Art

[0002] In today's manufacturing and product design fields, with the increasingly fierce market competition and the continuous growth of consumers' personalized needs, the production of customized products has become one of the key strategies for enterprises to enhance their competitiveness. In the development stage of customized products, many design elements, modules or structures may be similar in different product designs. If designers are not aware of the existing design situations and repeat the design work, it will not only increase the design cost and extend the product development cycle, but also introduce quality risks due to the immaturity of the new design. Therefore, efficiently managing and utilizing existing product design resources is of great significance for shortening the product market time, reducing the R & D cost, and improving customer satisfaction.

[0003] However, most of the existing product design resource management systems search based on simple text tags or basic file attributes, lacking the ability to deeply understand and analyze the content of design drawings. When faced with complex customization requirements put forward by customers, such as specific shape features, function combinations or aesthetic styles, it may not be able to effectively screen out the reference drawings that can be used as the starting point of the design, resulting in low design efficiency and resource waste, and seriously affecting the development efficiency of customized products.

[0004] Therefore, an optimized customized product data management system and method based on a cloud platform are expected. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. Embodiments of this application provide a customized product data management system and method based on a cloud platform. First, existing product design drawings are uploaded to the cloud platform, and image features of each existing product design drawing are extracted through a deep neural network deployed on the cloud platform to capture the key visual elements and design intents in each existing product design drawing. On this basis, the product customization requirements input by the user are further subjected to efficient semantic query matching analysis with each existing product design drawing, so as to accurately retrieve the existing product design drawings that match the user's customization requirements. In this way, it can help designers quickly locate the design drawings that meet specific requirements, and significantly improve the efficiency and quality of the development of customized products.

[0006] According to one aspect of this application, a customized product data management method based on a cloud platform is provided, which includes:

[0007] Obtain the natural language description of the product customization requirements input by the user;

[0008] Upload the existing product design drawings and the natural language description of the product customization requirements to the cloud platform;

[0009] On the cloud platform, perform image feature extraction on each of the existing product design drawings in the set of existing product design drawings to obtain a set of semantic encoding feature vectors of the existing product design images;

[0010] On the cloud platform, perform semantic encoding on the natural language description of the product customization requirements to obtain a semantic understanding encoding vector of the product customization requirements;

[0011] On the cloud platform, perform fast location semantic query encoding on the semantic understanding encoding vector of the product customization requirements and the set of semantic encoding feature vectors of the existing product design images to obtain a semantic encoding vector of the product customization requirements query response;

[0012] On the cloud platform, based on the semantic encoding vector of the product customization requirements query response, return the existing product design drawings that match the user's customization requirements as the retrieval result.

[0013] According to another aspect of the present application, there is provided a customized product data management system based on a cloud platform, which includes:

[0014] A product customization requirements acquisition module for acquiring the natural language description of the product customization requirements input by the user;

[0015] An upload synchronization module for uploading the existing product design drawings and the natural language description of the product customization requirements to the cloud platform;

[0016] An image feature extraction module for performing image feature extraction on each of the existing product design drawings in the set of existing product design drawings on the cloud platform to obtain a set of semantic encoding feature vectors of the existing product design images;

[0017] A semantic encoding module for performing semantic encoding on the natural language description of the product customization requirements on the cloud platform to obtain a semantic understanding encoding vector of the product customization requirements;

[0018] A semantic query encoding module for performing fast location semantic query encoding on the semantic understanding encoding vector of the product customization requirements and the set of semantic encoding feature vectors of the existing product design images on the cloud platform to obtain a semantic encoding vector of the product customization requirements query response;

[0019] A retrieval result return module for returning the existing product design drawings that match the user's customization requirements as the retrieval result on the cloud platform based on the semantic encoding vector of the product customization requirements query response.

[0020] Compared with the prior art, the customized product data management system and method based on the cloud platform provided by the present application first upload the existing product design drawings to the cloud platform, and extract image features from each existing product design drawing through a deep neural network deployed on the cloud platform to capture the key visual elements and design intentions in each existing product design drawing. On this basis, the product customization requirements input by the user are further subjected to efficient semantic query matching analysis with each existing product design drawing, so as to accurately retrieve the existing product design drawings that match the user's customization requirements. In this way, it can help designers quickly locate the design drawings that meet specific requirements, and significantly improve the efficiency and quality of customized product development. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By describing the embodiments of the present application in more detail with reference to the accompanying drawings, the above and other objects, features, and advantages of the present application will become more obvious. The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0022] Figure 1 It is a flowchart of a customized product data management method based on the cloud platform according to an embodiment of the present application.

[0023] Figure 2 It is a schematic diagram of data flow of a customized product data management method based on the cloud platform according to an embodiment of the present application.

[0024] Figure 3 It is a flowchart of sub-step S5 of a customized product data management method based on the cloud platform according to an embodiment of the present application.

[0025] Figure 4 It is a flowchart of sub-step S51 of a customized product data management method based on the cloud platform according to an embodiment of the present application.

[0026] Figure 5 It is a block diagram of a customized product data management system based on the cloud platform according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] As shown in the present application and the claims, unless the context clearly indicates otherwise, words such as "a", "an", "one", and / or "the" are not specifically singular and may also include plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the clearly identified steps and elements, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0028] Although the present application makes various references to certain modules in the system according to embodiments of the present application, any number of different modules may be used and run on a user terminal and / or a server. The modules are merely illustrative, and different aspects of the system and method may use different modules.

[0029] Flowcharts are used in the present application to illustrate the operations performed by the system according to embodiments of the present application. It should be understood that the operations above or below do not necessarily have to be performed precisely in order. On the contrary, various steps may be processed in reverse order or simultaneously as needed. At the same time, other operations may also be added to these processes, or one or more steps may be removed from these processes.

