Typically designed digital primitive library construction method and aided design method

Through the digital element library construction method, semantic drivers and graph neural networks are used to process typical design data and generate related information, which solves the problem of inconsistent design standards of power grid companies and improves design consistency and quality.

CN119047100APending Publication Date: 2024-11-29CHINA SOUTHERN POWER GRID COMPANY
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410966330.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

In typical design design of power grid companies, there are problems such as inconsistent design standards and insufficient refinement, which affects the quality of project implementation.

Method used

Using the digital primitive library construction method, the annotated typical design data is obtained and the semantic-driven prediction model and graph neural network model are used to generate a bidirectional index data set and association map, extract and classify association information, and form Digital primitive library that provides auxiliary design methods to improve design consistency and quality.

Benefits of technology

The data accuracy and coverage of the digital primitive library are improved, helping designers reduce design errors, improve design standards consistency and the quality of grid engineering design drawings.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119047100A_ABST
    Figure CN119047100A_ABST
Patent Text Reader

Abstract

The invention discloses a typical design digital primitive library construction method and an aided design method, and the method comprises the steps: inputting typical design data of a power grid into a splitting model for splitting, and obtaining the information of a belonging engineering field, primitives, the connection information between the primitives, and primitive marks; then, processing through a prediction model based on semantic driving to obtain a bidirectional index data set, processing through a graph neural network model to obtain a correlation graph, extracting correlation information in the correlation graph by utilizing two index directions of the bidirectional index data set, and classifying the correlation information according to the information of the engineering field to which the correlation information belongs; according to the digitized primitive library, the relation information is extracted from the association graph by utilizing the bidirectional index and then is merged into the database, so that a designer can be helped to design from a single primitive, the designer is guided by utilizing the association information between the primitives, design errors are reduced, and the consistency of design standards is improved; and the design quality of the design drawing of the power grid project is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method for constructing a digital graphic element library of a typical design and an auxiliary design method. Background Art

[0002] The standardized construction of distribution networks is a prerequisite for power grid companies to implement group operations, intensive development, and refined management. For this purpose, normative documents related to typical designs of distribution networks have been formulated with the aim of unifying construction standards and equipment specifications, facilitating operation and maintenance, facilitating equipment bidding, improving work efficiency, reducing construction and operation costs, giving play to scale advantages, and improving overall benefits.

[0003] Although the typical design of the distribution network project of the State Grid Corporation of China provides a standardized design plan, the requirements of various professional departments of the power grid company are not unified, and the requirements of different construction units and operating units of similar projects are inconsistent. Even if the design is carried out in accordance with the relevant specification documents of the typical design of the distribution network, it is difficult to avoid the problems of inconsistent design standards and insufficient design refinement, which will affect the quality of project implementation. Summary of the invention

[0004] The following is a summary of the subject matter described in detail herein. This summary is not intended to limit the scope of the claims.

[0005] The embodiments of the present application provide a method for constructing a digital graphic element library of a typical design and an auxiliary design method, which can construct a digital graphic element library of graphic elements and assist designers in designing.

[0006] In a first aspect, an embodiment of the present application provides a method for constructing a digital primitive library of a typical design, comprising:

[0007] Acquire annotated typical design data, wherein the typical design data is annotated with at least a drawing title and a drawing legend;

[0008] Input the annotated typical design data into a preset splitting model for splitting, and obtain the engineering field information corresponding to the drawing title, and obtain the graphic elements corresponding to the drawing legend, the connection information between the graphic elements, and the graphic element marks;

[0009] According to the primitive, the primitive tag and the engineering field information, a semantic-driven prediction model is inputted to output a bidirectional index data set, wherein the bidirectional index data set has a direction of pattern indexing with the primitive as a directory and a direction of character indexing with the primitive tag as a directory;

[0010] According to the connection information between the graph elements and the graph elements, the graph neural network model is input, and an association graph with the graph elements as nodes and the connection information between the graph elements as edges is output;

[0011] The bidirectional index data set and the association map are input into a fusion network for processing, so that the fusion network extracts the association information in the association map based on the directories in two directions in the bidirectional index data set, and classifies the association information according to the engineering field information to which it belongs, so as to obtain a digital graphic element library.

[0012] In some embodiments, the split model includes a text recognition module, a boundary recognition module and a primitive regeneration module; the step of inputting the annotated typical design data into the preset split model for splitting includes:

[0013] Input the annotated typical design data into the text recognition module, the boundary recognition module and the graphic element regeneration module for parallel processing, wherein the text recognition module outputs the engineering field information and the graphic element label, the boundary recognition module outputs the range of a single design drawing, and the graphic element regeneration module outputs the connection information between the graphic elements;

[0014] The scope of each single design drawing is divided according to the engineering field information, and the graphic element labels, the graphic elements and the connection information between the graphic elements are divided according to the scope of the single design drawing, wherein the graphic elements, the graphic element labels and the engineering field information obtained by division are constructed into a semi-structured data set, and the connection information between the graphic elements obtained by division is constructed into a graphic network data set.

[0015] In some embodiments, the inputting of the annotated typical design data into the text recognition module, the boundary recognition module and the primitive regeneration module for parallel processing includes:

[0016] Recognize the text of the drawing title and the text corresponding to the drawing legend of the design drawing in the typical design data by the text recognition module, match the engineering field according to the recognized text of the drawing title, and determine the graphic element mark according to the recognized text of the graphic element legend, wherein the engineering field includes the field of power transmission engineering, the field of power distribution engineering and the field of substation engineering;

[0017] Identifying the boundary area of ​​the design drawings in the typical design data by the boundary identification module to determine the scope of a single design drawing;

[0018] The drawing legend of the design drawings in the typical design data is identified through the graphic element regeneration module, the graphic element is identified based on the drawing legend and the graphic element feature value corresponding to the graphic element is determined, the graphic element in the design drawing that is identical to the graphic element is searched based on the graphic element feature value, and the connection information between the graphic elements is determined according to the connecting lines between the graphic elements.

[0019] In some embodiments, the step of inputting the primitives, the primitive tags and the engineering domain information into a semantically driven prediction model and outputting a bidirectional index dataset comprises:

[0020] Perform semantic analysis based on the engineering field information and the graphic element mark to determine the engineering field corresponding to the engineering field information and determine the graphic element type corresponding to the graphic element mark;

[0021] Convert the engineering field corresponding to the engineering field information into an identifier, and bind the identifier to the graphic element and the graphic element tag respectively;

[0022] Establishing a correspondence between the graphic element and the graphic element tag through the graphic element type;

[0023] After extracting the graphic elements according to the corresponding relationship, similar items are merged, and corresponding graphic element tags are merged according to the principle of maximum compatibility during the merging process to obtain a first data set with the graphic elements as directories for pattern indexing;

[0024] After extracting the graphic element mark according to the corresponding relationship, merge similar items, and merge corresponding graphic elements according to the principle of maximum compatibility during the merging process to obtain a second data set with the graphic element mark as a directory for character indexing;

[0025] The directory of the first data set and the directory of the second data set are adjusted and merged to obtain a bidirectional index data set.

