AI collaborative building drawing auxiliary review method and system

Abstract

The application discloses an AI-assisted building drawing review method and system, and relates to the technical field of drawing digitization processing.The method comprises the following steps: inputting a target building drawing into an AI review module of an intelligent review system, filling in drawing information to trigger a floor table creation instruction, then extracting drawing information from a floor list obtained through iteration and identification to obtain a floor drawing information set, and after verification and passing, obtaining an AI review result set and feeding back and checking the AI review result set.The application solves the technical problems of low drawing information extraction efficiency, difficulty in adapting non-standardized drawings, lack of standardization and automation in data processing, and incomplete review data and insufficient accuracy in traditional building drawing review, achieves efficient building drawing information extraction, adaptability of non-standardized drawings, and standardization and automation of data processing, and makes the building drawing review data more comprehensive and accurate, and improves the accuracy and collaboration of the review.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of digital processing of drawings, in particular to an AI-assisted building drawing review method and system. BACKGROUND

[0002] Building drawing review is a key link for controlling the quality of building engineering, and its accuracy directly affects the safety and compliance of the project. In the prior art, drawing review mainly relies on manual comparison with specifications for page-by-page checking, or is assisted by primary tools, but manual sorting of floor drawings and extraction of component information are required. These methods can work in small projects, but as the complexity of building projects increases, the traditional review methods have limitations in data processing: manual processing of drawing information is low in efficiency and easy to miss, and primary tools are difficult to adapt to non-standardized drawing formats, resulting in incomplete data extraction and insufficient accuracy, which cannot meet the standardized and automated data processing requirements of building drawing review and is difficult to support subsequent accurate review and quality control. SUMMARY

[0003] The application provides an AI-assisted building drawing review method and system, which solves the technical problems of low efficiency of drawing information extraction, difficulty in adapting to non-standardized drawings, and lack of standardization and automation in data processing in traditional building drawing review, resulting in incomplete and inaccurate review data.

[0004] In a first aspect, the application provides an AI-assisted building drawing review method, which comprises: inputting a target building drawing into an AI review module of an intelligent review system, and filling in drawing information according to a preset project category, and triggering a floor table creation instruction when the drawing information is obtained; based on the floor table creation instruction, automatically identifying the drawing frame of the target building drawing according to the drawing information, and constructing a floor list, wherein each floor corresponds to a floor drawing; traversing the floor list to extract the drawing information template or non-template of the corresponding floor drawing, and obtaining a floor drawing information set; from the dimensions of accuracy and integrity, traversing the floor drawing information set to check the quality of the drawing information, and if the check is passed, triggering a review instruction, and performing AI review on the floor drawing information set through the AI review module to obtain an AI review result set; feeding back the AI review result set to a drawing designer for self-review annotation and drawing modification, submitting the modified target modified building drawing and the AI self-review result set containing the self-review annotation to a design collaboration system, and checking the target modified building drawing and the AI self-review result set containing the self-review annotation by a reviewer.

[0005] In a second aspect of the present application, an AI-assisted building drawing review system is provided, which comprises: a floor table creation instruction acquisition module, configured to input a target building drawing into an AI review module of an intelligent review system, fill in drawing information according to a preset project category, and trigger a floor table creation instruction when the drawing information is obtained; a floor list construction module, configured to perform automatic frame recognition on the target building drawing according to the drawing information based on the floor table creation instruction, and construct a floor list, wherein each floor corresponds to a floor drawing; a floor drawing information set acquisition module, configured to perform template-based extraction or non-template-based extraction of drawing information of corresponding floor drawings by traversing the floor list, and obtain a floor drawing information set; an AI review result set acquisition module, configured to perform quality checking of graphic element information by traversing the floor drawing information set from two dimensions of accuracy and integrity, trigger a review instruction if the checking is passed, perform AI review on the floor drawing information set by the AI review module, and obtain an AI review result set; and an AI review result set feedback module, configured to feed back the AI review result set to a drawing designer for self-review annotation and drawing modification, submit the modified target modified building drawing and the AI self-review result set containing the self-review annotation to a design collaboration system, and check the target modified building drawing and the AI self-review result set containing the self-review annotation by a checker.

[0006] The one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0007] In the present application, the target building drawing is input into the AI review module and the information is filled in according to the preset project category to trigger the floor table creation instruction, the floor list is constructed through automatic frame recognition, the floor drawing information set is obtained through template-based and non-template-based dual-track extraction, the information quality is checked from the dimensions of accuracy and integrity, the checking is passed, the AI review module is used for review and the result set is output, and the review process is connected with the design collaboration system by relying on the supporting system, so that the multi-dimensional review of building drawing fire protection, accessibility and professional consistency is accurately completed, the building drawing review is more efficient, accurate and meets the collaborative needs, the technical effects of efficient building drawing information extraction, adaptability of non-standardized drawings and standardization and automation of data processing are achieved, the building drawing review data is more comprehensive and accurate, and the accuracy and collaboration of the review are improved. BRIEF DESCRIPTION OF DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0009] Figure 1 is a flow diagram of an AI-assisted building drawing review method provided by an embodiment of the present application.

[0010] Figure 2 is a structural diagram of an AI-assisted building drawing review system provided by an embodiment of the present application.

[0011] Legend: floor table creation instruction acquisition module 1, floor list construction module 2, floor drawing information set acquisition module 3, AI review result set acquisition module 4, AI review result set feedback module 5. DETAILED DESCRIPTION

[0012] The present application provides an AI-assisted building drawing review method and system, which is used to solve the technical problems of low efficiency of drawing information extraction, difficulty of adapting non-standardized drawings, lack of standardization and automation of data processing in traditional building drawing review, and resulting in incomplete and inaccurate review data.

[0013] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0014] It should be noted that the terms "first", "second", etc. in the specification and the above drawings of the present application are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0015] Embodiment one, as shown in a kind of AI-assisted building drawing review method, wherein the method comprises: Figure 1

[0016] Step A100: input the target building drawing into the AI review module of the intelligent review system, and fill in the drawing information according to the preset project category, and trigger the floor table creation instruction when the drawing information is obtained.

[0017] ​In the embodiments of the present application, the AI review module is the core functional module of the intelligent review system, which is composed of a neural network model and is mainly used for receiving incoming target architectural drawings. After the system completes the floor drawing information set extraction and quality verification, the review work is carried out according to the triggered review instruction.

[0018] Specifically, first, the skilled person needs to input the target architectural drawings into the AI review module of the intelligent review system to ensure that the AI review module can obtain the source file of the drawings to be reviewed. After the target architectural drawings are input, the intelligent review system will guide the technical personnel to fill in the drawing information according to the preset project categories, which include sub-item names, professional names, engineering categories, building types, fire resistance grades, total building areas, above-ground building areas, underground building areas, whether to use an automatic fire extinguishing system, and selection of review specifications. These categories are set according to the core needs of architectural drawing review. For example, the selection of professional names will determine the subsequent review logic of building professionals or water supply and drainage professionals, and the determination of fire resistance grades will affect the application judgment of fire-related specification clauses.

[0019] Next, during the filling of drawing information, the technical personnel need to supplement each item according to the actual situation of the target architectural project to ensure the authenticity and integrity of the information. Taking the professional name as an example, the technical personnel need to select the corresponding category in the building, water supply and drainage, etc. options corresponding to the target drawing; the building type needs to distinguish between residential buildings, office buildings, industrial buildings, underground garages, etc.; the fire resistance grade needs to be selected from grade one, grade two, grade three, and grade four according to the design of the project; whether to use an automatic fire extinguishing system needs to be selected as yes or no according to whether the project is equipped with the system, and this information will be directly related to the judgment standard of the specification requirements such as fire compartment area. When selecting the review specification, the intelligent review system will provide multiple existing valid specifications such as the Architectural Design Fire Protection Specification, the Fire Water Supply and Fire Hydrant System Technical Specification, and the Residential Design Specification. The technical personnel can select the required specification according to the actual project needs, or select the full selection self-review list item to ensure that the review covers all key specification clauses related to the project.