[0030] Next, exemplary embodiments of the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0031] It is worth noting that in the present application, all actions of obtaining data are carried out on the premise of complying with the corresponding data protection regulations and policies of the country where the data is located and obtaining authorization from the owner of the corresponding device.

[0032] To address the technical problems described in the above background art, the present application proposes a customized product data management method based on a cloud platform. First, existing product design drawings are uploaded to the cloud platform, and a deep neural network deployed on the cloud platform extracts image features from each existing product design drawing to capture key visual elements and design intentions in each existing product design drawing. On this basis, the product customization requirements input by the user are further subjected to an efficient semantic query matching analysis with each existing product design drawing, so as to accurately retrieve the existing product design drawings that match the user's customization requirements. In this way, it can help designers quickly locate the design drawings that meet specific requirements, and significantly improve the efficiency and quality of customized product development.

[0033] Figure 1 FIG. is a flowchart of a customized product data management method based on a cloud platform according to an embodiment of the present application. Figure 2 FIG. is a schematic diagram of data flow of a customized product data management method based on a cloud platform according to an embodiment of the present application. As Figure 1 and Figure 2As shown, the customized product data management method based on a cloud platform includes the steps of: S1, obtaining a natural language description of product customization requirements input by a user; S2, uploading existing product design drawings and the natural language description of the product customization requirements to the cloud platform; S3, on the cloud platform, respectively performing image feature extraction on each of the existing product design drawings in the set of the existing product design drawings to obtain a set of semantic encoding feature vectors of the existing product design images; S4, on the cloud platform, performing semantic encoding on the natural language description of the product customization requirements to obtain a semantic understanding encoding vector of the product customization requirements; S5, on the cloud platform, performing fast positioning semantic query encoding on the semantic understanding encoding vector of the product customization requirements and the set of the semantic encoding feature vectors of the existing product design images to obtain a semantic encoding vector of a product customization requirements query response; S6, on the cloud platform, based on the semantic encoding vector of the product customization requirements query response, returning an existing product design drawing that matches the user's customization requirements as a retrieval result.

[0034] In the above customized product data management method based on a cloud platform, in step S1, a natural language description of product customization requirements input by a user is obtained. It should be understood that the natural language description of the product customization requirements input by the user can be any detailed description of product design, such as dimensions, shape, color, material, function, etc.

[0035] Specifically, in order to obtain a natural language description of product customization requirements input by a user, first, an intuitive and easy-to-use user interface (UI) needs to be constructed to enable the user to conveniently express customization requirements. When designing the user interface, various input methods such as a text input box, voice recognition, and graphic selection should be considered. The text input box allows the user to directly type in detailed descriptions or specific requirements, such as attributes of the product like dimensions, color, material, etc. At the same time, by introducing advanced voice recognition technology, the user can verbally express requirements, which is especially suitable for those user groups who are more inclined to speak rather than type, and also improves efficiency, especially when the requirement description is complex or multi-step. In addition, the graphic selection function enables the user to select and adjust product characteristics by clicking, dragging, or drawing, such as dragging components onto the canvas to define the product layout, or selecting and adjusting the size and position from a preset shape library. Such an operation method is intuitive and easy to get started.

[0036] To enable users to describe their customization requirements as detailed and accurately as possible, a guided questionnaire or dialogue system can be embedded in the UI. These questions can be customized according to the product category, gradually guiding the thinking and answering specific requirements regarding dimensions, functions, materials, colors, etc. At the same time, based on the existing input, the system can give real-time intelligent tips or recommendations to help improve the requirement description. For example, if a certain specific function module is selected, the system can automatically suggest matching other components or features to achieve the best matching effect.

[0037] In addition, throughout the process, the security and privacy protection of user data must be highly emphasized. All data involving personal information should be strictly encrypted and comply with relevant laws and regulations. Additionally, a sound permission management mechanism needs to be established to ensure that only authorized personnel can access sensitive information. At the same time, transparency is also an important point, that is, to let users clearly know how the data will be used and what protection measures will be taken. By supporting diverse input methods, providing visualization options and templates, and enhancing the overall user experience and accuracy, this user interface can significantly improve the user experience when expressing customization requirements, while ensuring the accuracy and integrity of information transmission.

[0038] In the above customized product data management method based on the cloud platform, in step S2, the existing product design drawings and the natural language description of the product customization requirements are uploaded to the cloud platform. It should be understood that this application takes into account that the traditional product design drawing management method usually relies on local servers or storage devices. However, local storage not only has limited data storage capacity and is difficult to cope with the management requirements of large-scale data, but also has certain risks in terms of data security and backup and recovery capabilities. As a new technology that has developed very rapidly in recent years, cloud computing technology is developed and applied in the Internet, which can achieve efficient utilization and flexible management of resources. Through network connection, cloud computing technology can use low-cost computing systems for large-scale data processing and computing, and provide powerful data backup and recovery functions. Based on this, in order to provide a more secure, reliable and easily expandable data management environment, this application further uploads the existing product design drawings and the natural language description of the product customization requirements to the cloud platform to utilize the advantages of cloud computing technology to achieve centralized storage and processing of data.