[0026] In some embodiments, establishing a correspondence between the primitive and the primitive tag through the primitive type includes:

[0027] Generate a candidate pattern according to the semantic features of the primitive type, wherein the candidate pattern is represented by feature points;

[0028] Performing image processing on the primitive to obtain a primitive feature value, matching feature points of the candidate pattern according to the primitive feature value, and obtaining a corresponding relationship between the primitive and the primitive type;

[0029] The correspondence between the graphic element label and the graphic element is obtained according to the correspondence between the graphic element and the graphic element type.

[0030] In some embodiments, merging corresponding graphic element tags according to the principle of maximum compatibility during the merging process includes:

[0031] For two different primitive labels corresponding to the same primitive, determining the similarity of the two different primitive labels according to the feature points of the candidate patterns corresponding to the two different primitive labels;

[0032] If the similarity is lower than a first maximum compatibility threshold, two different primitive tags are not merged during the merging process, and the first maximum compatibility threshold is not lower than 50%;

[0033] The merging of corresponding primitives according to the principle of maximum compatibility during the merging process includes:

[0034] For the corresponding Figure 1 two different primitives marked by the two different primitives, and determining the similarity of the two different primitives according to the feature points of the candidate pattern corresponding to the two different primitives;

[0035] If the similarity is lower than a second maximum compatible threshold, two different primitives are not merged during the merging process, and the second maximum compatible threshold is not lower than 50%.

[0036] In some embodiments, the step of inputting the connection information between the graph elements into a graph neural network model and outputting an association graph with the graph elements as nodes and the connection information between the graph elements as edges includes:

[0037] Constructing an original matrix according to the primitives and the connection information between the primitives and inputting it into the graph neural network;

[0038] The encoder of the graph neural network model learns the mean and variance of the graph element represented in the low-dimensional vector, and the decoder of the graph neural network model performs link prediction to obtain an association graph;

[0039] Among them, the encoder is composed of multiple layers of graph convolutional layers, the decoder is used to calculate the probability that there is a connection relationship between two of the graph elements, and the loss function of the graph neural network model includes the distance information between the original matrix and the associated graph.

[0040] In some embodiments, the fusion network includes a relationship extraction module, a merging and deduplication module, and a reclassification module; the inputting the bidirectional index data set and the association map into the fusion network for processing includes:

[0041] Traversing the pattern index directory of the bidirectional index data set through the relationship extraction module, extracting first association information from the association map according to the graphic element of the pattern index directory, wherein the first association information represents a tree-like relationship diagram starting from the graphic element;

[0042] Traversing the character index directory of the bidirectional index data set through the relationship extraction module, extracting second association information from the association map according to the graphic element tag of the character index directory, wherein the second association information represents a tree-like relationship diagram starting from the graphic element corresponding to the graphic element tag;

[0043] Merge the first association information and the second association information and remove duplicates by the merging and deduplication module;

[0044] The data output by the merging and deduplication module is classified according to the engineering field information through the reclassification module to obtain the digital graphic element library.

[0045] In some embodiments, the merging the first association information and the second association information and removing duplicates by the merging and deduplication module includes:

[0046] Determine a first relationship chain of graphic element-connection relationships in the first association information, and determine a second relationship chain of graphic element-connection relationships in the second association information;

[0047] For a first relationship chain and a second relationship chain having the same starting point graphic element and the same engineering field, determining the same relationship links and different relationship links in the first relationship chain and the second relationship chain, merging the same relationship links, adding the different relationship links to the merged links, and marking the same relationship links as reliable links;

[0048] The merged relationship links are sorted out, and a directory is formed according to the starting point primitive and the primitive tag corresponding to the starting point primitive, so as to obtain a digital primitive library.

[0049] In a second aspect, an embodiment of the present application provides an auxiliary design method, which is applied to a digital primitive library obtained by the method for constructing a digital primitive library of a typical design as described above; the auxiliary design method comprises:

[0050] Establishing an index interface for the digital graphic element library, the index interface including an engineering field index item, a graphic element index item for pattern indexing, a graphic element mark index item for character indexing, and a relationship diagram display area for displaying association information obtained by indexing;

[0051] Receiving a first input key value of the engineering field index item, and preloading index data from the digital graphic primitive library into a memory according to the engineering field corresponding to the first input key value;

[0052] Receive a second input key value of the primitive index item and / or the primitive tag index item, perform indexing in the preloaded data according to the second input key value, obtain target association information, and display the target association information in the relationship diagram display area, wherein the target association information at least includes a target primitive corresponding to the second input key value and a next-level primitive connected to the target primitive;

[0053] If the next level graphic element is selected, the next level graphic element connected to the next level graphic element is displayed in the relationship diagram display area.

[0054] In some embodiments, the preloaded data is a graphic element tag of the engineering field corresponding to the first input key value; and indexing the preloaded data according to the second input key value to obtain the target association information includes:

[0055] If the second input key value is the graphic element tag input in the graphic element tag index item, searching the preloaded data according to the graphic element tag to obtain target association information starting from the graphic element tag;

[0056] If the second input key value is the primitive input in the primitive index item, image processing is performed on the primitive to obtain a primitive feature value, and the preloaded data is searched according to the primitive feature value to obtain target association information with the primitive as the starting point.

[0057] In a third aspect, an embodiment of the present application provides a controller comprising at least one processor and a memory for communicating with the at least one processor; the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the digital graphic library construction method or auxiliary design method for typical designs as described above.

[0058] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute a digital graphic library construction method or an auxiliary design method for a typical design as described above.

[0059] The method for constructing a digital primitive library of a typical design and the auxiliary design method provided in the embodiments of the present application have at least the following beneficial effects: inputting typical design data of a power grid into a splitting model for splitting, obtaining information on the engineering field, primitives, connection information between primitives, and primitive labels, and then processing the engineering field information, primitives, and primitive labels through a semantic-driven prediction model to obtain a bidirectional index data set, and processing the connection relationship between primitives through a graph neural network model to obtain an associated graph, extracting associated information from the associated graph using the two index directions of the bidirectional index data set, and classifying the associated information according to the engineering field information. The digital graphic element library can improve the data accuracy and coverage of the digital graphic element library by extracting the relationship information from the association map and merging it into the database through the bidirectional index. Designers can index according to engineering fields, graphics elements and graphics element tags through the index interface of the digital graphic element library, and can get the prompt of the related information in the relationship diagram display area of ​​the index interface, so as to help designers to design from a single graphics element, guide designers by the association information between graphics elements, reduce design errors, improve the consistency of design standards, and improve the design quality of design drawings of power grid projects.

[0060] Other features and advantages of the present application will be described in the following description, and partly become apparent from the description, or understood by practicing the present application. The purpose and other advantages of the present application can be realized and obtained by the structures specifically pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The accompanying drawings are used to provide further understanding of the technical solution of the present application and constitute a part of the specification. Together with the examples of the present application, they are used to explain the technical solution of the present application and do not constitute a limitation on the technical solution of the present application.