[0020] Finally, when the user completes the filling of drawing information for all preset project categories and confirms that there is no error, the intelligent review system will preliminarily verify the filled information to confirm that complete and valid drawing information has been obtained. After the system confirms that the information is complete and valid, it will trigger the floor table creation instruction to prepare for the next step of building the floor list.

[0021] The process of first transmitting the target building drawing into the AI review module, then completely filling in the drawing information according to the preset project category, and triggering the floor table creation instruction after the information is confirmed, ensures that the basic data of the subsequent review process is accurate and meets the review requirements, and achieves the effect of making the building drawing assisted review have standardization and pertinence from the initial stage, and providing a reliable premise for subsequent floor list construction and drawing information extraction.

[0022] Step A200: Based on the floor table creation instruction, automatically identifying the drawing frame of the target building drawing according to the drawing information, and constructing a floor list, wherein each floor corresponds to a floor drawing.

[0023] Optionally, first, a preset floor table template containing drawing name, floor number, floor height, floor elevation and floor type is obtained, then a pattern analysis intelligent agent scans and detects the drawing frame of the target building drawing according to the drawing information to obtain an image quantity and an image detection attribute set, and finally a floor list is created by combining the set and the preset floor table template. The specific steps are described in detail in A210-A230.

[0024] Step A300: Iterating through the floor list to perform drawing information template extraction or non-template extraction of the corresponding floor drawing, and obtaining a floor drawing information set.

[0025] In an embodiment of the present application, first, a drawing feature extractor is called to extract features of each floor drawing to obtain a floor drawing feature set, then similarity matching is performed between the floor drawing feature set and template features in a drawing information template library, floor drawings with a matching result greater than or equal to a preset threshold are classified into a template extraction set, and floor drawings with a matching result less than the threshold are classified into a non-template extraction set, finally, drawing information is extracted from the two sets to obtain a floor drawing information set. The specific steps are described in detail in A310-A340.

[0026] Step A400: Combining the preset specification review rules, iterating through the floor drawing information set to perform graph element information quality verification from the accuracy and integrity dimensions, if the verification passes, triggering a review instruction, performing AI review on the floor drawing information set through the AI review module, and obtaining an AI review result set.

[0027] Specifically, first, combining the preset specification review rules, from the accuracy and integrity dimensions, first, a preset verification index is obtained, then the accuracy verification index is used to compare the floor drawing information with the preset specification review rules to obtain an information accuracy verification result, then the preset integrity verification index is used to verify the information coverage and cross-professional consistency to obtain an information integrity verification result, both of which meet the requirements, then the verification passes. The specific steps are described in detail in A410-A440.

[0028] When the information quality check passes, the intelligent review system triggers the review instruction, and the AI review module immediately starts the AI review of the floor drawing information set, which includes specification review, intersection review, and design basis review. Due to the limited space of the specification, the review points of specification review and intersection review are as follows:

[0029] In the specification review section, in the building professional field, the AI review module will extract the fire compartment area data in the drawing, such as a high-rise office building drawing marking the fire compartment area as 400 . Comparing with the requirement in the building design fire protection specification that the fire compartment area of high-rise public building should not be greater than 300 , it identifies the over-limit problem; at the same time, it checks the front room area. If the front room area of residential building is marked as 5 , it does not comply with the regulation that the front room area of residential building should not be less than 6 . In the water supply and drainage professional field, the AI review module extracts the fire hydrant arrangement place, such as the toilet of commercial building floor. If it is not arranged with fire hydrant, it does not comply with the specification requirement of fire hydrant arrangement place; it can also calculate the fire hydrant spacing. If the fire hydrant spacing is marked as 35m in the drawing, which exceeds the limit value of not more than 30m in the specification, it is determined as a problem item.

[0030] In the intersection review, the building professional compares the plane information, such as the wall column in the building drawing is located between axis 1-2, and the shear wall in the corresponding position of the structure drawing is also located between axis 1-2 and has the same size, which indicates that the plane consistency is consistent; it checks the building opening and structure opening. If the size of building door opening is 1.0m×2.0m, and the size of structure reserved opening is 1.05m×2.05m, the construction and decoration allowance should be considered, and the position and size matching are determined. In the water supply and drainage professional field, the AI review module compares the water supply and drainage pipeline arrangement drawing and the civil structure plane drawing. If the drainage pipeline is arranged directly above the beam body without avoiding space, it is determined that the plane position conflicts; it compares the bottom distance. If the bottom of the drainage pipeline is marked as 150mm from the bottom of the structure plate, and the thickness of the structure plate is 100mm, it verifies whether it meets the installation requirements. If the space is insufficient, it is marked as a problem.

[0031] In the design basis review process, the AI review module extracts the specification name and number marked in the drawing, such as marking the technical specification for fire water supply and fire hydrant system. The intelligent review system checks that the specification is currently valid and the name and number are accurate, and determines that the design basis is compliant; if the marked specification version is abolished, it is marked as an error.

[0032] Finally, the AI review module analyzes each detail of the floor drawing information set according to the logic of specification, intersection, and design basis. The items that meet the requirements are marked as passed, and the items that do not meet the requirements are marked as corresponding problem points, such as fire compartment over-limit, professional intersection conflict, etc. Finally, all the results are summarized to form the AI review result set.

[0033] The AI review module conducts comprehensive intelligent review on the floor drawing information set that passes the information quality verification according to the requirements of specification review, intersection review, design basis review, etc., so as to quickly and accurately identify the specification inconsistencies, professional intersection problems and design basis errors in the drawings, and provide clear problem orientation for subsequent work.

[0034] Step A500: The AI review result set is fed back to the drawing designer for self-approval annotation and drawing modification. The modified target modified building drawing and the AI self-review result set containing the self-approval annotation are submitted to the design collaboration system, and the target modified building drawing and the AI self-review result set containing the self-approval annotation are checked by the reviewer.

[0035] In the embodiment of the application, the design collaboration system is a comprehensive information management platform for the engineering design field, aiming to improve design efficiency, ensure design quality and realize full-process collaborative management.

[0036] Specifically, after the AI review module generates the AI review result set, it is fed back to the drawing designer in real time. The drawing designer refers to the result set, and carries out self-review on the drawing. In the corresponding position of the drawing or the special annotation interface, the cause of the problem, the modification idea to be adopted and other self-review annotation contents are marked. Then, according to these annotations and the prompts of the AI review, the building drawing is modified to obtain the target modified building drawing.

[0037] Subsequently, the drawing designer submits the modified target modified building drawing and the AI self-review result set containing the self-approval annotation to the design collaboration system. The review record integrating the AI preliminary review conclusion and the designer's self-review opinion. After the reviewer enters the system, the modification of the target modified building drawing is checked whether it is reasonable and complete, and whether the designer's annotation in the AI self-review result set containing the self-approval annotation responds to the AI review problem and whether the problem forms an effective closed loop, so as to ensure that the drawing modification meets the specification requirements and design intention.

[0038] Further, the step A100 in the method provided in the embodiment of the application comprises:

[0039] A110: The preset project category includes sub-item name, professional name, engineering category, building type, fire resistance rating, total building area, above-ground building area, underground building area, whether to use automatic fire extinguishing system and selected review specification.

[0040] Specifically, after the target architectural drawings are input into the AI review module of the intelligent review system, to ensure that the subsequent AI review can accurately match the actual needs of the project, the system will guide the technical personnel in the field to complete the drawing information filling according to the preset project category. The preset project category covers sub-item name, professional name, engineering category, building type, fire resistance grade, total building area, above-ground building area, underground building area, whether to use an automatic fire extinguishing system, and selection of review standards.

[0041] First, fill in the sub-item name to clearly specify the specific project unit to which the drawings to be reviewed belong, avoiding confusion between different project drawings; then select the professional name to determine the professional field on which the review focuses, such as building or water supply and drainage, which is directly related to the review rule library called by the system subsequently; then determine the engineering category and building type, for example, the engineering category is selected as residential building, and the building type is further refined as two high-rise buildings, which defines the applicable building category standard for review through these two pieces of information; then fill in the fire resistance grade, which is selected from first, second, third, and fourth levels according to the project design, and this grade is an important prerequisite for judging the compliance of fire-related components, such as the grade of fireproof doors and the area of fireproof partitions, which are all review items that need to be determined in combination with the fire resistance grade.