[0039] Specifically, the cloud platform provides almost unlimited storage space, easily meeting the growing demand for massive data. Manufacturers and designers no longer need to worry about storage capacity issues and can flexibly expand storage resources according to actual needs to ensure that all historical design drawings and the latest descriptions of customization requirements are properly preserved. At the same time, cloud computing technology emphasizes resource sharing and efficient utilization. The computing resources on the cloud platform can be dynamically allocated according to task loads, which means that even in the face of data processing requests during peak hours, the system can maintain high performance and stable operation. This flexibility is crucial for quickly responding to market changes and shortening the product development cycle. In addition, the cloud platform also has a perfect data protection mechanism, including measures such as encrypted transmission, access control, and regular backups, greatly improving data security. Even in case of unexpected situations such as hardware failures or natural disasters, data can be quickly restored from backups, minimizing losses to the greatest extent. With the powerful computing ability of the cloud platform, it is possible to perform in-depth analysis and feature extraction on the uploaded product design drawings, as well as semantic encoding on natural language descriptions, etc. This provides a solid foundation for subsequent intelligent retrieval and matching, helping to improve the development efficiency and service quality of customized products.

[0040] To achieve the above goals, uploading existing product design drawings and natural language descriptions of product customization requirements to the cloud platform involves several key steps. First of all, it is crucial to select a stable, reliable cloud service provider with strong technical support and a good reputation. This requires evaluating factors such as the service level agreement (SLA), data privacy policy, cost structure, etc. of each vendor to ensure that the selected platform can meet specific requirements and provide long-term support. Before uploading, it is necessary to sort out and preprocess the existing product design drawings and natural language descriptions provided by users. Ensure that the file formats are unified, the information is complete and accurate, and any unnecessary sensitive information is removed. For large files or complex 3D models, consider compressing or splitting them to optimize the upload speed. Establish a secure transmission channel using encryption protocols such as SSL / TLS to ensure that data will not be stolen or tampered with during network transmission. At the same time, set strict access permissions so that only authorized personnel can upload and download relevant files to prevent unauthorized access. Upload the prepared files to the selected cloud platform through the API interface or graphical interface. Depending on the file size and quantity, it may be necessary to upload in batches to avoid network congestion. Monitor the upload progress to ensure that all files are successfully uploaded without errors. After the upload is completed, immediately conduct verification to check whether all files have reached the cloud platform intact. The file consistency can be confirmed by comparing hash values or other verification methods. Once any problems are found, take corrective measures in a timely manner to re-upload the affected parts. Configure automatic backup and recovery strategies on the cloud platform, setting reasonable backup frequencies and retention periods. Regularly test the effectiveness of the backups to ensure that data can be quickly restored in case of an emergency. In addition, consider cross-region replication of important files to further enhance data redundancy and disaster recovery capabilities. As the business develops and technology advances, continuously evaluate the performance of the currently used cloud service platform to find possible bottlenecks or deficiencies. Based on feedback and technological trends, adjust configuration parameters in a timely manner or replace them with more advanced solutions to maintain competitiveness and maximize the return on investment.

[0041] Uploading the existing product design drawings and the natural language descriptions of product customization requirements to the cloud platform is not only an innovation to the traditional local storage method, but also a key step in building a more secure, reliable and easily expandable data management environment by leveraging the advantages of cloud computing technology. This can not only solve the current storage and management problems, but also bring more efficient resource allocation and stronger competitive advantages to enterprises. Through centralized storage and processing, the cloud platform enables enterprises to better manage and utilize valuable intellectual property assets, accelerate the innovation process, and meet the personalized needs of the market. The cloud platform allows team members to access the latest versions of design drawings and requirement documents in real time regardless of their locations, promoting communication and collaboration across departments, reducing misunderstandings and rework caused by version asynchronization, and improving the overall work efficiency. At the same time, management can directly view detailed project materials in the cloud, obtain comprehensive information support, make wiser decisions more quickly, reduce the time consumption of paper document transmission, and promote the ability of agile management and rapid response to market changes. After adopting the cloud platform, enterprises no longer need to invest a large amount of funds in purchasing and maintaining local server hardware, saving the cost of upfront construction and later operation and maintenance. The pay-as-you-go model allows enterprises to flexibly select the required service level within the budget. In addition, the cloud platform strictly complies with international standard data protection regulations and adopts multiple protection measures to ensure the security of customer information. Even in case of emergencies, it can quickly resume normal operation with a perfect disaster recovery system, protecting the core interests of enterprises from damage to the greatest extent.

[0042] In the above customized product data management method based on the cloud platform, in step S3, on the cloud platform, image feature extraction is respectively performed on each existing product design drawing in the set of existing product design drawings to obtain a set of semantic encoding feature vectors of the existing product design images. In a specific example of the present application, step S3 includes: respectively inputting each existing product design drawing in the set of existing product design drawings into a product design image feature extractor based on the FPT model to obtain a set of semantic encoding feature vectors of the existing product design images. It should be understood that the present application takes into account that product design drawings usually contain different levels of design information (such as overall shape, local structure, texture, etc.). Therefore, in order to effectively capture the key visual elements and design intentions of each existing product design drawing, the present application uses the FPT model to construct a product design image feature extractor, and performs image feature extraction on each existing product design drawing respectively, so as to utilize the multi-scale feature extraction and cross-scale feature interaction capabilities of the FPT model to fully excavate the key semantic information of each existing product design drawing, thereby obtaining a set of semantic encoding feature vectors of the existing product design images. Those of ordinary skill in the art should know that the FPT model is a Feature Pyramid Transformer, which can achieve cross-space and cross-scale feature extraction by combining Transformer and Feature Pyramid. Specifically, the FPT model first extracts different levels of visual information from the product design drawing, such as overall shape, local structure and texture, etc., by constructing a multi-scale feature pyramid, and then uses the self-attention mechanism of the Transformer architecture to perform global interaction of cross-scale features, so as to be able to effectively capture the key visual elements and design intentions in the image, improve the understanding depth of complex design drawings, and provide a solid foundation for subsequent semantic query matching and retrieval.