[0062] Figure 1 This is an overall flow chart of a method for constructing a digital primitive library provided by an embodiment of the present application;

[0063] Figure 2 It is a drawing schematic diagram of typical design data provided by an embodiment of the present application;

[0064] Figure 3 This is a flowchart of a splitting model provided in one embodiment of the present application for splitting processing;

[0065] Figure 4 It is a processing flow chart of a text recognition module, a boundary recognition module and a primitive regeneration module provided by an embodiment of the present application;

[0066] Figure 5A flowchart of obtaining a bidirectional index data set according to primitives and primitive tags provided by an embodiment of the present application;

[0067] Figure 6 This is a flow chart of establishing a correspondence between a graphic element and a graphic element tag provided by an embodiment of the present application;

[0068] Figure 7 This is a flowchart of merging graphic element tags provided by an embodiment of the present application;

[0069] Figure 8 This is a flowchart of merging graphic elements provided by an embodiment of the present application;

[0070] Fig. 9 This is a flow chart of obtaining a correlation graph through a graph neural network provided by an embodiment of the present application;

[0071] Fig.10 This is a flow chart of obtaining a digital primitive library by extracting relationships provided by an embodiment of the present application;

[0072] Fig.11 This is a flow chart of obtaining a digital primitive library based on relationship link merging provided by an embodiment of the present application;

[0073] Fig.12 is a schematic diagram of an index interface in an auxiliary design method provided by an embodiment of the present application;

[0074] Fig.13 This is a flowchart of searching and displaying on an index interface provided by an embodiment of the present application;

[0075] Fig.14 This is a flowchart of searching according to primitives or primitive tags provided by an embodiment of the present application;

[0076] Fig.15 It is a schematic diagram of the connection structure of a controller provided in one embodiment of the present application. DETAILED DESCRIPTION

[0077] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. In addition, the characteristics, operations or features described in the specification can be combined in any appropriate manner to form various implementation methods. At the same time, the steps or actions in the method description can also be replaced or adjusted in order in a manner that is obvious to those skilled in the art. Therefore, the various sequences in the specification and the accompanying drawings are only for the purpose of clearly describing a certain embodiment and are not meant to be a necessary sequence, unless otherwise specified that a certain sequence must be followed.

[0078] In the description of this application, "several" means one or more, "more" means more than two, "greater than", "less than", "exceed", etc. are understood to exclude the number itself, and "above", "below", "within", etc. are understood to include the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0079] The serial numbers of the components in this document, such as "first", "second", etc., are only used to distinguish the objects described and do not have any order or technical meaning. The "connection" and "coupling" mentioned in this application, unless otherwise specified, include direct and indirect connections (couplings).

[0080] The embodiment of the present application provides a method for constructing a digital graphic element library of a typical design and an auxiliary design method, wherein typical design data of a power grid is input into a splitting model for splitting to obtain information on the engineering field, graphic elements, connection information between graphic elements, and graphic element tags, and then the engineering field information, graphic elements, and graphic element tags are processed by a semantic-driven prediction model to obtain a bidirectional index data set, and the connection relationship between graphic elements is processed by a graph neural network model to obtain an association map, and the two index directions of the bidirectional index data set are used to extract the association information in the association map, and the association information is classified according to the engineering field information to obtain a digital graphic element library. Since the digital graphic element library uses a bidirectional index to extract the relationship information from the association map and then merge it into a database, the data accuracy and coverage of the digital graphic element library can be improved. Designers can index according to the engineering field, graphic elements, and graphic element tags through the index interface of the digital graphic element library, and can get prompts of the association information in the relationship diagram display area of ​​the index interface, thereby helping designers to design from a single graphic element, and using the association information between graphic elements to guide designers, reduce design errors, improve the consistency of design standards, and improve the design quality of design drawings of power grid projects.

[0081] Reference Figure 1 As shown, the embodiment of the present application provides a method for constructing a digital primitive library of a typical design, including but not limited to the following steps:

[0082] Step S110, obtaining annotated typical design data, where the typical design data is annotated with at least a drawing title and a drawing legend;

[0083] The State Grid Corporation of China provides typical design information, including specification documents and typical design drawings; the specification documents involve areas such as distribution substations and overhead lines. The specification documents record various design specifications, load specifications, technical conditions, etc., and some of the contents are given in the form of diagrams, such as various engineering drawings (refer to Figure 2 As shown), the specification document describes the engineering drawings by annotation, so the corresponding drawing title (the drawing title is the file name of the specification document and / or the illustration of the engineering drawing) and the drawing legend (the drawing legend is the component annotation and / or dimension annotation in the engineering drawing) can be obtained in the specification document; the typical design drawing is an electronic file of a specific engineering drawing, including a CAD drawing or a three-dimensional drawing, etc. The typical design drawing includes one or more construction drawings, and the construction drawings are marked with the engineering drawing name and the design details and legends of each engineering component, so the corresponding drawing title (the drawing title is the file name and / or engineering drawing name of the typical design drawing) and the drawing legend (the drawing legend is the component name and / or size of the engineering component in the typical design drawing, etc.) can be obtained in the typical design drawing. In order to enable the subsequent split model to simplify the recognition process, quickly complete the splitting and have a high splitting accuracy, the typical design data are annotated, the above-mentioned drawing title and drawing legend are annotated, and the annotated typical design data are input into the splitting model as input data.

[0084] Step S120, inputting the annotated typical design data into a preset splitting model for splitting, obtaining the engineering field information corresponding to the drawing title, and obtaining the graphic elements corresponding to the drawing legend, the connection information between the graphic elements, and the graphic element marks;

[0085] The annotated typical design data can be quickly split in the splitting model, which can be pre-trained and used to identify the drawing titles and drawing legends in the annotated typical design data, wherein the drawing title is identified to obtain the corresponding engineering field information, which is mainly used to divide the engineering field to which the specification document or drawing of the current typical design belongs, namely, the power transmission engineering field, the power distribution engineering field and the substation engineering field, etc., and the drawing legend is identified to obtain the corresponding graphic element, the connection information between the graphic elements and the graphic element mark. The graphic element exists in the form of an image file in the embodiment of the present application, such as the cross-sectional view of the drainage pipe, the single-phase socket, the distribution box, etc. The connection information between the graphic elements represents the relationship between the graphic elements, for example, a connected power cable is arranged between the graphic element of the distribution box and the graphic element of the socket, and another graphic element of the drainage pipe or a concrete graphic element is arranged near the graphic element of the drainage pipe, etc. The graphic element mark represents the text description of the graphic element, which is mainly obtained in the form of an engineering drawing legend, indicating the name of the current graphic element and the design numerical standard, etc. For example, the graphic element of the drainage pipe is marked as "drain pipe" and the inner diameter / outer diameter of the drainage pipe, etc. By splitting the model, we can obtain the engineering field information, primitives, connection information between primitives and primitive labels, which are convenient for subsequent processing in other models.