[0042] Then supplement the total building area, above-ground building area, and underground building area, which will directly affect the determination of review content such as fire partition division and evacuation distance calculation, for example, there are differences in the maximum allowable area of fire partitions corresponding to different building areas; then determine whether to use an automatic fire extinguishing system, if yes, the system will adjust some review standards according to the specifications, such as expanding the area of fire partitions as required; finally, select the review standards, the system will provide a list of currently effective standards, such as the Building Design Fire Protection Specification and the Fire Water Supply and Fire Hydrant System Technical Specification, and the technical personnel can select the relevant standards according to the actual needs of the project, or select all the self-review items to ensure that the review covers all key specification clauses related to the project.

[0043] During the filling process, the intelligent review system will preliminarily prompt the completeness and reasonableness of each piece of information, for example, if the sub-item name or professional name is not filled in, the system will remind the user to supplement the required information; if the building area and building type do not match, such as the total building area of a two high-rise residential building far exceeds the normal range, the system will prompt the user to check the accuracy of the data; after completing the information filling of all the preset project categories, the system confirms that the information is complete and reasonable, laying a foundation for subsequent floor table creation instructions.

[0044] By guiding the technical personnel to gradually complete the drawing information filling according to the preset project category and performing preliminary verification, it ensures that the AI review module obtains highly adaptive basic data for the project, achieving the effect of providing standardized and targeted data support for subsequent floor table creation and AI accurate review.

[0045] Further, the method provided by the embodiment of the present application comprises the following steps A200:

[0046] A210: obtaining a preset floor table template, wherein the preset floor table template comprises a drawing name, a floor number, a floor height, a floor elevation, and a floor type.

[0047] A220: performing frame scanning and detection on the target architectural drawing according to the drawing information by a schema parsing agent to obtain an image number and an image detection attribute set.

[0048] A230: creating a floor table according to the image number and the image detection attribute set and the preset floor table template to obtain a floor list.

[0049] In the embodiment of the present application, the schema parsing agent is a functional component in the intelligent review system for realizing automatic identification of the frame of the architectural drawing.

[0050] Optionally, after receiving the floor table creation instruction, in order to ensure that the subsequently constructed floor list has a unified information dimension and a standard format, a preset floor table template is first obtained. The template includes five core fields of a drawing name, a floor number, a floor height, a floor elevation, and a floor type. These fields are set by the person skilled in the art according to the basic requirements of floor information management in the review of the architectural drawing. The drawing name is used to clearly identify the identification of a single floor drawing. The floor number is used to distinguish the order of different floors. The floor height and the floor elevation provide key data for subsequent evacuation distance calculation and component height compliance review. The floor type includes a standard floor, an underground floor, a roof floor, and the like, which can assist the intelligent review system in matching the review specification of the corresponding type of floor to avoid data gaps in the subsequent review link due to the lack of information dimension.

[0051] Then, the intelligent review system calls the schema parsing agent to perform frame scanning and detection on the target architectural drawing according to the drawing information filled in step A100. The core role of the schema parsing agent is to accurately identify the frame that defines the range of a single drawing in the drawing according to the project features in the drawing information, so as to avoid misidentifying the non-frame area as valid drawing. During the scanning and detection process, the schema parsing agent will traverse the target architectural drawing page by page, count the total number of valid frames, i.e., the image number, and extract the key attributes in each frame, such as the drawing number, the floor identification, and the drawing scale, to form an image detection attribute set. Taking a 17-story residential project as an example, the schema parsing agent scans and detects 18 valid frames according to the sub-item name of the residential and the professional name of the building, which correspond to the ground floors 1-17 and the roof floor, respectively, and extracts the drawing numbers such as one-story plan and roof floor plan and the floor identifications such as 1F and roof in each frame to form an image detection attribute set containing 18 attribute items.

[0052] After that, according to the number of images and the set of image detection attributes obtained by the mode analysis agent, and in combination with the field requirements of the preset floor table template, the floor list creation is carried out. In the specific process, the intelligent review system will match the image detection attributes of each frame with the field requirements of the preset floor table template one by one: fill the floor plan in the frame in the above example into the drawing name field, fill 1F into the floor number field, and then fill the floor height filled in the drawing information into the floor height field, calculate and fill the relative elevation of the outdoor terrace into the floor elevation field, and determine the floor type as a standard floor on the ground according to the attributes of the floor plan. If the frame corresponding to the image quantity has no need for review, such as duplicate standby drawings, the floor entry corresponding to the frame can be deleted; if there are multiple same floors, such as 4-17 floors being the same standard floor, the specific number can be input in the copy floor number and the upward addition can be selected to complete the supplement of the floor entry. Finally, a floor list corresponding to one floor drawing for each floor is formed, such as the 17-story residential project mentioned above, which will form a floor list containing 18 entries, covering all floor drawings from the first floor to the roof floor.

[0053] Through the process of first obtaining the preset floor table template containing the key field, then scanning and detecting the frame by the mode analysis agent to obtain the number of images and the set of attributes according to the drawing information, and finally creating the floor list in combination with the template and the detection result, the standardized sorting and management of floor drawings are realized, and the effect of providing clear and standardized data basis for subsequent floor drawing information extraction is achieved.

[0054] Further, the step A300 in the method provided by the embodiment of the present application comprises:

[0055] A310: calling a drawing feature extractor to perform feature extraction on the floor drawing of each floor in the floor list, and obtaining a set of floor drawing features.

[0056] A320: respectively performing similarity matching between the set of floor drawing features and the template features of each drawing information template in the drawing information template library, and if the matching result is greater than or equal to a preset similarity threshold, associating the corresponding floor drawing with the matching drawing information template and adding it into a set of template extraction floor drawings.

[0057] A330: if the matching result is less than the preset similarity threshold, adding the corresponding floor drawing into a set of non-template extraction floor drawings.

[0058] A340: performing drawing information extraction according to the set of template extraction floor drawings and the set of non-template extraction floor drawings, and obtaining a set of floor drawing information.

[0059] In the embodiments of the present application, the drawing feature extractor is a functional component in the intelligent review system for extracting key features of floor drawings. The drawing information template library is a database for storing standardized drawing templates.

[0060] Specifically, the drawing feature extractor is first called to extract features of the floor drawings one by one in the floor list, wherein the core function of the drawing feature extractor is to identify and extract key component features and layout information related to the review from the floor drawings, including but not limited to the position and size of walls and columns, the distribution and spacing of axes, the identification and specifications of doors and windows, the direction and pipe diameter of water supply and drainage pipes, the arrangement position of fire hydrants or sprinkler heads, etc. These features are the core basis for subsequent judgment of whether the drawings conform to the template and can be standardized extracted.

[0061] In the extraction process, the drawing feature extractor will perform full-image scanning and feature analysis on each floor drawing according to the component recognition rules preset by those skilled in the art, such as identifying walls and columns based on geometric features and identifying pipes based on layer attributes, and finally aggregate the extraction results of all floor drawings to form a floor drawing feature set containing key features of each drawing. For example, the floor list of a certain residential project includes a one-story plan, standard layer plans from the second to the seventeenth floor, and a roof layer plan. The drawing feature extractor will extract the size of the entrance door and the wall thickness of the living room area in the one-story plan, the specifications of the bedroom doors and windows and the axis spacing of the stairwell in the standard layer plans from the second to the seventeenth floor, and the position of the fire water tank in the roof layer plan, and finally integrate them into a floor drawing feature set covering the features of 18 drawings.

[0062] Next, the obtained floor drawing feature set is matched with the template features of each drawing information template stored in the drawing information template library one by one. The drawing information template library pre-stores standardized drawing templates corresponding to different building types, different specialties, and different floor types, such as standard layer plan templates for residential buildings and water supply and drainage drawing templates for underground garages. Each template contains standard features that should be included in this type of drawing, such as the template feature that the entrance door in the standard layer template for residential buildings is a Class B fire door and the axis spacing of the stairwell is not less than 2.4 meters.