[0043] In the above customized product data management method based on the cloud platform, in step S4, on the cloud platform, semantic encoding is performed on the natural language description of the product customization requirement to obtain a semantic understanding encoding vector of the product customization requirement. That is, in order to implement the retrieval of design drawings based on context semantic information to improve the relevance and accuracy of image retrieval, the present application further uses a pre-trained language model to perform semantic encoding on the natural language description of the product customization requirement input by the user to capture the context semantics of the user's product customization requirement and convert it into a semantic understanding encoding vector of the product customization requirement in a high-dimensional semantic feature space. In a specific example of the present application, step S4 includes: inputting the natural language description of the product customization requirement into a semantic encoder based on the Bert model for semantic encoding to obtain the semantic understanding encoding vector of the product customization requirement. Those of ordinary skill in the art should know that the Bert model is based on a bidirectional Transformer encoder architecture and can perform in-depth semantic understanding and context awareness on the input text. Through pre-training on a large amount of text data, it learns to understand the meanings of words in different contexts, so as to accurately represent the user's demand intention. This semantic encoding method not only improves the relevance and accuracy of context-based design drawing retrieval but also enables the system to better match the specific needs of users. Even in the face of complex customization requirements, it can accurately parse and provide corresponding design references. Therefore, using the Bert model for semantic encoding significantly enhances the ability to understand natural language descriptions and lays a foundation for realizing efficient and intelligent product design drawing retrieval.

[0044] In the above customized product data management method based on the cloud platform, in step S5, on the cloud platform, fast location semantic query encoding is performed on the set of the semantic understanding encoding vector of the product customization requirement and the semantic encoding feature vectors of the existing product design images to obtain a semantic query response encoding vector of the product customization requirement. That is, the present application further performs semantic query matching on the set of the semantic understanding encoding vector of the product customization requirement and the semantic encoding feature vectors of the existing product design images to retrieve the existing product design drawings that match the user's customization requirements. In particular, to improve the retrieval efficiency, the present application proposes a fast location semantic query encoding method, which can quickly and accurately screen out a subset of design drawings that match the user's customization requirements by constructing an efficient index structure and an optimized query algorithm to reduce the data dimension and improve the calculation efficiency, and realize a more refined query matching process. Among them, Figure 3 is a flowchart of sub-step S5 of the customized product data management method based on the cloud platform according to an embodiment of the present application. As Figure 3As shown, step S5 includes the steps of: S51, based on the semantic differences between the semantic understanding encoding vector of the product customization requirement and each semantic encoding feature vector of the existing product design images in the set of the semantic encoding feature vectors of the existing product design images, determining the positioning center existing product design image semantic encoding feature vector of the set of the semantic encoding feature vectors of the existing product design images; S52, based on the positioning center existing product design image semantic encoding feature vector, determining a fine-grained matching search window for the existing product design images, where the vector at the center position of the fine-grained matching search window for the existing product design images is the positioning center existing product design image semantic encoding feature vector, and each semantic encoding feature vector of the existing product design images in the fine-grained matching search window for the existing product design images is defined as a fine-grained query existing product design image semantic encoding feature vector; S53, performing semantic query response encoding on the semantic understanding encoding vector of the product customization requirement and each fine-grained query existing product design image semantic encoding feature vector in the fine-grained matching search window for the existing product design images to obtain the product customization requirement query response semantic encoding vector.

[0045] Specifically, step S51, based on the semantic differences between the semantic understanding encoding vector of the product customization requirement and each semantic encoding feature vector of the existing product design images in the set of the semantic encoding feature vectors of the existing product design images, determines the positioning center existing product design image semantic encoding feature vector of the set of the semantic encoding feature vectors of the existing product design images. Among them, Figure 4 is a flowchart of sub-step S51 of the customized product data management method based on a cloud platform according to an embodiment of the present application. As Figure 4 shown, step S51 includes the steps of: S511, calculating the cross-entropy of the semantic understanding encoding vector of the product customization requirement with respect to each semantic encoding feature vector of the existing product design images in the set of the semantic encoding feature vectors of the existing product design images to obtain a set of existing product design image fast query positioning factors; S512, taking the semantic encoding feature vector of the existing product design image corresponding to the minimum value in the set of the existing product design image fast query positioning factors as the positioning center existing product design image semantic encoding feature vector.

[0046] More specifically, step S511 is represented by the formula:

[0047] D = {d1, d2,..., d i ,..., d n}

[0048]

[0049] F = {H(q, d1), H(q, d2),..., H(q, d i ),..., H(q, d n )}

[0050] where D represents the set of semantic coding feature vectors of the existing product design images, d1, d2, d i and d n respectively represent the 1st, 2nd, ith, and nth existing product design image semantic coding feature vectors in the set of the existing product design image semantic coding feature vectors, n is the number of feature vectors in the set of the existing product design image semantic coding feature vectors, represents the eigenvalue at the kth position in the ith existing product design image semantic coding feature vector, q represents the product customization requirement semantic understanding coding vector, q k represents the eigenvalue at the kth position in the product customization requirement semantic understanding coding vector, log2(·) represents the logarithmic function with base 2, H(q, d1), H(q, d2), H(q, d i ) and H(q, d n ) respectively represent the cross-entropy of q with respect to d1, d2, d i and d n , and F represents the set of the existing product design image fast query and positioning factors.