[0086] Step S130, according to the information of the primitive, the primitive tag and the engineering field to which it belongs, input into the semantic-driven prediction model, outputting a bidirectional index data set, wherein the bidirectional index data set has a direction of pattern indexing with the primitive as the directory and a direction of character indexing with the primitive tag as the directory;

[0087] Step S140, inputting the connection information between the graph elements into the graph neural network model, and outputting an association graph with the graph elements as nodes and the connection information between the graph elements as edges;

[0088] In order to quickly complete the construction of the digital primitive library, the engineering domain information, primitives, connection information between primitives and primitive labels obtained in step S120 are divided into two parts, one part is primitives, primitive labels and engineering domain information, which are processed by the semantic-driven prediction model in step S130 to obtain a bidirectional index data set, and the other part is the connection information between primitives, which is processed by the graph neural network in step S140. Among them, the semantic-driven prediction model organizes primitive labels and engineering domain information according to primitives as directories, and organizes primitives and engineering domain information according to primitive labels as directories, thereby obtaining a bidirectional index data set, which allows indexing by primitives and indexing by primitive labels, which can facilitate designers to search according to primitives or primitive labels, can strengthen the retrieval entry function of the digital primitive library constructed subsequently, and can also improve the fusion effect between the associated graphs output by the subsequent graph neural network; the graph neural network model is suitable for processing the relationship between primitives, and the connection relationship between primitives can be effectively processed through the graph neural network model, so as to obtain the associated graph. Therefore, the above steps S130 and S140 are to process the data after splitting by the splitting model respectively, and arrange the processing with the semantic-driven prediction model and the graph neural network model according to the type of data, which can avoid the problem of using a single model to process multiple types of data and increase the calculation complexity and long calculation time.

[0089] Step S150, input the bidirectional index data set and the association map into the fusion network for processing, so that the fusion network extracts the association information in the association map based on the two-directional directories in the bidirectional index data set, and classifies the association information according to the engineering field information to which it belongs, to obtain a digital graphic element library.

[0090] The fusion network is used to fuse two parts of data, that is, to fuse the bidirectional index data set and the association map. The fusion process is based on the directory of the two directions in the bidirectional index data set. According to the primitives and primitive tags in the directory, the association information of the primitives and the association information of the primitive tags are extracted from the association map, so that the association information similar to the tree can be formed. Finally, the engineering field information is attached to these association information for classification, and a digital primitive library can be obtained. The obtained digital primitive library includes primitives, primitive tags, and the connection relationship between primitives, and is divided according to the engineering field. When searching in the digital primitive library, the search scope is initially narrowed by the engineering field, and then the connection relationship between primitives can be retrieved by further inputting primitives or primitive tags, so as to provide guidance information for engineering design to designers, assisting designers to design drawings according to the standard direction of typical design, reduce the probability of design errors, and improve design speed.

[0091] Reference Figure 3 As shown, in some embodiments, the splitting model includes a text recognition module, a boundary recognition module and a primitive regeneration module; the annotated typical design data is input into the preset splitting model for splitting, including:

[0092] Step S210, inputting the annotated typical design data into the text recognition module, the boundary recognition module and the graphic element regeneration module for parallel processing, wherein the text recognition module outputs the engineering field information and the graphic element mark, the boundary recognition module outputs the range of a single design drawing, and the graphic element regeneration module outputs the connection information between graphic elements;

[0093] Step S220, divide the scope of each single design drawing according to the engineering field information to which it belongs, and divide the graphic element labels and the connection information between graphic elements according to the scope of the single design drawing, wherein the divided graphic elements, graphic element labels and the engineering field information to which they belong are constructed into a semi-structured data set, and the divided connection information between graphic elements is constructed into a graphic network data set.

[0094] The split model consists of three main modules, namely the text recognition module, the boundary recognition module and the primitive regeneration module. The three modules process the input typical design data in parallel and obtain corresponding information, namely the engineering field information, primitive tags, and the connection information between primitives. Since the typical design data have commonalities, they can be processed separately by the three modules. The output results of each module are for all drawings in the typical design data. The output results of the three modules are independent of each other in data type but are related. Therefore, these data can be classified and combined and input into the semantic-driven prediction model and graph neural network model for processing.

[0095] Among them, the text recognition module can integrate character extraction and optical character recognition functions. For the text in vector files such as CAD drawings and editable text in PDF, the text in these files can be obtained through the character extraction function. For the text in three-dimensional drawings and bitmap text in PDF, the text can be recognized through the optical character recognition function. Since the drawing titles and drawing legends in typical design materials are pre-marked, the text recognition module can directly process the marked drawing titles and drawing legends to obtain the engineering field information and element tags. The boundary recognition module mainly provides line recognition function, which is mainly used to identify the boundaries of design drawings in CAD drawings. When there are multiple design drawings in a single CAD drawing, each design drawing is relatively independent. In order to clearly define the connection information between elements, the boundary recognition module needs to determine the scope of a single design drawing, which is used to divide the connection information between elements output by the element regeneration module. The graphic element regeneration module is mainly responsible for image recognition and recognition of connection lines between images. When processing CAD drawings, it searches for similar images / vector graphics in the design drawings according to the drawing legend, and then identifies connection lines based on the found images / vector graphics. When processing PDF materials, it identifies the drawings in the materials according to the drawing legend, divides the graphics elements in the drawings and determines the connection relationship between the graphics elements. After integration, it can output the connection information between the graphics elements.

[0096] After the output of the three modules, the three output results are further processed, and the primitive labels, primitives and connection information between primitives are divided according to the scope of each single design drawing. Among them, primitives, primitive labels and engineering field information are used to input into the semantic-driven prediction model. These three types of data constitute a semi-structured data set, that is, primitives are unstructured data, and primitive labels and engineering field information are structured data; the connection information between primitives is used to input into the graph neural network model, and is provided to the graph neural network model as nodes and edges for processing.

[0097] Reference Figure 4 As shown, in some embodiments, the annotated typical design data is input into a text recognition module, a boundary recognition module and a primitive regeneration module for parallel processing, including:

[0098] Step S310, identifying the text of the drawing title and the text corresponding to the drawing legend of the design drawing in the typical design data through a text recognition module, matching the engineering field according to the recognized text of the drawing title, and determining the graphic element mark according to the recognized text of the graphic element legend, the engineering field includes the power transmission engineering field, the power distribution engineering field and the substation engineering field;

[0099] Step S320, identifying the boundary area of ​​the design drawings in the typical design data through the boundary recognition module to determine the range of a single design drawing;

[0100] Step S330, identifying the drawing legend of the design drawings in the typical design data through the graphic element regeneration module, obtaining the graphic element according to the drawing legend and determining the graphic element feature value corresponding to the graphic element, searching for the graphic element in the design drawing that is identical to the graphic element according to the graphic element feature value, and determining the connection information between the graphic elements according to the connection lines between the graphic elements.