[0063] During the matching process, the intelligent review system calculates the coincidence degree, i.e. the similarity, of the features of each drawing in the floor drawing feature set and the template features of each template in the template library, and compares the calculation result with the preset similarity threshold. During the similarity calculation process, the intelligent review system first performs standardization processing on the features of a single drawing in the floor drawing feature set and the template features of a single template in the drawing information template library, and uniformly converts the features of both into a structured data format, such as the coordinate parameters, specification attributes, and relative position relationships of components, covering key review features such as wall column position, axis spacing, door and window specifications, and fire hydrant arrangement.

[0064] Subsequently, comparison is carried out one by one according to the unified feature dimension, first, it is judged whether the core component type of the drawing and the template matches, such as the template contains the fireproof door feature, it is judged whether the fireproof door component is extracted from the drawing, then it is verified whether the key attribute of the matched component conforms, such as whether the fireproof door grade is consistent with the preset B-grade standard of the template, whether the axis spacing is within the reasonable range specified by the template, and whether the relative position relationship between components fits, such as whether the wall column corresponds to the axis position marked by the template, whether the fire hydrant is located near the evacuation passage required by the template. Then, according to the importance of different features, different weights are given, such as the weight of strong feature such as fire division and evacuation door is higher than that of ordinary component size feature, the matching results of each dimension are weighted and calculated, and the overall coincidence degree, i.e. the similarity, of the two is obtained, finally, the similarity result is compared with the preset similarity threshold, and it is judged whether the drawing is suitable for the corresponding template.

[0065] If the features of a floor drawing have a similarity with the template features of a certain drawing information template greater than or equal to the preset similarity threshold, for example, the preset threshold is 80%, it is determined that the floor drawing meets the requirements of the standardized template, it is associated with the matched drawing information template, and is added to the template extraction floor drawing set, and subsequent standardized information extraction can be directly based on the template; if the features of a floor drawing have a similarity with all template features in the template library less than the preset similarity threshold, for example, the similarity of the features of a certain roof floor plan with the standard floor template is only 65% because it contains special components such as fire water tank and ventilation room, it is determined that the floor drawing is in a non-standardized format, it is added to the non-template extraction floor drawing set separately, and subsequent non-standardized extraction method needs to be used.

[0066] Finally, according to the obtained template extraction floor drawing set and non-template extraction floor drawing set, the former is template extraction to obtain a first floor drawing information set, and the latter is non-template extraction to obtain a second floor drawing information set, and then the two sets are combined to obtain a floor drawing information set, and the specific steps are described in detail in A341-A343.

[0067] Further, the method provided in the embodiment of the application comprises the following steps A340:

[0068] A341: Template extraction is performed on the template extraction floor drawing set to obtain a first floor drawing information set.

[0069] A342: Non-template extraction is performed on the non-template extraction floor drawing set to obtain a second floor drawing information set.

[0070] A343: aggregating the first and second floor plan information sets to obtain a floor plan information set.

[0071] Specifically, after completing the templated and non-templated classification of floor plans, the templated extraction operation is first carried out for the templated floor plan set. Since the floor plans in this set have been associated with the corresponding templates in the template library, the extraction process will strictly follow the information dimensions and extraction rules preset by the matching templates. The templates explicitly indicate the component information to be extracted, such as wall column position and size, door and window specifications and quantity, axis spacing, fire hydrant arrangement position, etc.; layer attributes, such as wall line layers of building specialty, pipe layers of water and drainage specialty; and data format. The intelligent review system will automatically associate the corresponding layers and components in the drawing, accurately extract the required information and format it according to the format.

[0072] Taking the second to seventeenth floor standard layer plan of a residential project as an example, which is classified into the templated extraction floor plan set and matches the residential building standard layer plan template, the intelligent review system will automatically extract the information such as entrance door size, stair shaft axis spacing, fire door grade, and fire hydrant protection distance for each floor according to the template, and arrange these information according to the floor number to form the first floor plan information set containing the key review data of the second to seventeenth floor standard layer. At the same time, during the extraction process, the system will refer to the operation logic of the extraction drawing information link to allow technical personnel to check the extraction results and confirm whether the axis, wall column and other graphic element information is accurate. If there is a small amount of deviation, manual adjustment can be made to ensure the integrity and accuracy of the first floor plan information set.

[0073] Next, the non-templated extraction floor plan set is traversed to obtain a preliminary graphic element vector set based on the preliminary graphic element identification and vectorization identification. Based on this set, two types of prompt sets are obtained through associated guidance prompt retrieval and experience guidance prompt retrieval. The drawing template generator is called to generate a new drawing template set based on this. After the layer accuracy verification of the set is passed, the second floor plan information set is extracted and obtained. The specific steps are detailed in A342-1-A342-4.

[0074] Finally, the first floor plan information set and the second floor plan information set are integrated. The intelligent review system performs consistency checking on the information dimensions of the two sets to ensure that the common information in the two sets, such as floor number and plan name, is uniformly formatted, and to avoid data conflicts. Then, the information of the two sets is merged in order from the underground layer to the roof layer, and then sorted by floor number to form a complete data set covering all floors of the project, i.e., the floor plan information set. During the integration process, the system automatically removes duplicate information, such as accidental duplicate component annotations in different sets, and supplements missing associated data, such as matching the location of the underground water collection well with the water supply and drainage pipeline layout information, to ensure that the final floor plan information set can comprehensively cover the data required for building, water supply and drainage, and other professional reviews.

[0075] By the process of extracting the first set from the templated drawings according to the preset template, extracting the second set from the non-templated drawings after adapting to the newly added template, and then checking and merging the two sets, the comprehensive extraction of information from different types of floor plans is realized, and the effect of providing complete and accurate data support for subsequent information quality checking and AI review is achieved.

[0076] Further, the step A342 in the method provided by the embodiment of the application comprises:

[0077] A342-1: Traverse the non-templated extraction floor plan set to perform preliminary graphic element identification and vectorization identification, and obtain a preliminary graphic element vector group set.

[0078] A342-2: Perform associated guidance prompt retrieval and experience guidance prompt retrieval based on the preliminary graphic element vector group set, and obtain an associated guidance prompt set and an experience guidance prompt set.

[0079] A342-3: Call a drawing template generator to generate drawing templates based on the associated guidance prompt set and the experience guidance prompt set, and obtain a newly added drawing template set.

[0080] A342-4: Perform layer accuracy checking on the newly added drawing template set, and if the checking is passed, perform non-templated extraction based on the newly added drawing template set to obtain a second floor plan information set.

[0081] In the embodiment of the application, the drawing template generator is a functional component for generating newly added drawing templates that adapt to non-templated extraction floor plans.

[0082] Specifically, first, based on the preset geometric feature rules and the graphic element detection network, the non-templated extracted floor drawing set is divided into components and the component features are extracted, and then the divided component feature group set is standardized vector coded and associated with the corresponding divided components to obtain a preliminary graphic element vector group set, which is described in detail in A342-1A-A342-1B.

[0083] Next, the drawings associated with the non-templated floor drawing set are obtained from the templated extracted floor drawing set to form an associated templated extracted floor drawing group set, and the graphic element vectors in the set are analyzed with the preliminary graphic element vector group as an index to obtain an associated guidance prompt set; then the experience guidance prompt set is retrieved from the experience database containing the typical component graphic element vector and the application scenario information, and the specific steps are described in detail in A342-2A-A342-2D.

[0084] Then the drawing template generator is called, which takes the associated guidance prompt set and the experience guidance prompt set as the core input to construct the new drawing template. The associated guidance prompt set contains the highest associated degree of the non-templated drawing template, and the experience guidance prompt set covers the application scenario information of the typical component, such as the size label requirement of the roof fire water tank and the component association rule of the ventilation room. The drawing template generator first analyzes the features of the two types of prompt sets, extracts the core framework of the associated template including the drawing information dimension, the component classification logic and the data format standard, and then integrates the special component requirements in the experience prompts, such as the depth parameter extraction rule of the underground water collecting well and the pipeline connection attribute of the equipment room, etc. Through template structure reorganization and information dimension completion, a special new template is generated for each drawing in the non-templated extracted floor drawing set, and finally the new drawing template set is formed.

