AI drawing creation system, AI drawing creation method, and AI drawing creation program

Abstract

The objective is to provide novel technologies that contribute to shortening construction periods in the design of structures. [Solution] AI drawing creation system, It comprises an acquisition unit and a three-dimensional model generation unit. The acquisition unit acquires two-dimensional drawing data relating to a two-dimensional architectural drawing, which includes figures and / or text indicating the components. The three-dimensional model generation unit acquires member information from the two-dimensional drawing data, A three-dimensional model is generated based on the aforementioned component information. AI drawing creation system.

Description

Technical Field

[0001] The present invention relates to an AI drawing creation system, an AI drawing creation method, and an AI drawing creation program.

Background Art

[0002] In the fields of architectural design and construction, drawings such as design drawings and construction drawings are used. However, since construction cannot be actually carried out only with design drawings that describe only the design and the main structure, based on the design drawings, detailed construction drawings that describe the detailed layout of the structure (such as a building) and dimensions required for incidental work are created.

[0003] In the work related to the creation of such construction drawings, conventionally, workers such as construction drawing creators used general-purpose systems (such as 2D CAD systems and 3D CAD systems) to create them, which took a lot of time.

[0004] In Patent Document 1, a new design method and design system capable of shortening the construction period are disclosed for the design of steel structures.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] However, the invention described in Patent Document 1 does not disclose a method for generating a three-dimensional model from a two-dimensional drawing.

[0007] An object of the present invention is to provide a novel technique that contributes to shortening the construction period in the design of a structure.

Means for Solving the Problems

[0008] [1] An AI drawing creation system, It comprises an acquisition unit and a three-dimensional model generation unit. The acquisition unit acquires two-dimensional drawing data relating to a two-dimensional architectural drawing, which includes figures and / or text indicating the components. The three-dimensional model generation unit generates member information based on the two-dimensional drawing data, A three-dimensional model is generated based on the aforementioned component information. AI drawing creation system. [2] The three-dimensional model generation unit obtains text information related to text and graphic information related to shapes from the two-dimensional drawing data, and performs matching of shapes representing members with text, The AI ​​drawing creation system according to [1] generates the three-dimensional model based on the text information relating to the matched shapes and text, and the shape information. [3] The AI ​​drawing creation system according to [2], wherein the three-dimensional model generation unit performs matching of the figures and the text based on the position information of the figures and the position information of the text obtained from the two-dimensional drawing data as a process related to the matching. [4] The architectural drawings include material drawings, The AI ​​drawing creation system according to any one of [1] to [3], wherein the three-dimensional model generation unit generates member information including member parameters and part numbers based on two-dimensional drawing data relating to the material drawing, and generates the three-dimensional model based on the member information. [5] The architectural drawings further include a plan view and a side view, The three-dimensional model generation unit further identifies the code and / or model number of the member indicated by the figure included in the two-dimensional drawing data based on text information obtained from the two-dimensional drawing data relating to the plan view and / or side view. The AI ​​drawing creation system according to [4] generates the three-dimensional model based on the identified part information of the code and / or model number. [6] The architectural drawings include a plan view and a side view, The three-dimensional model generation unit further identifies the hierarchy of the plan view and side view included in the architectural drawing based on the two-dimensional drawing data, An AI drawing creation system according to any one of [1] to [5], which identifies the correspondence between figures in a plan view and a side view based on a specified hierarchy, and generates the three-dimensional model by integrating the figure information and text information of the corresponding figures. [7] The three-dimensional model generation unit performs a process to identify correspondences between figures belonging to the same hierarchical level in the plan view and the side view, as part of the process for identifying correspondences, the AI ​​drawing creation system as described in [6]. [8] The AI ​​drawing creation system is further equipped with a fee calculation unit, The AI ​​drawing creation system according to any one of [1] to [7], wherein the fee calculation unit determines the volume and / or weight of the members included in the structure based on the three-dimensional model and calculates the fee. [9] The AI ​​drawing creation system further includes a two-dimensional drawing generation unit, The AI ​​drawing creation system according to any one of [1] to [8], wherein the two-dimensional drawing generation unit generates a two-dimensional construction drawing based on the three-dimensional model.

[10] A method for creating AI drawings that can be performed by a computer, It comprises an acquisition process and a three-dimensional model generation process. In the acquisition process described above, two-dimensional drawing data relating to a two-dimensional architectural drawing is acquired, which includes figures and / or text indicating the components. In the three-dimensional model generation process, component information is generated based on the two-dimensional drawing data. A three-dimensional model is generated based on the aforementioned component information. AI drawing creation method. An AI drawing creation program that causes the computer to execute the AI ​​drawing creation method described in

[11]

[10] .