[0101] Drawing titles usually contain information about the engineering field, such as "10KV distribution network engineering power line laying diagram". The text recognition module can mine drawing title information through NLP (Natural Language Processing) and directly identify which engineering field the key text in the drawing title belongs to, or identify the key text and then use the key text to match the words of the three preset engineering fields. For drawing legends, the text is directly recognized from the drawing legend to obtain the graphic element tag, which is the character extraction and optical character recognition function mentioned above.

[0102] The boundary recognition module recognizes large boundaries in CAD drawings. Since the boundary of a single design drawing is a regular rectangle, the scope of a single design drawing can be well identified according to the distance and shape between boundaries.

[0103] The primitive regeneration module identifies the connection relationship between primitives in the above manner, and mainly searches for similar primitives in the design drawing based on the annotated drawing legend. Since the drawing legend has been annotated (the annotated legend is the primitive, for example, the pattern in the legend of the distribution box is the primitive of the distribution box), the primitive regeneration module processes the annotated drawing legend to generate primitive feature values, and then uses the primitive feature values ​​to search for similar primitives, so as to determine all the primitives in the drawing, and then obtain the connection relationship between the primitives through the connecting lines between the primitives.

[0104] Reference Figure 5 As shown, in some embodiments, according to the input of the primitives, primitive tags and the engineering field information to which they belong, into the semantic-driven prediction model, a bidirectional index data set is output, including:

[0105] Step S410, performing semantic analysis based on the engineering field information and the graphic element tag, determining the engineering field corresponding to the engineering field information, and determining the graphic element type corresponding to the graphic element tag;

[0106] Step S420, converting the engineering field corresponding to the engineering field information into an identifier, and binding the identifier to the graphic element and the graphic element tag respectively;

[0107] Step S430, establishing a correspondence between the primitive and the primitive tag through the primitive type;

[0108] Step S440, extracting primitives according to the corresponding relationship and merging similar items, and merging corresponding primitive tags according to the principle of maximum compatibility during the merging process to obtain a first data set with primitives as directories for pattern indexing;

[0109] Step S450, after extracting the graphic element tags according to the corresponding relationship, similar items are merged, and in the merging process, corresponding graphic elements are merged according to the principle of maximum compatibility to obtain a second data set with the graphic element tags as the directory for character indexing.

[0110] Step S460: adjust and merge the directory of the first data set and the directory of the second data set to obtain a bidirectional index data set.

[0111] The primitives, primitive tags and engineering domain information obtained through splitting the splitting model form a semi-structured data set and are input into the semantic-driven prediction model. The prediction model again performs semantic analysis on the engineering domain information to determine the engineering domain corresponding to the engineering domain information. The prediction model also performs semantic analysis on the primitive tags to learn the primitive types of primitive tags. For example, the primitive of "distribution box" belongs to the primitive type of "distribution cabinet", which should have a large number of connecting lines. For example, the primitive of "load switch" belongs to the primitive type of "switch", which should have two or more connecting lines.

[0112] Afterwards, the information of the engineering field is converted into identifiers according to the division of engineering fields. As we know, the engineering fields are divided into three types, namely, power transmission engineering field, power distribution engineering field and substation engineering field. Therefore, three identifiers can be set accordingly. Then, the corresponding identifiers are bound to the graphics elements and graphics element tags. In the subsequent processing process, the identifiers always follow the graphics elements and graphics element tags.

[0113] After the corresponding relationship between the primitive and the primitive tag is established through the primitive type, in the merging process, it is divided into two parts for merging, one part takes the primitive as the directory, merges the primitive tag, and obtains the first data set, and the other part takes the primitive tag as the directory, merges the primitive, and obtains the second data set; in order to avoid merging errors, when merging with the primitive as the directory, the dissimilar primitive tags are retained to the maximum extent, and when merging with the primitive tag as the directory, the dissimilar primitives are retained to the maximum extent, so that the bidirectional index database can have the maximum compatibility. Then merge the first data set and the second data set to form a bidirectional index data set with two directories. The bidirectional index of this application refers to the two directions from the primitive-primitive tag and from the primitive tag-primitive, by retrieving the primitive, the primitive tag can be retrieved, and the primitive and the primitive tag of the bidirectional index data set can be accompanied by the engineering field information. This bidirectional index data set can facilitate designers to search according to the primitive tag or primitive, can strengthen the retrieval entry function of the digital primitive library constructed later, and can also improve the fusion effect between the associated graphs output by the subsequent graph neural network.

[0114] Reference Figure 6 As shown, in some embodiments, a correspondence is established between a primitive and a primitive tag through a primitive type, including:

[0115] Step S510, generating candidate patterns according to the semantic features of the primitive type, wherein the candidate patterns are represented by feature points;

[0116] Step S520, performing image processing on the primitive to obtain a primitive feature value, matching feature points of the candidate pattern according to the primitive feature value, and obtaining a correspondence between the primitive and the primitive type;

[0117] Step S530 , obtaining the correspondence between the graphic element label and the graphic element according to the correspondence between the graphic element and the graphic element type.

[0118] The primitives exist in the bidirectional index data set in the form of images. In the process of establishing the correspondence between primitives and primitive tags, multiple candidate patterns are generated according to the semantic features of the primitive tags to match the primitives existing in the form of images; for example, for the primitive tag of "distribution box", multiple candidate patterns related to "distribution box" are generated. The candidate patterns should be line drawings (generated based on matching primitives in CAD and primitives in materials such as PDF), and then the feature points of the candidate patterns are calculated. It can be understood that the candidate pattern generation module can be pre-trained and suitable for text-to-image type neural networks. Of course, the neural network for text-to-image generation here needs to be trained to generate line drawings. This type of text-to-image function is relatively mature and will not be explained in detail here.

[0119] On the other hand, primitive feature values ​​are also generated for primitives, and then the primitive feature values ​​are matched with feature points of candidate patterns. The successfully matched primitives and candidate patterns establish a correspondence between primitives and primitive types, and then the correspondence between primitive labels and primitives is obtained based on the correspondence between primitives and primitive types.

[0120] Reference Figure 7 As shown, in some embodiments, corresponding primitive tags are merged according to the principle of maximum compatibility during the merging process, including:

[0121] Step S610, for two different primitive labels corresponding to the same primitive, determine the similarity of the two different primitive labels according to the feature points of the candidate patterns corresponding to the two different primitive labels;

[0122] Step S620, if the similarity is lower than a first maximum compatibility threshold, two different primitive tags are not merged in the merging process, and the first maximum compatibility threshold is not lower than 50%;

[0123] Reference Figure 8 As shown, in the merging process, corresponding primitives are merged according to the principle of maximum compatibility, including:

[0124] Step S630, for the corresponding Figure 1 Two different primitives marked by the primitives are determined, and the similarity of the two different primitives is determined according to the feature points of the candidate patterns corresponding to the two different primitives;

[0125] Step S640: if the similarity is lower than a second maximum compatible threshold, two different primitives are not merged in the merging process, and the second maximum compatible threshold is not lower than 50%.