[0085] Subsequently, the layer accuracy of the new drawing template set is verified, which is the key to ensure that the subsequent information extraction does not appear layer confusion. In the verification process, the intelligent review system will call the layer matching algorithm to compare the preset extraction layer corresponding to each component in the new drawing template, such as the building component corresponding to the building professional layer, the fire fighting component corresponding to the water supply and drainage professional layer, etc. with the actual layer attribute of the non-templated drawing: on the one hand, it checks whether the association relationship between the component and the layer in the drawing template is correct, for example, the template stipulates that the fire hydrant information is extracted from the water supply and drainage-fire fighting layer, and it needs to be confirmed that the fire hydrant in the drawing is indeed drawn on this layer; on the other hand, it verifies the uniqueness and integrity of the layer identification, for example, to avoid mixing the building-wall layer with the structure-wall layer in the drawing template, and to ensure that there is no missing key extraction layer. If there is an error in the layer association of a new drawing template, such as setting the spray head extraction layer as the building layer, the intelligent review system will prompt that the verification fails, and return to the drawing template generator to adjust the layer parameters; if the layer association of all new drawing templates is accurate, the verification is passed.

[0086] After the check passes, information extraction is performed on the non-templated extraction floor plan set based on the newly added floor plan template set. During the extraction, the intelligent review system will automatically locate the corresponding components and data in the non-templated extraction floor plan according to the information dimensions and extraction rules preset by each newly added template: for example, according to the roof layer template, the length x width x height of the waterproof box, the plan size of the ventilation room, and the specific position coordinates of the roof evacuation exit are extracted from the non-templated extraction floor plan; according to the underground layer template, the diameter, depth and connection mode with the drainage pipeline of the water collecting well are extracted. The extraction results of each non-templated extraction floor plan are arranged into structured data according to the template dimensions, and the extraction data of all non-templated extraction floor plans are summarized to obtain the second floor plan information set.

[0087] Through the process of generating a new template by calling the floor plan template generator, integrating the association and experience prompt, performing layer accuracy verification on the new template, and extracting non-templated extraction floor plan information based on the template after the verification passes, the standardized information extraction of the non-templated extraction floor plan is realized, and the effect of providing accurate and comprehensive non-templated extraction floor plan data for subsequent formation of a complete floor plan information set is achieved.

[0088] Further, step A342-1 in the method provided by the embodiment of the application comprises:

[0089] A342-1A: based on the preset geometric feature rules and the graphic element detection network, component division and component feature extraction are performed on the non-templated extraction floor plan set to obtain a divided component group set and a divided component feature group set.

[0090] A342-1B: standardization vector coding is performed on the divided component feature group set, and the coding result is associated with the corresponding divided component in the divided component group set to obtain the preliminary graphic element vector group set.

[0091] In the embodiment of the application, the graphic element detection network is a technical component for component division and component feature extraction of the non-templated extraction floor plan set.

[0092] In an embodiment, when performing preliminary primitive identification and vectorization identification on the non-templated extraction floor plan set, first, based on preset geometric feature rules including line length, angle, closure, and primitive detection network, component division and component feature extraction are performed on each floor plan in the set. The primitive detection network scans the floor plan, identifies building components, including walls, columns, doors, windows, room text, etc.; potential areas of water supply and drainage / fire-fighting components, including water pipes, sprinklers, fire hydrants, fire extinguishers, etc.; and further subdivides component types in combination with preset rules such as line length, for example, to distinguish between long outline lines of walls and short outline lines of doors; angle, for example, to determine the right angle shape of walls and columns; and closure, for example, to identify the closed shape of doors and windows. For example, for a certain underground floor plan, after analysis by the primitive detection network and the preset rules, a fire door can be divided, the fire door meets the specific line length and closure, and a fire hydrant is also divided, the fire hydrant is circular and connected with pipes and other components. The division results of all floor plans are summarized to obtain a division component group set; at the same time, the dimensions of each component are extracted, such as wall thickness, door length and width, position coordinates, attribute text, such as power distribution room text, and other features to form a division component feature group set.

[0093] Subsequently, the division component feature group set is traversed, and the features of each component are standardized and vector encoded according to uniform rules to integrate the size values of the components, two-dimensional coordinates of the positions, semantic vectors of the attribute texts, etc. into fixed-dimension vectors. After encoding, each vector result is associated with the corresponding component in the division component group set to ensure that the vector accurately corresponds to a specific component. For example, the encoding vector of a component divided as a fire door in a certain floor plan is associated with the fire door in the division component group entry. After the above operations are completed for all components in the division component feature group set, a preliminary primitive vector group set is finally obtained.

[0094] By performing component division and feature extraction on floor plans that have not been completely successfully templated based on preset geometric feature rules and primitive detection networks, and then standardizing and encoding the component features and associating the corresponding components, accurate vectorization representation of non-templated floor plan components is achieved, providing a structured and computable primitive data basis for subsequent association-oriented prompt retrieval and experience-oriented prompt retrieval.

[0095] Further, step A342-2 in the method provided by the embodiments of the present application includes:

[0096] A342-2A: Obtain the floor plans associated with the non-templated extraction floor plan set in the templated extraction floor plan set to obtain an associated templated extraction floor plan group set.

[0097] A342-2B: indexing the preliminary set of graphic element vectors, respectively, the graphic element vectors in the set of associated template extraction floor plan are analyzed, the template corresponding to the highest correlation degree of the associated template extraction floor plan is taken as the associated guidance prompt, and the set of associated guidance prompts is obtained.

[0098] A342-2C: obtaining an experience database, wherein the experience database includes typical component graphic element vectors and application scenario information.

[0099] A342-2D: indexing the preliminary set of graphic element vectors, retrieving and analyzing the experience database, and obtaining the experience guidance prompt set.

[0100] Optionally, first, the drawings associated with the set of non-template extraction floor plans in the set of template extraction floor plans are obtained. Since the building drawings have inter-floor and functional area association logic, for example, roof layer drawings are often associated with top standard layer drawings, underground layer drawings are associated with first floor drawings, and intelligent review systems can filter out drawings with association relationships according to the floor number, building function annotation and other information of the drawings, and form a set of associated template extraction floor plan groups. For example, the underground second floor drawing of a certain commercial complex belongs to the non-template extraction set, and the system can identify that the associated underground first floor and first floor drawings belong to the template extraction set, and these drawings are summarized as the set of associated template extraction floor plan groups.

[0101] Then, indexing the preliminary set of graphic element vectors, the graphic element vectors in the set of associated template extraction floor plan groups are analyzed, wherein the preliminary set of graphic element vectors is the component vectors extracted from the non-template extraction floor plan, and the correlation analysis calculates the similarity between the non-template plan component vectors and the associated template plan component vectors. The specific process is as follows:

[0102] When calculating the similarity between the non-template plan component vectors and the associated template plan component vectors, first, the same type of components in the two types of drawings are selected for comparison to avoid invalid calculation between different types of components, for example, the sump well component vector in the non-template plan is only compared with the sump well component vector in the associated template plan, and the drainage pipe component vector in the non-template plan is only compared with the drainage pipe component vector in the associated template plan.

[0103] In the comparison of a single homogeneous component vector, the similarity is calculated dimension by dimension: in the size feature dimension, the size deviation rate of the non-templated component and the templated component is calculated, for example, the diameter of the non-templated catch basin is 1.2 meters, and the diameter of the templated catch basin is 1.1 meters, the deviation rate is about 8.3%, the smaller the deviation rate, the higher the similarity of this dimension; in the position feature dimension, the straight-line distance of the two types of components in the drawing coordinate system is calculated, the closer the distance, the higher the similarity of this dimension; in the attribute text feature dimension, the matching degree of the attribute texts of the two types of components is calculated by a semantic similarity algorithm, for example, the semantic similarity of the catch basin and the fire water catch basin is higher than that of the catch basin and the sewage well, the closer the semantics, the higher the similarity of this dimension.