[0009] The invention described in [1] makes it possible to easily generate a three-dimensional model from a two-dimensional drawing.

[0010] According to the invention according to [2], a three-dimensional model can be easily generated using the figures and text described in a two-dimensional drawing.

[0011] According to the invention according to [3], it is possible to create a three-dimensional model using information on highly relevant figures and text by utilizing the position of the text.

[0012] According to the invention according to [4], information on members can be obtained from a material drawing to generate a three-dimensional model.

[0013] According to the invention according to [5], the model number of a member can be specified, and a three-dimensional model can be generated using the member information corresponding to the model number.

[0014] According to the invention according to [6], a three-dimensional model can be generated by integrating data on corresponding figures in a plan view and a side view.

[0015] According to the invention according to [7], figures at the same level in a plan view and a side view can be associated with each other.

[0016] According to the invention according to [8], it is possible to calculate usage fees using a three-dimensional model.

[0017] According to the invention according to [9], it is possible to generate construction drawings using a three-dimensional model.

Effect of the Invention

[0018] According to the present invention, in designing a structure, it is possible to provide a novel technique that contributes to shortening the construction period.

Brief Description of the Drawings

[0019] [Figure 1] A diagram showing an example of a plan view in one embodiment [Figure 2] A diagram showing an example of a side view in one embodiment [Figure 3] A diagram showing an example of a material drawing in one embodiment [Figure 4] A diagram showing the system configuration in one embodiment. [Figure 5] Figure showing the hardware configuration in one embodiment. [Figure 6] Figure showing an example of a three-dimensional model in one embodiment. [Figure 7] A diagram showing the processing flow in one embodiment. [Modes for carrying out the invention]

[0020] The AI ​​drawing creation system, AI drawing creation method, and AI drawing creation program of the present invention will be described below with reference to the drawings. Note that the embodiments shown below are examples of the present invention, and the present invention is not limited to these embodiments; various configurations can be adopted.

[0021] This embodiment describes the configuration and operation of an AI drawing creation system, but similar configurations, devices, computer programs, and program recording media storing such programs can achieve similar effects. The series of processes according to this embodiment, described below, are provided as a computer-executable program and can be provided via non-transient computer-readable recording media such as CD-ROMs and flexible disks, as well as via communication lines.

[0022] The AI ​​drawing creation system in this embodiment is primarily used to create construction drawings for steel frame structures, which are mainly composed of steel frames.

[0023] This AI drawing creation system uses architectural drawings, including design drawings and construction drawings, to perform the processing described below.

[0024] A design drawing is a drawing created by an architect, primarily depicting the design of a structure (such as a building). Because attempting to construct a structure based solely on the design drawing can lead to minor discrepancies, construction drawings are created based on the design drawing, detailing even the smallest aspects. In this embodiment, the design drawing is a two-dimensional drawing.