[0051] That is, based on the metric of cross-entropy, the semantic distance and mismatch degree between the semantic understanding encoding vector of the product customization requirement and the semantic encoding feature vectors of each existing product design image are quantified to generate a set of fast query and positioning factors for the existing product design images, so as to guide the search process to be closer to the target query area of the existing product design images and provide an accurate reference for subsequent fast query and positioning. It is worth mentioning that cross-entropy essentially measures the difference between two probability distributions. In the technical solution of this application, the product customization requirement encoding vector and the image encoding feature vector are normalized into probability distributions (for example, through Softmax processing), and cross-entropy can directly quantify the matching degree between the two. At the same time, the asymmetry of cross-entropy naturally adapts to the matching directionality of requirements and designs. The user requirement is the target distribution, and the existing design is the candidate distribution. Cross-entropy measures the additional information required to express the requirement with the design distribution. This directionality makes the retrieval process more in line with the actual application scenario, that is, screening designs based on requirements, while asymmetric metrics (such as cosine similarity) may ignore this logic. Additionally, in a high-dimensional feature space, the Euclidean distance is vulnerable to the curse of dimensionality, while cross-entropy captures semantic differences from the perspective of probability distributions and is more sensitive to small changes in key features. For example, if the requirement emphasizes "environmentally friendly materials", cross-entropy will significantly amplify the feature differences related to this attribute, thereby improving the retrieval accuracy.

[0052] More specifically, the step S512 is expressed by the formula:

[0053]

[0054] where arg min represents taking the semantic encoding feature vector of the existing product design image corresponding to the minimum value in the set of fast query and positioning factors for the existing product design images, and d * represents the semantic encoding feature vector of the existing product design image at the positioning center.

[0055] That is, by selecting the minimum value in the set of fast query and positioning factors for the existing product design images calculated above, the existing product design image that best matches the user requirement is determined to ensure that the semantic encoding feature vector of the existing product design image preliminarily screened has a high semantic correlation with the user requirement.

[0056] Specifically, the step S52 is expressed by the formula:

[0057]

[0058] W = {d j-m / 2 , d j-m / 2+1 ,..., d * ,..., d j+m / 2-1 , d j+m / 2}

[0059] Among them, j represents the position index of the semantic encoding feature vector of the existing product design image of the positioning center in the set of semantic encoding feature vectors of the existing product design images, ||·|| represents calculating the norm of the vector, represents rounding down, m represents the width of the fine-grained matching search window of the existing product design image, d j-m / 2 、d j-m / 2+1 、d j+m / 2-1 and d j+m / 2 are the semantic encoding feature vectors of the fine-grained query existing product design images in the fine-grained matching search window of the existing product design image, and W is the set composed of the semantic encoding feature vectors of the fine-grained query existing product design images in the fine-grained matching search window of the existing product design image.

[0060] That is, taking the semantic encoding feature vector of the existing product design image preliminarily screened above as the positioning center, a fine-grained matching search window of the existing product design image is delimited. Since the semantic encoding feature vectors of the existing product design images within the window have a small distance from the positioning center in the semantic feature space, they represent potential matching items for the user's customization requirements. Therefore, in this application, the semantic encoding feature vectors of the existing product design images within the window are further designated as the semantic encoding feature vectors of the fine-grained query existing product design images, serving as candidate objects for the next semantic interaction response encoding, and are used for more refined local context awareness and semantic query interaction. In this way, resources can be concentrated for in-depth analysis in a smaller and more targeted data subset, which not only improves the retrieval accuracy but also maintains the computational efficiency.

[0061] Specifically, in a specific example of this application, step S53 includes: performing a linear transformation on the semantic understanding encoding vector of the product customization requirement to obtain a query vector and a value vector, and using the semantic encoding feature vectors of the fine-grained query existing product design images in the fine-grained matching search window of the existing product design image as key vectors, and inputting the query vector, the value vector, and the key vectors into a fine-grained query encoding module based on a heterogeneous transformer structure to obtain the semantic encoding vector of the product customization requirement query response, which is expressed by the formula:

[0062] v q =qW q +b q

[0063] v v =qW v +b v

[0064]

[0065] Among them, W q and W v represent the query embedding matrix and the value embedding matrix respectively, b q and b v represent different bias terms respectively, v q and v v represent the query vector and the value vector respectively, d l represents the l-th fine-grained query existing product design image semantic encoding feature vector in the fine-grained matching search window of the existing product design image, represents matrix multiplication, softmax represents the softmax function, (·) T represents the transpose of a vector, S represents the feature scale value of the l-th fine-grained query existing product design image semantic encoding feature vector, v r represents the product customization requirement query response semantic encoding vector.

[0066] That is, by performing a linear transformation on the product customization requirement semantic understanding encoding vector to convert it into a query vector and a value vector suitable for the attention mechanism, and at the same time combining each key vector in the fine-grained matching search window of the existing product design image, using the fine-grained query encoding module based on the heterogeneous transformer structure to simulate the attention interaction between the query, the key, and the value, and through the multi-head self-attention layer in the fine-grained query encoding module to parallelly process the semantic association between the user customization requirement and each existing product design image, so as to realize the deep semantic interaction and fusion between the user customization requirement and each existing product design image, and generate a product customization requirement query response semantic encoding vector that synthesizes the semantic association information between the user customization requirement and each existing product design image, thereby providing a more accurate semantic context representation for the subsequent matching and retrieval of the existing product design image.

[0067] In the above-mentioned customized product data management method based on the cloud platform, in step S6, on the cloud platform, based on the product customization requirement query response semantic encoding vector, return the existing product design drawing that matches the user's customization requirement as the retrieval result. In a specific example of the present application, step S6 includes: inputting the product customization requirement query response semantic encoding vector into a search optimizer based on a classifier to obtain a search optimization result, and the search optimization result is the sequence label of the existing product design drawing; returning the existing product design drawing corresponding to the sequence label as the retrieval result.