[0104] Further, different weights are given according to the importance of the features to the building review, for example, the weight of the component function feature corresponding to the attribute text is higher than that of the size deviation feature, the weight of the position feature is higher than that of the secondary size feature, and finally the similarity results of each dimension are weighted and summed according to the weights to obtain the similarity of a single homogeneous component. Then, the similarity of all homogeneous components in the associated templated drawing of the non-templated extraction floor drawing is averaged to obtain the component vector similarity of the two drawings as a whole.

[0105] Next, the associated templated extraction floor drawing with the highest correlation degree is found, and the template corresponding to the drawing is taken as the associated guidance prompt. The highest correlation degree means that in the set of associated templated extraction floor drawings, the overall component vector similarity value of a certain templated drawing and the current non-templated drawing is the largest. Taking the underground second floor non-templated drawing as an example, its set of associated templated extraction floor drawings includes the templated drawings of the underground first floor and the first floor. After calculation, if the overall component vector similarity of the underground second floor drawing and the underground first floor drawing is 82% and the overall component vector similarity of the underground second floor drawing and the first floor drawing is 65%, then the underground first floor drawing is the associated templated extraction floor drawing with the highest correlation degree. At this time, the template corresponding to the associated templated drawing with the highest correlation degree is taken as the associated guidance prompt of the underground second floor non-templated drawing. According to the same process, the associated templated drawing with the highest correlation degree is found for each drawing in the set of non-templated extraction floor drawings, and the corresponding template is obtained. After all these templates are collected, the set of associated guidance prompts is formed.

[0106] Subsequently, an experience database is obtained, which stores typical component graph vector and application scenario information. The typical components cover special components of different building types including residential, office, industrial, etc., such as roof fire water tank of residential building, core tube stairwell of office building, equipment foundation of industrial plant, etc. Each component vector corresponds to a specific application scenario description, for example, the effective volume of the roof fire water tank should meet the fire water demand within the fire duration time of the building, etc.

[0107] After that, the experience database is searched and analyzed with the preliminary set of primitive vector groups as index. The intelligent review system matches the component vectors in the non-templated drawing with the typical component vectors in the experience database, finds the entries with high similarity, and extracts the corresponding application scenario information as experience-oriented prompts. For example, the escalator component vector extracted from the non-templated drawing is matched with the typical vector of the commercial building escalator in the experience database, and the information such as the escalator inclination angle and net width requirement under the entry is extracted to form the experience-oriented prompt. Finally, all such prompts are aggregated into an experience-oriented prompt set.

[0108] By first screening the associated templated drawings and analyzing to obtain the association-oriented prompts, and then retrieving the experience-oriented prompts from the experience database storing the typical components and scenarios, a double guidance based on the building association logic and the past engineering experience is provided for the subsequent generation of new templates adapted to the non-templated drawings, which guarantees the rationality and applicability of the new drawing templates.

[0109] Further, the step A400 in the method provided by the embodiment of the present application comprises:

[0110] A410: obtaining preset accuracy verification indicators and preset integrity verification indicators.

[0111] A420: performing information comparison verification on the floor drawing information set based on the preset accuracy verification indicators and preset specification review rules to obtain an information accuracy verification result.

[0112] A430: performing drawing information coverage and cross-professional consistency verification on the floor drawing information set based on the preset integrity verification indicators to obtain an information integrity verification result.

[0113] A440: if the information accuracy verification result and the information integrity verification result meet the requirements, the verification is passed.

[0114] In one embodiment, first, the preset accuracy verification indicators and the preset integrity verification indicators are obtained in combination with the preset specification review rules. The preset specification review rules are international specification review rules, specifically covering international general specifications in the fields of architectural design, structural engineering, mechanical and electrical installation, etc., such as the provisions on structural safety in the International Building Code (IBC), the requirements on pipeline layout in the International Mechanical Code (IMC), etc.

[0115] In one aspect, the preset accuracy checking index is set around the dimensions of geometric property accuracy, layer and property accuracy, topological relationship accuracy, and knowledge rule accuracy, and each dimension is combined with international standard review rules: the geometric property accuracy clearly defines the wall thickness, such as the residential partition wall thickness not less than 200 mm in the International Building Code (IBC), and the door hole size and door block size consistency need to meet the international mapping coordination standard; the layer and property accuracy needs to check whether the corresponding drawing is correct, for example, according to the international layer classification standard, the building-wall layer should not contain pipeline graphics, and the line type, color, and elevation symbol need to comply with the International Engineering Drawing Standardization; the topological relationship accuracy requires verification of the spatial relationship between graphics, such as the connection of columns and floors, foundations needs to meet the international structure connection node standard, the nesting of doors and windows needs to comply with the international building component installation gap standard, and the spatial conflict of pipelines and structures needs to meet the international mechanical and electrical pipeline avoidance standard; the knowledge rule accuracy is based on international standards to set the width of the evacuation door not less than 1 m per 100 people, the fire compartment boundary closure needs to meet the international fire code (IFC), and the mechanical and electrical pipeline slope needs to comply with the international drainage engineering code (IDEC) drainage requirements.

[0116] On the other hand, the preset integrity checking index is developed from element coverage integrity, structural logic integrity, system configuration integrity, and information description integrity, and each dimension is combined with international standard review rules: the element coverage integrity needs to traverse the room / function area to check whether the components such as doors and windows in the room, handrails in the stairwell, and fire doors, and water supply and drainage points exist, in compliance with the International Building Component Integrity Verification Guide; the structural logic integrity checks the absence or breakage of components, such as whether the shear wall needs to span the full layer to meet the international high-rise building structure standard, whether the frame beam needs to cover all spans to comply with the international beam structure design standard, and whether the floor has gaps or is suspended to comply with the international floor structure safety standard; the system configuration integrity checks the system closure for mechanical / electrical / heating systems, such as checking the head-to-tail closure of water supply and drainage pipelines to comply with the international water supply and drainage system installation standard, the electrical circuit connection to meet the international electrical wiring standard, and the full-area coverage of fire sprinkler to comply with the international fire sprinkler system design standard; the information description integrity checks whether the attribute fields are complete, requiring graphics to contain material, number, elevation, and other attributes, and key components to have BIM-ID or unique coding, in compliance with the International BIM Information Delivery Standard.

[0117] Then, combined with the preset specification review rules, the information quality of the floor drawing information set is checked based on the above preset checking indicators, and each preset checking indicator corresponds to a module. Taking a certain residential floor drawing as an example, when accuracy checking is performed: the geometric attribute accuracy module extracts the thickness of the household wall as 240mm, which meets the specification value of not less than 200mm in the international building code (IBC), and the door size 900mm is compared with the door block 880mm, and the reserved installation gap needs to meet the international building component installation gap standard (such as allowing 20mm installation adjustment amount) to determine consistency; the layer and attribute accuracy module scans the building-wall layer to confirm that there is no pipeline element, and the fire evacuation indication line is a yellow solid line, which meets the provisions of the international engineering drawing unified specification regarding fire identification line type; the topological relationship accuracy module verifies that the column and floor board connection meets the international structure connection node specification through three-dimensional coordinate comparison, checks that the door and window are nested in the wall to meet the international building component installation gap standard, and detects that the vertical distance between the drainage pipeline and the beam body is 150mm, which meets the international mechanical and electrical pipeline avoidance specification regarding installation gap without conflict; the knowledge rule accuracy module calculates the evacuation door width of 1.2m, which corresponds to 100 people, which meets the requirement of the International Fire Protection Association (IFPA) that not less than 1m per 100 people, and detects that the fire compartment boundary is closed to meet the international fire protection specification (IFC), and the drainage horizontal pipe slope is 0.005, which meets the requirement of not less than 0.003 in the international drainage engineering specification (IDEC).