[0126] There are multiple candidate patterns, so there are two situations when merging: the pattern can be merged and the pattern cannot be merged. At this time, there may be two different primitive tags corresponding to the same primitive, or there may be two different primitives corresponding to the same primitive tag. In the above steps S610 and S620, the similarity threshold of the primitive tag, that is, the first maximum compatibility threshold, is set, the primitive type of the two primitive tags is considered, and then the candidate pattern corresponding to the primitive type is derived. The candidate pattern corresponds to the primitive tag, and the similarity of the feature points of the two candidate patterns is compared, which is equivalent to comparing the similarity of the two primitive tags (the primitive tags are text, and the similarity cannot be actually obtained, so the similarity needs to be determined by the feature points of the candidate patterns). If the similarity is high, the two primitive tags are merged into the corresponding primitive (one of the two primitive tags is selected), and if the similarity is low, the two primitive tags are not merged. In order to adapt to the above-mentioned maximum compatibility principle, the first maximum compatibility threshold setting here can be set to be larger, and none of them will be merged if it is lower than this first maximum compatibility threshold. Similarly, in the above steps S630 and S640, the similarity threshold of the primitives, i.e., the second maximum compatibility threshold, is set, and the candidate patterns corresponding to the two primitives are considered, and the similarity of the feature points of the two candidate patterns is compared as the similarity of the two primitives (the two primitives can also be directly compared to obtain the similarity). If the similarity is high, the two primitives are merged into the corresponding primitive mark (one of the two primitives is the standard), and if the similarity is low, the two primitives are not merged. In order to adapt to the above maximum compatibility principle, the second maximum compatibility threshold setting here can be set larger, and no merging will be done below this second maximum compatibility threshold.

[0127] Reference Fig. 9 As shown, in some embodiments, according to the connection information between the graph elements, the graph neural network model is inputted, and the association graph with the graph elements as nodes and the connection information between the graph elements as edges is outputted, including:

[0128] Step S710, constructing an original matrix based on the connection information between the graph elements and inputting it into the graph neural network;

[0129] Step S720, learning the mean and variance of the graph element represented in the low-dimensional vector through the encoder of the graph neural network model, and performing link prediction through the decoder of the graph neural network model to obtain an associated graph;

[0130] Among them, the encoder is composed of multiple layers of graph convolutional layers, the decoder is used to calculate the probability of a connection relationship between two graph elements, and the loss function of the graph neural network model includes the distance information between the original matrix and the associated graph.

[0131] In the graph neural network model, the original matrix is ​​processed by the encoder. The original matrix consists of the connection information between the graph elements and the graph elements, which can be transformed into the adjacency matrix and the feature matrix of the node. After the encoder processes the original matrix, it enters the decoder to process the probability of the connection relationship between the two graph elements, and the corresponding loss function is used to optimize the training process. In addition to the distance information between the original matrix and the associated graph, the loss function can also include the node representation vector distribution and the normal distribution KL divergence. It can be seen that the graph neural network model of this embodiment adopts the variational graph autoencoder (Var iationa lAuto-Encoder), in which the encoder can only use two layers of GCN (Graph Convo l ut ionNetworks, graph convolution network).

[0132] Reference Fig.10 As shown, in some embodiments, the fusion network includes a relationship extraction module, a merging and deduplication module, and a reclassification module; the bidirectional index data set and the association map are input into the fusion network for processing, including:

[0133] Step S810, traversing the pattern index directory of the bidirectional index data set through a relationship extraction module, extracting first association information from the association map according to the graphic element of the pattern index directory, wherein the first association information represents a tree-like relationship map starting from the graphic element;

[0134] Step S820, traversing the character index directory of the bidirectional index data set through the relationship extraction module, extracting second association information from the association map according to the graphic element label of the character index directory, wherein the second association information represents a tree-like relationship graph starting from the graphic element corresponding to the graphic element label;

[0135] Step S830, merging the first association information and the second association information and removing duplicates through a merging and deduplication module;

[0136] Step S840: The data output by the merging and deduplication module is classified according to the engineering field information through the reclassification module to obtain a digital primitive library.

[0137] For the bidirectional index data set, the directory in two directions is traversed, and the associated information is extracted from the associated map for each item in the directory, that is, for each element of the bidirectional index data set, the first associated information is extracted from the associated map to obtain multiple tree-like relationship diagrams starting from the element, and for each element tag of the bidirectional index data set, the second associated information is extracted from the associated map to obtain multiple tree-like relationship diagrams starting from the element corresponding to the element tag. Since the bidirectional index data set is traversed from two directions, there is a possibility of duplication in the first associated information and the second associated information extracted from the associated map, so in step S830, the duplicate items are removed by merging the deduplication module, and then the remaining associated information is classified according to the engineering field information to which it belongs, to obtain a digital element library. As can be seen from the above, the engineering field information to which it belongs is bound to the element and the element tag through the identifier, so the associated information extracted from the associated map through the bidirectional index data set can be classified according to the bound engineering field information. It is understandable that although the association information is extracted from the association graph to obtain a digital graph library, the form of the association information is actually still the matrix form output by the graph neural network model, and it is actually a simple tree relationship diagram, but it can be regarded as a tree relationship diagram starting from a certain node.

[0138] Reference Fig.11 As shown, in some embodiments, merging the first association information and the second association information and removing duplicates by a merging and deduplication module includes:

[0139] Step S910, determining a first relationship chain of a graphic element-connection relationship in the first association information, and determining a second relationship chain of a graphic element-connection relationship in the second association information;

[0140] Step S920, for the first relationship chain and the second relationship chain having the same starting point graphic element and the same engineering field, determining the same relationship links and the different relationship links in the first relationship chain and the second relationship chain, merging the same relationship links, adding the different relationship links to the merged links, and marking the same relationship links as reliable links;

[0141] Step S930, sorting out the merged relationship links, and forming a directory according to the starting point primitive and the primitive mark corresponding to the starting point primitive, to obtain a digital primitive library.

[0142] The first association information starts from the primitive, and the second association information starts from the primitive corresponding to the primitive tag. Therefore, the relationship links in the first association information and the second association information are both in the form of primitive-connection relationship-primitive-connection relationship-primitive. Therefore, the first association information and the second association information can be directly merged, but repeated links need to be considered. In this embodiment, a simpler comparison method is adopted. For two relationship chains with the same starting point, the same relationship links are merged, and different relationship links are directly attached to the same relationship links as tree branches. In order to improve the confidence, the merged same relationship links are set as reliable links. This is a link in which the first association information and the second association information both give the same relationship, which has a higher confidence.

[0143] In fact, the graph element / graph element tag of each node in the merged relationship link may become a search object. Therefore, in order to provide search for users, a directory is formed for the merged relationship link. The directory records the graph elements and graph element tags, and the directory is used to provide search keywords.