[0068] Specifically, the classifier is based on a neural network architecture and learns features from the semantic encoding vector of the product customization requirement query response. Based on the semantic association information between the user customization requirements contained in the semantic encoding vector of the product customization requirement query response and each existing product design image, it analyzes the matching degree between each existing product design drawing and the user customization requirements and outputs a corresponding probability distribution, which represents the confidence of each existing product design drawing as a retrieval result. Then, according to the output of the classifier, the search optimizer selects the sequence label corresponding to the existing product design drawing with the highest probability as the search optimization result. Finally, through the retrieval mechanism of the cloud platform, the obtained sequence label and the corresponding existing product design drawing stored in the cloud are extracted and returned to the user as the retrieval result. In this way, designers can carry out further product design and customization based on the retrieval result, thus greatly improving the design efficiency and satisfaction.

[0069] More specifically, inputting the semantic encoding vector of the product customization requirement query response into the search optimizer based on the classifier to obtain the search optimization result, where the search optimization result is the sequence label of the existing product design drawing, including: using the fully connected layer of the search optimizer to perform a fully connected encoding on the semantic encoding vector of the product customization requirement query response to obtain the fully connected encoding vector of the product customization requirement query response semantics; inputting the fully connected encoding vector of the product customization requirement query response semantics into the Softmax classification function of the search optimizer to obtain the probability values of the semantic encoding vector of the product customization requirement query response belonging to the sequence labels of each existing product design drawing; and determining the sequence label corresponding to the largest of the probability values as the search optimization result.

[0070] In the technical solution of this application, the set of the semantic understanding encoding vector of the product customization requirement and the semantic encoding feature vector of the existing product design image respectively represents the text semantic encoding feature of the product customization requirement and the image semantic encoding feature of each existing product design drawing in the set of the existing product design drawings. In the dynamic semantic query encoding based on the fast positioning mechanism, the feature modality difference between the query feature and the query target feature will cause the long-distance dynamic semantic search encoding to be insufficient, thus reducing the expression effect of the semantic encoding vector of the product customization requirement query response and affecting the accuracy of the search optimization result obtained by inputting it into the search optimizer based on the classifier.

[0071] Based on this, in a preferred embodiment of this application, inputting the semantic encoding vector of the product customization requirement query response into the search optimizer based on the classifier to obtain the search optimization result includes:

[0072] Query the global feature mean and feature variance of the semantic encoding vector of the product customization requirement response, and perform morphological deviation modulation on the semantic encoding vector of the product customization requirement response to obtain the morphological encoding vector of the product customization requirement response deviation, which is expressed as:

[0073]

[0074] where V represents the semantic encoding vector of the product customization requirement response, and μ and σ 2 respectively represent the global feature mean and feature variance of the semantic encoding vector of the product customization requirement response, ⊙, respectively represent element-wise multiplication, element-wise addition, and element-wise subtraction by position, and V' represents the morphological encoding vector of the product customization requirement response deviation;

[0075] Based on the morphological encoding vector of the product customization requirement response deviation, construct the global structure dependence matrix of the product customization requirement response, which is expressed as:

[0076]

[0077] where T represents the transpose of the vector, represents matrix multiplication, ReLU represents the rectified linear unit function, and M represents the global structure dependence matrix of the product customization requirement response;

[0078] Calculate the transpose vector of the semantic encoding vector of the product customization requirement response, and couple its structure into the feature manifold of the global structure dependence matrix of the product customization requirement response to obtain the first structure-coupled modulation vector of the product customization requirement response, which is expressed as:

[0079]

[0080] where Softmax represents the softmax function, and V1 represents the first structure-coupled modulation vector of the product customization requirement response;

[0081] Project the semantic encoding vector of the product customization requirement response into the feature space of the global structure dependence matrix of the product customization requirement response to obtain the second structure-coupled modulation vector of the product customization requirement response, which is expressed as:

[0082]

[0083] where V2 represents the second structure-coupled modulation vector of the product customization requirement response;

[0084] Fuse the first product customization requirement query response structure coupling modulation vector and the second product customization requirement query response structure coupling modulation vector to obtain an optimized product customization requirement query response semantic encoding vector, expressed as:

[0085]

[0086] where ω1 and ω2 respectively represent the first weight hyperparameter and the second weight hyperparameter, and V o represents the optimized product customization requirement query response semantic encoding vector;

[0087] Input the optimized product customization requirement query response semantic encoding vector into the classifier-based search optimizer to obtain the search optimization result.

[0088] In this way, for the global structure dependency graph of the prior attribute statistical distribution of the product customization requirement query response semantic encoding vector under the specified feature organization scheme, due to the cross-domain link construction obstacle problem caused by the significant spacing of the local clustering threshold, the adaptive channel fusion high-dimensional descriptor of the product customization requirement query response semantic encoding vector is used to deconstruct the composite interaction architecture of its global structure dependency graph. Furthermore, by simulating and utilizing the high-dimensional descriptor of manifold embedding to reconstruct the feature distribution of the product customization requirement query response semantic encoding vector, the core mode refinement enhancement of the product customization requirement query response semantic encoding vector in the inherent dynamic trajectory is realized, enhancing the feature expression effect of the product customization requirement query response semantic encoding vector. In this way, the accuracy of the search optimization result obtained by inputting it into the classifier-based search optimizer is improved.