[0118] When integrity checking is performed, the international specification review rules also need to be combined, which will not be described here: the element coverage integrity module checks whether the doors and windows in the bedroom are complete, whether the stairwell contains handrails and fire doors, and whether the toilet is arranged with drainage points; the structural logic integrity module confirms that the shear wall spans the full layer, the frame beam covers a 6m span, and the floor has no gaps or overhangs; the system configuration integrity module verifies that the water supply and drainage pipeline is closed at both ends, the electrical circuit is connected, the fire sprinkler head spacing is 3m, which meets the specification, and covers the entire area; the information description integrity module checks that the wall is labeled with material sintered shale brick, number Q1, elevation ± 0.000, and shear wall has BIM-ID. If the floor drawing information set meets all the above requirements of the preset accuracy and integrity checking indicators combined with the international specification review rules, the information quality checking is passed.

[0119] By pre-setting multi-dimensional accuracy and integrity checking indicators, combining with the preset specification review rules, and then checking the floor drawing information set based on these indicators, the effect of ensuring that the floor drawing information meets the building industry specifications, drawing standards and functional logic requirements in terms of accuracy and integrity is achieved.

[0120] In summary, the AI-assisted building drawing review method provided by the embodiments of the present application has the following technical effects:

[0121] This application transmits the target architectural drawings to the AI ​​review module of an intelligent review system. After the drawing information is filled in according to the preset project category, a floor list creation command is triggered. Based on this command, the target architectural drawings are automatically identified to construct a floor list. The floor list is traversed, and drawing information is extracted in a templated or non-templated manner according to the corresponding floor drawings to obtain a set of floor drawing information. The information quality of the floor drawing information set is verified from two dimensions: accuracy and completeness. After the verification is passed, a review command is triggered, and the AI ​​review module performs AI review on the set of floor drawing information to obtain a set of AI review results. This achieves efficient auxiliary review of architectural drawings, making the architectural drawing review process more standardized and the review results more accurate and reliable. It achieves the technical effects of making architectural drawing information extraction more efficient, adapting to non-standardized drawings, and standardizing and automating data processing, making architectural drawing review data more comprehensive and accurate, and improving the accuracy and collaboration of the review.

[0122] Example 2, as Figure 2 As shown, based on the same inventive concept as in Embodiment 1 above, this application provides an AI-collaborative architectural drawing review system, the system comprising:

[0123] Floor table creation instruction acquisition module 1 is used to input the target building drawings into the AI ​​review module of the intelligent review system, and fill in the drawing information according to the preset project category. When the drawing information is obtained, the floor table creation instruction is triggered.

[0124] The floor list construction module 2, based on the floor table creation instruction, automatically identifies the drawing frame of the target building drawing according to the drawing information to construct a floor list, wherein each floor corresponds to a floor drawing.

[0125] Floor drawing information acquisition module 3 is used to traverse the floor list to extract the drawing information of the corresponding floor drawings in a templated or non-templated manner, so as to obtain the floor drawing information set.

[0126] AI review result set acquisition module 4 is used to combine preset standard review rules to traverse the floor drawing information set from two dimensions: accuracy and completeness to perform element information quality verification. If the verification passes, a review instruction is triggered, and the AI ​​review module performs AI review on the floor drawing information set to obtain an AI review result set.

[0127] An AI review result set feedback module 5 is configured to feed the AI review result set to a draftsman for self-checking and drawing modification, and submit the modified target modified building drawing and the AI self-checking result set containing self-checking notes to a design collaboration system, and check the target modified building drawing and the AI self-checking result set containing self-checking notes by a reviewer.

[0128] Further, the floor table creation instruction acquisition module 1 is configured to perform the following steps:

[0129] The preset project category includes a sub-item name, a professional name, an engineering category, a building type, a fire resistance rating, a total building area, an above-ground building area, an underground building area, whether to use an automatic fire extinguishing system, and a selected review specification.

[0130] Further, the floor list construction module 2 is configured to perform the following steps:

[0131] A preset floor table template is acquired, wherein the preset floor table template includes a drawing name, a floor number, a floor height, a floor elevation, and a floor type; a mode analysis intelligent agent performs frame scanning and detection on the target building drawing according to the drawing information to obtain an image quantity and an image detection attribute set; and a floor table is created according to the image quantity and the image detection attribute set and the preset floor table template to obtain a floor list.

[0132] Further, the floor drawing information set acquisition module 3 is configured to perform the following steps:

[0133] A drawing feature extractor is called to extract features of the floor drawing of each floor in the floor list to obtain a floor drawing feature set; the floor drawing feature set is matched with template features of each drawing information template in a drawing information template library in terms of similarity, if a matching result is greater than or equal to a preset similarity threshold, the corresponding floor drawing is associated with the matching drawing information template, and added to a template extraction floor drawing set; if the matching result is less than the preset similarity threshold, the corresponding floor drawing is added to a non-template extraction floor drawing set; and drawing information is extracted according to the template extraction floor drawing set and the non-template extraction floor drawing set to obtain the floor drawing information set.

[0134] Further, the floor drawing information set acquisition module 3 is configured to perform the following steps:

[0135] Template extraction is performed on the template extraction floor plan set to obtain a first floor plan information set; non-template extraction is performed on the non-template extraction floor plan set to obtain a second floor plan information set; the first floor plan information set and the second floor plan information set are summarized to obtain a floor plan information set.

[0136] Further, the floor plan information set acquisition module 3 is configured to perform the following steps:

[0137] Preliminary graphic element vector group sets are obtained by performing preliminary graphic element identification and vectorization identification on the non-template extraction floor plan set; the associated guidance prompts set and the experience guidance prompts set are obtained by performing associated guidance prompt retrieval and experience guidance prompt retrieval based on the preliminary graphic element vector group sets; the floor plan template generator is called to generate floor plan templates based on the associated guidance prompts set and the experience guidance prompts set to obtain a new floor plan template set; the second floor plan information set is obtained by performing non-template extraction based on the new floor plan template set.

[0138] Further, the floor plan information set acquisition module 3 is configured to perform the following steps:

[0139] The component division and component feature extraction are performed on the non-template extraction floor plan set based on the preset geometric feature rules and the graphic element detection network to obtain a division component group set and a division component feature group set; the preliminary graphic element vector group sets are obtained by performing standardized vector coding on the division component feature group set and associating the coding results with the corresponding division components in the division component group set.

[0140] Further, the floor plan information set acquisition module 3 is configured to perform the following steps:

[0141] The floor plans associated with the non-template extraction floor plan set in the template extraction floor plan set are obtained to obtain an associated template extraction floor plan group set; the graphic element vectors in the associated template extraction floor plan group set are analyzed based on the preliminary graphic element vector group sets as indexes, and the template corresponding to the associated template extraction floor plan with the highest association degree is taken as the associated guidance prompt to obtain the associated guidance prompt set; the experience database is obtained, wherein the experience database includes typical component graphic element vectors and application scenario information; the experience database is retrieved and analyzed based on the preliminary graphic element vector group sets as indexes to obtain the experience guidance prompts set.

[0142] Further, the AI review result set acquisition module 4 is configured to perform the following steps:

[0143] The preset accuracy check index and the preset integrity check index are acquired; information comparison and checking is performed on the set of floor plan information based on the preset accuracy check index and preset specification review rules to obtain an information accuracy check result; the set of floor plan information is checked for plan information coverage and cross-professional consistency based on the preset integrity check index to obtain an information integrity check result; if the information accuracy check result and the information integrity check result meet the requirements, the checking is passed.

[0144] The AI-assisted building plan review system provided by the embodiment of the application can execute the AI-assisted building plan review method provided by any embodiment of the application, and has the corresponding function modules and beneficial effects of the execution method.

[0145] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or server, and each unit and module included is only divided according to the functional logic, but is not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of each functional unit are only for easy mutual differentiation, and do not limit the protection scope of the present application.