[0025] Furthermore, design drawings are drawings created by architects and structural engineers that represent the structure, shape, dimensions, materials used, and equipment of a building. These drawings include a plan view, which is a top-down view of the structure; a side view, which is a cross-section of the structure; and a materials drawing, which shows the model numbers and shapes of the components used in the construction of the structure. Design drawings include architectural drawings related to the design of the building and structural drawings related to the structure of the building. In the process described later, both architectural and structural drawings are used as design drawings to create construction drawings. Architectural drawings include plan views, framing plans, elevations, side views, cross-sections, and detail drawings. Structural drawings include framing plans, floor plans, frame plans, list drawings, and detail drawings. Construction drawings include beam layouts, frame plans, joint reference drawings, and detail drawings. Furthermore, while drawings such as structural drawings and detailed drawings may be included in both the design drawings and the construction drawings, even if the drawings have the same name, the detailed drawings included in the construction drawings are more precise than the drawings included in the design drawings. A structural member is an element that makes up a structure, such as a column, beam, or joint. For structural design, it is preferable to use a set of drawings that include all of the following as design drawings: plan view, side view, and material drawings. However, if all the necessary information about the target structure can be obtained, it is acceptable if any of these are not included. Figure 1 shows an example of a plan view, and Figure 2 shows an example of a side view. A floor plan is a drawing of each floor of a structure viewed from above. A floor plan is a drawing that shows the layout of rooms, walls, doors, dimensions, etc. In this embodiment, a beam layout drawing, which shows the arrangement of steel columns and beams from a top-down perspective, is used as the floor plan for the processing described later. The beam layout drawing is a drawing that shows the symbols of each column and beam (e.g., G1, B2, etc.), the cross-section of the steel material (H-beam, channel steel, etc.), the length and arrangement dimensions of the beams and the dimensions between grid centers, the joints (how each member connects), the position and type of horizontal braces (X-type, V-type, etc.), the position of the joints (center of the beam, in the middle of the column span, etc.), the position of openings and beam penetration holes, etc. A symbol is an individual symbol (identification information) used to identify a member. A side view is a drawing of a cross-section of a structure viewed from the side, assuming a virtual vertical cut of the structure. In this embodiment, drawings such as frame drawings and cross-sectional drawings, which show the arrangement of columns, beams, braces, etc., when the building is viewed vertically, are used as side views in the processing described later. A frame drawing is a drawing that shows the cross-sectional size of columns and beams, the height and position of each member, the dimensions between grid centers, the height of each floor (floor height, beam top, etc.), the shape and arrangement of column bases and column tops, the position and type of axial braces (X-type, V-type, etc.), and the position of joints (center of beams, in the middle of column spans, etc.). Multiple side views, such as frame drawings, may exist for a single structure, for example, by creating one for each grid center. Multiple side views may include side views of other cross-sections parallel to the cross-section in one side view, or side views of other cross-sections that intersect the cross-section in one side view at a right angle. Furthermore, in this embodiment, the frame diagram is included in both the design drawing and the construction drawing, but the frame diagram included in the construction drawing is a more detailed drawing that can withstand construction. The plan view and side view in this embodiment include figures showing the shape of the structural members and text describing the members (e.g., model number and dimensions (height, width, depth, etc.)), figures showing spaces (e.g., rooms), auxiliary lines to assist in design (e.g., lines showing height), and text indicating floor levels. The plan view and side view may also include the model number of each member (e.g., H-400×200×8×12, etc.) along with the reference numeral of each member. Figure 3 shows an example of a material diagram. A material drawing is a drawing that contains detailed information (dimensions and joining methods) of the components used when constructing the structure, and includes figures showing the shape of the components and text describing the components (model number, parameters indicating the shape of the components (height, length of each part, depth, width, etc.)). Components with the model numbers and shapes described in this material drawing are placed in plan views, side views, etc., and used in the construction of the structure. In this embodiment, drawings such as component lists and detailed drawings that enlarge specific parts to depict the structure and component arrangement in detail are used as material drawings in the processing described later. A detailed drawing is a drawing that describes the joint structure of columns and beams, the notch shape and groove of each component, steel plates, reinforcement methods, the size and arrangement of components such as stud bolts, dimensions and tolerances for manufacturing and processing, details of welded parts (type, length, position), and the arrangement and type of high-strength bolts. If a common component master (a component master including model number, parameters, shape, etc.) that can be used at all sites exists, a material drawing does not necessarily have to be included. In this case, the processing described later may refer to information from the component master, such as using the component data from the corresponding component master based on the model number. These drawings preferably include information about the drawing type, such as metadata or the drawing name, but this is not required. In this case, the AI ​​drawing creation system 0 may classify the drawings based on the two-dimensional drawing data and assign a drawing type to each drawing, or it may display a screen instructing the user to input the drawing type and accept the input. Furthermore, in the drawing creation process described later, a set of plan views, side views, and material drawings related to the target structure is used for drawing creation. The number of each type of drawing included in this set is not limited.

[0026] Two-dimensional drawing data is a form of one or more design drawing data including a plan view, a side view, and a material drawing. In this embodiment, it is drawing data in DWG format related to two-dimensional CAD, but the format is not limited. In this embodiment, a three-dimensional model is generated from this two-dimensional drawing data related to the design drawing as construction drawing data (particularly steel frame construction drawings in this embodiment) necessary for actual construction. Furthermore, the two-dimensional drawing data in this embodiment may include information about the members as metadata.

[0027] Construction drawings are drawings created based on design drawings that include detailed information for construction purposes (detailed structure, material model numbers and shapes, equipment information, etc.). In this embodiment, the construction drawings include a three-dimensional model showing the structure and two-dimensional drawings further generated from the three-dimensional model. In this embodiment, steel frame construction drawings related to the construction of steel structures are generated, but the generated drawings are not limited to steel frame drawings; drawings for other structures (e.g., reinforced concrete structures) may also be generated.