[0089] After obtaining the preliminary retrieval results, the retrieval results can be further screened and refined to remove those design solutions that are technically similar but actually do not meet the specific requirements of the user. For example, some design drawings may be similar in overall shape or function combination, but there are differences in material selection, color matching, or other details. To achieve this goal, more refined filtering rules can be introduced, such as the use limit of specific materials, the precise control of size ranges, etc. In addition, the secondary screening can also be carried out in combination with the additional input of the user (if provided), such as the feedback of the user on the preliminary results or further specific requirements.

[0090] Next is the interaction and confirmation stage with the user to ensure that the final design solution truly meets the user's personalized needs. To this end, the system provides an intuitive and easy-to-use interface that allows the user to view and compare the differences between different design drawings. Visualization tools can help highlight key design elements, such as shape features, structural layouts, material textures, etc., making it easier for the user to understand and evaluate each option. At the same time, the user is encouraged to provide immediate feedback, such as ratings or comments, which can not only help the user make decisions but also provide valuable data support for the continuous improvement of the system. For uncertainties, the user can be guided to express more detailed requirements through a question-and-answer form, thereby further narrowing down the selection range.

[0091] After determining the final design solution, the next step is to generate a detailed specification document. This document will serve as a guiding document for production and manufacturing, containing all necessary information, such as precise dimension data, material lists, processing technology requirements, etc. To ensure the accuracy and integrity of the document, automated tools can be used to extract relevant information from the design drawings and organize it into an easy-to-understand and executable text content in a standard format. In addition, considering the possible differences in operating habits and technical levels among different factories or suppliers, corresponding versions can also be customized for specific production environments to ensure that every participant can clearly understand and strictly execute the design intent.

[0092] After generating the detailed specifications, it enters the production preparation and coordination stage. This includes communicating and collaborating with all parties upstream and downstream of the supply chain to ensure the smooth connection of all links, such as raw material procurement, component processing, and assembly testing. To improve efficiency, a unified information sharing platform can be established to allow all stakeholders to access the latest project progress and change notifications in real time. At the same time, advanced project management software is used to track task progress and promptly identify and solve possible problems. In addition, considering the particularity of customized products, a flexible production and delivery plan also needs to be formulated to cope with possible uncertainties, such as sudden technical problems or changes in customer requirements.

[0093] In summary, the cloud platform-based customized product data management method according to the embodiments of the present application is elucidated. It first uploads the existing product design drawings to the cloud platform, and extracts image features from each existing product design drawing through a deep neural network deployed on the cloud platform to capture the key visual elements and design intents in each existing product design drawing. On this basis, the product customization requirements input by the user are further subjected to efficient semantic query matching analysis with each existing product design drawing, so as to accurately retrieve the existing product design drawings that match the user's customization requirements. In this way, it can help designers quickly locate the design drawings that meet specific requirements and significantly improve the efficiency and quality of customized product development.

[0094] Furthermore, a customized product data management system based on a cloud platform is also provided.

[0095] Figure 5 It is a block diagram of a customized product data management system based on a cloud platform according to an embodiment of the present application. As Figure 5 shown, a customized product data management system 100 based on a cloud platform according to an embodiment of the present application includes: a product customization requirement acquisition module 110, configured to acquire a natural language description of a product customization requirement input by a user; an upload and synchronization module 120, configured to upload existing product design drawings and the natural language description of the product customization requirement to the cloud platform; an image feature extraction module 130, configured to perform image feature extraction on each of the existing product design drawings in the set of the existing product design drawings on the cloud platform to obtain a set of semantic encoding feature vectors of the existing product design images; a semantic encoding module 140, configured to perform semantic encoding on the natural language description of the product customization requirement on the cloud platform to obtain a semantic understanding encoding vector of the product customization requirement; a semantic query encoding module 150, configured to perform fast positioning semantic query encoding on the semantic understanding encoding vector of the product customization requirement and the set of the semantic encoding feature vectors of the existing product design images on the cloud platform to obtain a semantic query response encoding vector of the product customization requirement; and a retrieval result return module 160, configured to return, on the cloud platform, an existing product design drawing that matches the user's customization requirement as a retrieval result based on the semantic query response encoding vector of the product customization requirement.

[0096] Here, those skilled in the art can understand that the specific operations of each module in the above-mentioned customized product data management system based on a cloud platform have been introduced in detail in the above description of the Figures 1 to 4 customized product data management method based on a cloud platform, and therefore, the repeated description thereof will be omitted.

[0097] The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, advantages, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, advantages, effects, etc. are essential for each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and easy understanding, rather than limitations, and the above details do not limit the present invention to necessarily adopt the above specific details for implementation.

[0098] In the above embodiments, the descriptions of the various embodiments each have their own emphasis. For parts not detailed or recorded in a certain embodiment, reference may be made to the relevant descriptions of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only illustrative. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0099] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claimed rights.

[0100] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0101] Finally, it should be noted that the above description has been given for purposes of illustration and description. In addition, the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A customized product data management method based on a cloud platform, characterized in that, Including: Obtain the natural language description of the product customization requirements input by the user; Upload the existing product design drawings and the natural language description of the product customization requirements to the cloud platform; On the cloud platform, perform image feature extraction on each of the existing product design drawings in the set of existing product design drawings to obtain a set of semantic encoding feature vectors of the existing product design images; On the cloud platform, perform semantic encoding on the natural language description of the product customization requirements to obtain a semantic understanding encoding vector of the product customization requirements; On the cloud platform, perform fast positioning semantic query encoding on the semantic understanding encoding vector of the product customization requirements and the set of semantic encoding feature vectors of the existing product design images to obtain a semantic encoding vector of the product customization requirements query response; On the cloud platform, based on the semantic encoding vector of the product customization requirements query response, return the existing product design drawings that match the user's customization requirements as the retrieval result.