[0144] Reference Fig.12 and Fig.13 As shown, the embodiment of the present application further provides an auxiliary design method, which is applied to the digital primitive library obtained by the digital primitive library construction method of the typical design of the above embodiment, and the auxiliary design method includes:

[0145] Step S1010, establishing an index interface for the digital primitive library, the index interface including an engineering field index item, a primitive index item for pattern indexing, a primitive mark index item for character indexing, and a relationship diagram display area for displaying association information obtained by indexing;

[0146] Step S1020, receiving a first input key value of an engineering field index item, and preloading index data from a digital graphic primitive library into a memory according to the engineering field corresponding to the first input key value;

[0147] Step S1030, receiving a second input key value of the primitive index item and / or the primitive tag index item, indexing the preloaded data according to the second input key value, obtaining target association information, and displaying the target association information in the relationship diagram display area, wherein the target association information at least includes a target primitive corresponding to the second input key value and a next-level primitive connected to the target primitive;

[0148] Step S1030: If the next level graphic element is selected, the next level graphic element connected to the next level graphic element is displayed in the relationship diagram display area.

[0149] The index interface includes an area for inputting key values ​​and an area for displaying a relationship diagram. The area for inputting key values ​​includes an engineering field index item, an element index item, and an element mark index item. The area for displaying a relationship diagram is a relationship diagram display area. In order to speed up the retrieval speed, the index interface guides the designer to first input the engineering field in the engineering field index item, i.e., the first input key value, and preload the index data into the memory according to the engineering field. These preloaded data are the data of the engineering field corresponding to the identifiers of the bound elements and element marks. Preloading can alleviate the problem of too large data volume and too slow retrieval speed in the digital element library. Afterwards, the second input key value of the index is received in the element index item and / or the element mark index item, and then the preloaded data is searched according to the second input key value to obtain the target association information with the element / element mark corresponding to the second input key value as the starting point, and the target element and the next level element of the target element of the target association information are displayed in the relationship diagram display area.

[0150] Since the target association information is complex, in order to allow designers to view the connection relationships between more graphics elements, the relationship diagram display area is dynamic. When the designer clicks / selects the next level of the target graphics element, the next level of the next level of the graphics element is displayed, so that the designer can click in the tree relationship diagram to view the tree branch structure of the association information. It is understandable that since the association information is multi-branched, the designer can only select one of the nodes (graphic elements) in multiple branches at a time, and then display multiple nodes (graphic elements) of the next level. In order to avoid the content displayed in the relationship diagram display area becoming complicated after clicking on the node (graphic element) multiple times, the nodes (graphic elements) of the first N levels can be set to be automatically hidden.

[0151] In addition, engineering field index items, graphic element index items and graphic element tag index items can be made into the form of drop-down menu options or input boxes; more practically, the engineering field index items use a drop-down menu to provide three engineering field options, the graphic element index items use an input box + import picture button to allow designers to fuzzy query through the input box and import picture query, and the graphic element tag index items use an input box + drop-down menu to allow designers to fuzzy query through the input box and manually select.

[0152] Reference Fig.14 As shown, in some embodiments, the preloaded data is a graphic element tag of the engineering field corresponding to the first input key value; the preloaded data is indexed according to the second input key value to obtain the target association information, including:

[0153] Step S1110, if the second input key value is the graphic element tag input in the graphic element tag index item, searching the preloaded data according to the graphic element tag to obtain target association information starting from the graphic element tag;

[0154] Step S1120, if the second input key value is the primitive input in the primitive index item, image processing is performed based on the primitive to obtain the primitive feature value, and the preloaded data is searched according to the primitive feature value to obtain the target association information starting from the primitive.

[0155] It can be known from the above that the second input key value can be a primitive or a primitive tag. If searching is performed according to the primitive tag, after the designer inputs the text, searching is performed in the preloaded data according to the text, and the target association information starting from the primitive corresponding to the primitive tag is output; if searching is performed according to the primitive, after the designer imports the primitive, the primitive feature value is obtained, searching is performed in the preloaded data according to the primitive feature value, and the target association information starting from the primitive corresponding to the primitive feature value is output.

[0156] An embodiment of the present application also provides a controller, comprising at least one processor and a memory for communicating with the at least one processor; the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the aforementioned digital graphic library construction method or auxiliary design method for typical designs.

[0157] Reference Fig.15 As shown, the control processor 1001 and the memory 1002 in the controller 1000 can be connected via a bus as an example. The memory 1002, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory 1002 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk memory, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory 1002 may optionally include a memory remotely arranged relative to the control processor 1001, and these remote memories may be connected to the controller 1000 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0158] Those skilled in the art will understand that Fig.15 The device structure shown in the figure does not constitute a limitation on the controller 1000, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0159] The embodiment of the present application also provides a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are executed by one or more control processors, for example, Fig.15One of the control processors 1001 executes, which can enable the above-mentioned one or more control processors to execute the digital graphic element library construction method or auxiliary design method of the typical design in the above-mentioned method embodiment.

[0160] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0161] It will be appreciated by those skilled in the art that all or some of the steps and systems in the disclosed method above may be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that may be used to store desired information and may be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0162] The above is a specific description of the preferred implementation of the present application, but the present application is not limited to the above-mentioned implementation mode. Technical personnel familiar with the field can also make various equivalent deformations or substitutions without violating the spirit of the present application. These equivalent deformations or substitutions are all included in the scope defined by the claims of the present application.

Claims

1. A method for constructing a digital graphic library of typical designs. It is characterized in that include: Acquire annotated typical design data, wherein the typical design data is annotated with at least a drawing title and a drawing legend; Input the annotated typical design data into a preset splitting model for splitting, and obtain the engineering field information corresponding to the drawing title, and obtain the graphic elements corresponding to the drawing legend, the connection information between the graphic elements, and the graphic element marks; According to the primitive, the primitive tag and the engineering field information, a semantic-driven prediction model is inputted to output a bidirectional index data set, wherein the bidirectional index data set has a direction of pattern indexing with the primitive as a directory and a direction of character indexing with the primitive tag as a directory; According to the connection information between the graph elements and the graph elements, the graph neural network model is input, and an association graph with the graph elements as nodes and the connection information between the graph elements as edges is output; The bidirectional index data set and the association map are input into a fusion network for processing, so that the fusion network extracts the association information in the association map based on the directories in two directions in the bidirectional index data set, and classifies the association information according to the engineering field information to which it belongs, so as to obtain a digital graphic element library.

2. The method according to claim 1, It is characterized in that The splitting model includes a text recognition module, a boundary recognition module and a primitive regeneration module; the inputting of the annotated typical design data into the preset splitting model for splitting includes: Input the annotated typical design data into the text recognition module, the boundary recognition module and the graphic element regeneration module for parallel processing, wherein the text recognition module outputs the engineering field information and the graphic element label, the boundary recognition module outputs the range of a single design drawing, and the graphic element regeneration module outputs the connection information between the graphic elements; The scope of each single design drawing is divided according to the engineering field information, and the graphic element labels, the graphic elements and the connection information between the graphic elements are divided according to the scope of the single design drawing, wherein the graphic elements, the graphic element labels and the engineering field information obtained by division are constructed into a semi-structured data set, and the connection information between the graphic elements obtained by division is constructed into a graphic network data set.