[0028] Furthermore, in architectural drawings, including design drawings and construction drawings, it is preferable that different types of elements (e.g., walls, beams, columns, etc.) are placed on different layers.

[0029] <System Configuration> Figure 4 is a block diagram showing the configuration of an AI drawing creation system according to one embodiment. As shown in Figure 4, the AI ​​drawing creation system 0 comprises a drawing creation device 1 and a user terminal 2, and each component is connected to communicate via a network NW. There may be one or more user terminals 2. The drawing creation device 1 may also be configured to be connectable via a network NW to an external server device storing a large-scale language model. In this embodiment, data transmission and reception are preferably performed using encryption to ensure secure communication.

[0030] The drawing creation device 1 is an information processing device for receiving input from a user terminal 2 and supporting design-related work, and comprises an acquisition unit 11, an inspection unit 12, a classification unit 13, a three-dimensional model generation unit 14, a fee calculation unit 15, and a display processing unit 16.

[0031] User terminal 2 is a terminal device operated by a user of the AI ​​drawing creation system 0. In this embodiment, the user of the system is a construction drawing creator engaged in the work of generating construction drawings. However, the AI ​​drawing creation system 0 may also be used by users who are not engaged in the work of generating construction drawings, such as those who create design drawings (e.g., architects). It is preferable that one user terminal 2 is operated by one user, but one user may operate multiple terminal devices, or one user terminal 2 may be operated by multiple users. In this case, it is preferable that the user operating user terminal 2 is identified by performing a prior login operation or the like.

[0032] User terminal 2 is a terminal device operated by the user, which connects to drawing device 1 by executing a drawing device utilization program such as a browser application, and performs predetermined processing. The drawing device utilization program may be a browser application pre-installed on the terminal device, or it may be a client application downloaded from a program provider device (not shown).

[0033] In this embodiment, the network NW is an IP (Internet Protocol) network, but there are no restrictions on the type of communication protocol, nor on the type or size of the network.

[0034] <Hardware Configuration> The hardware configuration of the drawing creation device 1 and user terminal 2 is shown below using Figure 5.

[0035] Figure 5(a) shows a hardware configuration diagram of a drawing creation device 1 in one embodiment. The drawing creation device 1 comprises a processing unit 101, a storage unit 102, and a communication unit 103 as its hardware configuration. In this embodiment, one or more computer devices such as a server device or a personal computer can be used as the drawing creation device 1. Note that the drawing creation device 1 is not limited to the configuration shown in Figure 5(a), as long as it can realize the functional elements described above as a whole.

[0036] The processing unit 101 is composed of one or more processors such as a CPU and a GPU (Graphics Processing Unit), and controls the overall processing in the drawing creation device 1 by executing application programs that enable the server device, etc., to function as the drawing creation device 1. The storage unit 102 is an HDD (Hard Disk Drive), SSD (Solid State Drive), flash memory, RAM (Random Access Memory), etc., and stores computer programs and various data. The communication unit 103 controls communication with the network NW and enables data communication with user terminals 2 and other devices.

[0037] Figure 5(b) shows the hardware configuration diagram of user terminal 2. User terminal 2 comprises a processing unit 201, a storage unit 202, a communication unit 203, an input unit 204, and an output unit 205 as its hardware configuration. In this embodiment, user terminal 2 can be a terminal device such as a smartphone, a personal computer, or a tablet terminal.

[0038] The processing unit 201 consists of one or more processors, such as a CPU, and controls the overall processing on the user terminal 2 by executing terminal programs, the OS, and other applications. The storage unit 202 is an HDD, SSD, flash memory, RAM, etc., and stores browser applications and various data. The communication unit 203 controls communication with the network NW and enables data communication with at least the drawing creation device 1. The input unit 204 is an input interface that accepts input operations from the user, and is composed of a microphone, touch panel, mouse, keyboard, etc. The output unit 205 consists of a display or the like for display output.

[0039] <Functional Configuration> The following describes the functional configuration of the drawing creation device 1 in this embodiment.