2. The customized product data management method based on a cloud platform according to claim 1, characterized in that Performing image feature extraction on each of the existing product design drawings in the set of existing product design drawings to obtain a set of semantic encoding feature vectors of the existing product design images, including: Input each of the existing product design drawings in the set of existing product design drawings into a product design image feature extractor based on the FPT model to obtain a set of semantic encoding feature vectors of the existing product design images.

3. The customized product data management method based on a cloud platform according to claim 2, wherein Performing semantic encoding on the natural language description of the product customization requirements to obtain a semantic understanding encoding vector of the product customization requirements, including: Input the natural language description of the product customization requirements into a semantic encoder based on the Bert model for semantic encoding to obtain the semantic understanding encoding vector of the product customization requirements.

4. The customized product data management method based on a cloud platform according to claim 3, wherein Performing fast positioning semantic query encoding on the semantic understanding encoding vector of the product customization requirements and the set of semantic encoding feature vectors of the existing product design images to obtain a semantic encoding vector of the product customization requirements query response, including: Based on the semantic differences between the semantic understanding encoding vector of the product customization requirements and each semantic encoding feature vector of the existing product design images in the set of semantic encoding feature vectors of the existing product design images, determine the positioning center semantic encoding feature vector of the existing product design images in the set of semantic encoding feature vectors of the existing product design images; Based on the positioning center semantic encoding feature vector of the existing product design images, determine a fine-grained matching search window for the existing product design images. The vector at the center position of the fine-grained matching search window for the existing product design images is the positioning center semantic encoding feature vector of the existing product design images, and each semantic encoding feature vector of the existing product design images in the fine-grained matching search window for the existing product design images is defined as a fine-grained query semantic encoding feature vector of the existing product design images; Perform semantic query response encoding on the semantic understanding encoding vector of the product customization requirements and each fine-grained query semantic encoding feature vector of the existing product design images in the fine-grained matching search window for the existing product design images to obtain the semantic encoding vector of the product customization requirements query response.

5. The customized product data management method based on a cloud platform according to claim 4, characterized in that Determine the positioning center existing product design image semantic coding feature vector of the set of existing product design image semantic coding feature vectors based on the semantic differences between the semantic understanding coding vector of the product customization requirement and each existing product design image semantic coding feature vector in the set, including: Calculate the cross entropy of the semantic understanding coding vector of the product customization requirement with respect to each existing product design image semantic coding feature vector in the set of existing product design image semantic coding feature vectors to obtain a set of existing product design image quick query positioning factors; Use the existing product design image semantic coding feature vector corresponding to the minimum value in the set of existing product design image quick query positioning factors as the positioning center existing product design image semantic coding feature vector.

6. The customized product data management method based on a cloud platform according to claim 5, wherein, Perform semantic query response encoding on the semantic understanding coding vector of the product customization requirement and each fine-grained query existing product design image semantic coding feature vector in the fine-grained matching search window of the existing product design image to obtain the semantic coding vector of the product customization requirement query response, including: Perform a linear transformation on the semantic understanding coding vector of the product customization requirement to obtain a query vector and a value vector, and use each fine-grained query existing product design image semantic coding feature vector in the fine-grained matching search window of the existing product design image as a key vector, and input the query vector, the value vector, and the key vector into a fine-grained query coding module based on a heterogeneous transformer structure to obtain the semantic coding vector of the product customization requirement query response.

7. The customized product data management method based on a cloud platform according to claim 6, characterized in that Based on the semantic coding vector of the product customization requirement query response, return the existing product design drawing that matches the user's customization requirement as the retrieval result, including: Input the semantic coding vector of the product customization requirement query response into a search optimizer based on a classifier to obtain a search optimization result, where the search optimization result is a sequence label of an existing product design drawing; Return the existing product design drawing corresponding to the sequence label as the retrieval result.

8. The customized product data management method based on a cloud platform according to claim 7, characterized in that Input the semantic coding vector of the product customization requirement query response into a search optimizer based on a classifier to obtain a search optimization result, where the search optimization result is a sequence label of an existing product design drawing, including: Perform a fully connected encoding on the semantic coding vector of the product customization requirement query response using the fully connected layer of the search optimizer to obtain a fully connected encoding vector of the product customization requirement query response semantics; Input the fully connected encoding vector of the product customization requirement query response semantics into the Softmax classification function of the search optimizer to obtain the probability values of the semantic coding vector of the product customization requirement query response belonging to the sequence labels of each existing product design drawing; Determine the sequence label corresponding to the largest of the probability values as the search optimization result.

9. A customized product data management system based on a cloud platform, characterized in that, Including: A product customization requirement acquisition module for acquiring a natural language description of the product customization requirement input by the user; An upload synchronization module for uploading the existing product design drawings and the natural language description of the product customization requirement to the cloud platform; An image feature extraction module, which is used to perform image feature extraction on each existing product design drawing in the set of the existing product design drawings on the cloud platform to obtain a set of semantic encoding feature vectors of the existing product design images; A semantic encoding module, which is used to perform semantic encoding on the natural language description of the product customization requirements on the cloud platform to obtain a semantic understanding encoding vector of the product customization requirements; A semantic query encoding module, which is used to perform fast positioning semantic query encoding on the semantic understanding encoding vector of the product customization requirements and the set of semantic encoding feature vectors of the existing product design images on the cloud platform to obtain a semantic encoding vector of the product customization requirement query response; A retrieval result return module, which is used to return the existing product design drawings that match the user's customization requirements as retrieval results on the cloud platform based on the semantic encoding vector of the product customization requirement query response.

Images