3. The method according to claim 2, It is characterized in that The step of inputting the annotated typical design data into the text recognition module, the boundary recognition module and the primitive regeneration module for parallel processing includes: Recognize the text of the drawing title and the text corresponding to the drawing legend of the design drawing in the typical design data by the text recognition module, match the engineering field according to the recognized text of the drawing title, and determine the graphic element mark according to the recognized text of the graphic element legend, wherein the engineering field includes the field of power transmission engineering, the field of power distribution engineering and the field of substation engineering; Identifying the boundary area of ​​the design drawings in the typical design data by the boundary identification module to determine the scope of a single design drawing; The drawing legend of the design drawings in the typical design data is identified through the graphic element regeneration module, the graphic element is identified based on the drawing legend and the graphic element feature value corresponding to the graphic element is determined, the graphic element in the design drawing that is identical to the graphic element is searched based on the graphic element feature value, and the connection information between the graphic elements is determined according to the connecting lines between the graphic elements.

4. The method according to claim 1, It is characterized in that The step of inputting the primitives, the primitive tags and the engineering domain information into a semantically driven prediction model and outputting a bidirectional index data set includes: Perform semantic analysis based on the engineering field information and the graphic element mark to determine the engineering field corresponding to the engineering field information and determine the graphic element type corresponding to the graphic element mark; Convert the engineering field corresponding to the engineering field information into an identifier, and bind the identifier to the graphic element and the graphic element tag respectively; Establishing a correspondence between the graphic element and the graphic element tag through the graphic element type; After extracting the graphic elements according to the corresponding relationship, similar items are merged, and corresponding graphic element tags are merged according to the principle of maximum compatibility during the merging process to obtain a first data set with the graphic elements as directories for pattern indexing; After extracting the graphic element mark according to the corresponding relationship, merge similar items, and merge corresponding graphic elements according to the principle of maximum compatibility during the merging process to obtain a second data set with the graphic element mark as a directory for character indexing; The directory of the first data set and the directory of the second data set are adjusted and merged to obtain a bidirectional index data set.

5. The method according to claim 4, It is characterized in that The establishing a correspondence between the graphic element and the graphic element tag through the graphic element type includes: Generate a candidate pattern according to the semantic features of the primitive type, wherein the candidate pattern is represented by feature points; Performing image processing on the primitive to obtain a primitive feature value, matching feature points of the candidate pattern according to the primitive feature value, and obtaining a corresponding relationship between the primitive and the primitive type; The correspondence between the graphic element label and the graphic element is obtained according to the correspondence between the graphic element and the graphic element type.

6. The method according to claim 5, It is characterized in that The merging of corresponding graphic element tags according to the principle of maximum compatibility during the merging process includes: For two different primitive labels corresponding to the same primitive, determining the similarity of the two different primitive labels according to the feature points of the candidate patterns corresponding to the two different primitive labels; If the similarity is lower than a first maximum compatibility threshold, two different primitive tags are not merged during the merging process, and the first maximum compatibility threshold is not lower than 50%; The merging of corresponding primitives according to the principle of maximum compatibility during the merging process includes: For two different primitives of a corresponding primitive label, determining the similarity of the two different primitives according to the feature points of the candidate pattern corresponding to the two different primitives; If the similarity is lower than a second maximum compatible threshold, two different primitives are not merged during the merging process, and the second maximum compatible threshold is not lower than 50%.

7. The method according to claim 1, It is characterized in that The step of inputting the connection information between the graph elements into a graph neural network model and outputting an association graph with the graph elements as nodes and the connection information between the graph elements as edges includes: Constructing an original matrix according to the primitives and the connection information between the primitives and inputting it into the graph neural network; The encoder of the graph neural network model learns the mean and variance of the graph element represented in the low-dimensional vector, and the decoder of the graph neural network model performs link prediction to obtain an association graph; Among them, the encoder is composed of multiple layers of graph convolutional layers, the decoder is used to calculate the probability that there is a connection relationship between two of the graph elements, and the loss function of the graph neural network model includes the distance information between the original matrix and the associated graph.

8. The method according to claim 1, It is characterized in that The fusion network includes a relationship extraction module, a merging and deduplication module, and a reclassification module; The step of inputting the bidirectional index data set and the association graph into a fusion network for processing comprises: Traversing the pattern index directory of the bidirectional index data set through the relationship extraction module, extracting first association information from the association map according to the graphic element of the pattern index directory, wherein the first association information represents a tree-like relationship diagram starting from the graphic element; Traversing the character index directory of the bidirectional index data set through the relationship extraction module, extracting second association information from the association map according to the graphic element tag of the character index directory, wherein the second association information represents a tree-like relationship diagram starting from the graphic element corresponding to the graphic element tag; Merge the first association information and the second association information and remove duplicates by the merging and deduplication module; The data output by the merging and deduplication module is classified according to the engineering field information through the reclassification module to obtain the digital graphic element library.

9. The method according to claim 8, It is characterized in that The merging the first association information and the second association information and removing duplicates by the merging and deduplication module includes: Determine a first relationship chain of graphic element-connection relationships in the first association information, and determine a second relationship chain of graphic element-connection relationships in the second association information; For a first relationship chain and a second relationship chain having the same starting point graphic element and the same engineering field, determining the same relationship links and different relationship links in the first relationship chain and the second relationship chain, merging the same relationship links, adding the different relationship links to the merged links, and marking the same relationship links as reliable links; The merged relationship links are sorted out, and a directory is formed according to the starting point primitive and the primitive tag corresponding to the starting point primitive, so as to obtain a digital primitive library.

10. An auxiliary design method, It is characterized in that A digital graphic primitive library obtained by the method for constructing a digital graphic primitive library of a typical design as claimed in any one of claims 1 to 9; the auxiliary design method comprises: Establishing an index interface for the digital graphic element library, the index interface including an engineering field index item, a graphic element index item for pattern indexing, a graphic element mark index item for character indexing, and a relationship diagram display area for displaying association information obtained by indexing; Receiving a first input key value of the engineering field index item, and preloading index data from the digital graphic primitive library into a memory according to the engineering field corresponding to the first input key value; Receive a second input key value of the primitive index item and / or the primitive tag index item, perform indexing in the preloaded data according to the second input key value, obtain target association information, and display the target association information in the relationship diagram display area, wherein the target association information at least includes a target primitive corresponding to the second input key value and a next-level primitive connected to the target primitive; If the next level graphic element is selected, the next level graphic element connected to the next level graphic element is displayed in the relationship diagram display area.

11. The auxiliary design method according to claim 10, It is characterized in that The preloaded data is a graphic element tag of the engineering field corresponding to the first input key value; The step of indexing the preloaded data according to the second input key value to obtain target association information includes: If the second input key value is the graphic element tag input in the graphic element tag index item, searching the preloaded data according to the graphic element tag to obtain target association information starting from the graphic element tag; If the second input key value is the primitive input in the primitive index item, image processing is performed on the primitive to obtain a primitive feature value, and the preloaded data is searched according to the primitive feature value to obtain target association information with the primitive as the starting point.

Citation Information

Cited By

  • Image-based data matching method, electronic device, storage medium and program

    CN121582616A