[0146] The above specific embodiments do not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps described in the present application can be executed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

Claims

1. An AI-assisted building drawing review method, characterized in that, The method comprises: The target building drawing is input into the AI review module of the intelligent review system, and the drawing information is filled in according to the preset project category; when the drawing information is obtained, a floor table creation instruction is triggered; Based on the floor table creation instruction, the target building drawing is automatically identified according to the drawing information, and a set of automatic identification results of the drawing frame is obtained, wherein each automatic identification result of the drawing frame corresponds to a floor drawing; Iterate through the set of automatic identification results of the drawing frame to extract the drawing information template or non-template, and obtain a set of floor drawing information; Combined with the preset specification review rules, the set of floor drawing information is iterated from the accuracy and integrity dimensions to perform information quality checking, if the checking is passed, the review instruction is triggered, the set of floor drawing information is reviewed by the AI review module, and a set of AI review results is obtained; The set of AI review results is fed back to the drawing designer for self-review annotation and drawing modification, the modified target modified building drawing and the AI self-review result set containing the self-review annotation are submitted to the design collaboration system, and the target modified building drawing and the AI self-review result set containing the self-review annotation are checked by the reviewer in the design collaboration system; Wherein, iterating through the set of automatic identification results of the drawing frame to extract the drawing information template or non-template, and obtaining a set of floor drawing information, comprises: Calling a drawing feature extractor, performing feature extraction on the set of automatic identification results of the drawing frame, and obtaining a set of floor drawing features; Respectively, the set of floor drawing features and the template features of each drawing information template in the drawing information template library are matched in similarity, if the matching result is greater than or equal to the preset similarity threshold, the corresponding floor drawing is associated with the matching drawing information template, and is added to the template extraction floor drawing set; If the matching result is less than the preset similarity threshold, the corresponding floor drawing is added to the non-template extraction floor drawing set; According to the template extraction floor drawing set and the non-template extraction floor drawing set, the drawing information is extracted, and the set of floor drawing information is obtained; Wherein, according to the template extraction floor drawing set and the non-template extraction floor drawing set, the drawing information is extracted, and the set of floor drawing information is obtained, comprising: Template extraction is performed on the template extraction floor drawing set to obtain a first set of floor drawing information; Non-template extraction is performed on the non-template extraction floor drawing set to obtain a second set of floor drawing information; The first set of floor drawing information and the second set of floor drawing information are summarized to obtain a set of floor drawing information.

2. The AI-cooperated building drawing auxiliary review method according to claim 1, wherein The preset project category includes sub-item name, professional name, engineering category, building type, fire resistance rating, total building area, aboveground building area, underground building area, whether to use automatic fire extinguishing system and selection of review specification.

3. The AI-cooperated building drawing auxiliary review method according to claim 1, wherein Based on the floor table creation instruction, the target building drawing is automatically identified according to the drawing information, and a set of automatic identification results of the drawing frame is obtained, comprising: Obtain a preset floor table template, wherein the preset floor table template includes a drawing name, a floor number, a floor height, a floor elevation, and a floor type; According to the drawing information, the agent performs frame scanning and detection on the target building drawing through mode analysis, obtains an image quantity and an image detection attribute set, and performs automatic recognition on the frame based on the image quantity and the image detection attribute set to obtain an automatic recognition result set. According to the drawing information, the agent performs frame scanning and detection on the target building drawing through mode analysis, obtains an image quantity and an image detection attribute set, and performs automatic recognition on the frame based on the image quantity and the image detection attribute set to obtain an automatic recognition result set.

4. The AI-cooperated building drawing auxiliary review method of claim 1, wherein, The non-templated extraction floor drawing set is subjected to non-templated extraction to obtain a second floor drawing information set, including: Iterate through the non-templated extraction floor drawing set to perform preliminary graph element identification and vectorization identification to obtain a preliminary graph element vector group set; Based on the preliminary graph element vector group set, perform associated guidance prompt retrieval and experience guidance prompt retrieval to obtain an associated guidance prompt set and an experience guidance prompt set; Call a drawing template generator to generate a drawing template based on the associated guidance prompt set and the experience guidance prompt set to obtain a new drawing template set; If the verification is passed, the non-templated extraction is performed based on the new drawing template set to obtain a second floor drawing information set.

5. The AI-assisted building drawing review method of claim 4, wherein, Iterate through the non-templated extraction floor drawing set to perform preliminary graph element identification and vectorization identification to obtain a preliminary graph element vector group set, including: Based on a preset geometric feature rule and a graph element detection network, the non-templated extraction floor drawing set is subjected to component division and component feature extraction to obtain a divided component group set and a divided component feature group set; Iterate through the divided component feature group set to perform standardized vector coding, and associate the coding result with the corresponding divided component in the divided component group set to obtain the preliminary graph element vector group set.

6. The AI-assisted building drawing review method of claim 4, wherein, Based on the preliminary graph element vector group set, perform associated guidance prompt retrieval and experience guidance prompt retrieval to obtain an associated guidance prompt set and an experience guidance prompt set, including: Obtain the drawing associated with the non-templated extraction floor drawing set in the templated extraction floor drawing set to obtain an associated templated extraction floor drawing group set; Take the preliminary graph element vector group set as an index to respectively perform associated analysis on the graph element vectors in the associated templated extraction floor drawing group set, and take the template corresponding to the associated templated extraction floor drawing with the highest correlation degree as the associated guidance prompt to obtain the associated guidance prompt set; Obtain an experience database, wherein the experience database includes typical component graph element vectors and application scenario information; Take the preliminary graph element vector group set as an index to perform retrieval analysis on the experience database to obtain the experience guidance prompt set.

7. The AI-cooperated building drawing auxiliary review method of claim 1, wherein, Combine the preset specification review rules to perform information quality verification on the floor drawing information set from the accuracy and completeness dimensions, including: Obtain a preset accuracy verification index and a preset completeness verification index; Based on the preset accuracy verification index and the preset specification review rules, perform information comparison verification on the floor drawing information set to obtain an information accuracy verification result; Performing integrity check on the floor plan information set based on the preset integrity check index to obtain information integrity check result; If the information accuracy check result and the information integrity check result meet the requirements, the check is passed.

8. An AI-assisted building drawing review system, characterized by, An AI-assisted building plan review method according to any one of claims 1-7, the system comprising: A floor table creation instruction acquisition module configured to input a target building plan into an AI review module of an intelligent review system, fill in plan information according to a preset project category, and trigger a floor table creation instruction when the plan information is obtained; A set of automatic frame recognition results acquisition module configured to perform automatic frame recognition on the target building plan based on the floor table creation instruction and the plan information, and obtain a set of automatic frame recognition results, wherein each automatic frame recognition result corresponds to a floor plan; A floor plan information set acquisition module configured to traverse the set of automatic frame recognition results to extract plan information in a template or non-template manner, and obtain a floor plan information set; An AI review result set acquisition module configured to perform plan element information quality check on the floor plan information set from the dimensions of accuracy and integrity, trigger a review instruction if the check is passed, perform AI review on the floor plan information set by the AI review module, and obtain an AI review result set; An AI review result set feedback module configured to feed back the AI review result set to a plan designer for self-review annotation and plan modification, submit the modified target modified building plan and the AI self-review result set containing the self-review annotation to a design collaboration system, and check the target modified building plan and the AI self-review result set containing the self-review annotation in the design collaboration system by a reviewer; The floor plan information set acquisition module is further configured to perform the following steps: Call a plan feature extractor to extract features from the set of automatic frame recognition results to obtain a floor plan feature set; Respectively match the floor plan feature set with template features of each plan information template in a plan information template library for similarity, if the matching result is greater than or equal to a preset similarity threshold, associate the corresponding floor plan with the matching plan information template, and add it to a template extraction floor plan set; If the matching result is less than the preset similarity threshold, add the corresponding floor plan to a non-template extraction floor plan set; Extract plan information according to the template extraction floor plan set and the non-template extraction floor plan set to obtain the floor plan information set; The floor plan information set acquisition module is further configured to perform the following steps: Template extraction on the template extraction floor plan set to obtain a first floor plan information set; Non-template extraction on the non-template extraction floor plan set to obtain a second floor plan information set; Summarize the first floor plan information set and the second floor plan information set to obtain a floor plan information set.

Images