[0040] <Acquisition part 11> The acquisition unit 11 acquires various necessary data, such as two-dimensional drawing data input from the user terminal 2, and performs processing to pass it on to other functional configurations or to store it in the storage unit 102. The acquisition unit 11 may acquire the necessary data from the storage unit 102, or from other devices such as the user terminal 2. The two-dimensional drawing data is data relating to two-dimensional architectural drawings such as design drawings. The format of the two-dimensional drawing data may be DWG format or other formats. Furthermore, in the processing described later, if the accuracy of the data output by the AI ​​drawing creation system 0 does not meet the standard, data input from the user may be accepted. In addition, the format of the two-dimensional drawing data of the design drawings used to create drawings (construction drawings in this embodiment) in the system and the format of the two-dimensional drawing data of the generated construction drawings may be the same.

[0041] <Inspection Department 12> The inspection unit 12 inspects whether the acquired drawing (design drawing in this embodiment) is appropriate based on the two-dimensional drawing data. In this embodiment, the inspection unit 12 inputs the acquired two-dimensional drawing data into an inspection model and outputs an inspection result ("Pass," "Fail," etc.). The inspection model is a model for determining whether the target drawing (design drawing in this embodiment) is appropriate, and is a model that has been trained using combinations of drawings and inspection result labels attached to the drawings as training data. Furthermore, in order to recognize what specific violations the drawing has committed, the training data also includes labels indicating detailed violation items (for example, "insufficient spacing between members," "missing material markings," etc.) and annotations for the drawing areas corresponding to the violation locations, which are used for training. Therefore, the inspection unit 12 may input the two-dimensional drawing data into the inspection model and output an inspection result of "Pass" or "Fail," along with detailed reasons such as "insufficient spacing between members." The drawing type output by the classification unit 13 may also be used for outputting the inspection result. Furthermore, if the drawing creation device 1 determines that the input two-dimensional drawing data is "unacceptable," it may display the inspection results and / or detailed reasons to the user and prompt them to input appropriate two-dimensional drawing data. In addition, the inspection unit 12 may determine that a drawing is unacceptable if it is not layered (for example, a drawing in which all reference lines and figures indicating members are on the same layer). Moreover, the inspection unit 12 may determine that a drawing is unacceptable if it contains figures or text that could not be read in the processing described later.

[0042] Furthermore, the inspection unit 12 may determine an inspection result such as "pass" or "fail" based on whether the drawing meets the conditions, as part of the process related to the inspection of the drawing. In this case, the inspection unit 12 may also determine whether a set of plan views, a set of side views, and a set of material drawings are included. If the acquired two-dimensional drawing data does not include the drawing type, the inspection unit 12 may determine whether it is a proper drawing based on the classification result by the classification unit 13. In addition, the inspection unit 12 may determine that it is inappropriate if a specific type of drawing (for example, a material drawing) is not included, or if part of the drawing set is missing (for example, if there is a plan view for the second floor but not for the first floor). Furthermore, the storage unit 102 may store the judgment conditions for rule-based judgments based on such conditions.

[0043] Furthermore, the inspection unit 12 may determine whether the conditions are met based on the elements in the drawing (such as axes and shapes) and output the inspection result. In this case, the inspection unit 12 may determine that the drawing is "passed" if the conditions are met, such as whether the axes in the drawing are perpendicular or parallel, or whether the axis numbers are consecutive (for example, when "axis B" is directly followed by "axis C"). In this case, for the purpose of inspecting and analyzing the drawing, the drawing and / or text obtained from the drawing may be input into a large-scale language model.

[0044] <Classification section 13> The classification unit 13 inputs the acquired two-dimensional drawing data into a drawing classification model and performs a process to classify it into one of the drawing types. In this embodiment, the classification unit 13 performs a process to classify the two-dimensional drawing data, which relates to design drawings, into one of the drawing types: plan view, side view, or material view. However, the drawing types used in the classification are not limited to these. The classification unit 13 does not need to perform a process to assign drawing types to two-dimensional drawing data that has already been assigned drawing types in the form of metadata or text within the drawing.

[0045] The drawing classification model is a model for classifying drawings, and when two-dimensional drawing data relating to a drawing is input, it outputs a drawing type. In this embodiment, the drawing classification model acquires information about the figures in the drawing (drawing information), the quantity of the figures, the distribution of the figures, and the spatial discreteness based on the two-dimensional drawing data, and outputs a drawing type (in this embodiment, a score for each candidate drawing type) based on this data. In this embodiment, the drawing classification model filters out text and elements (figures) in the drawing that do not contribute to determining the drawing type. Furthermore, the learning of the classification model in this embodiment includes information about the figures contained in the drawing (length, feature vectors, position, quantity, etc.). In addition, the drawing classification model may be a model that has been trained using the two-dimensional drawing data itself and the combination of drawing types as training data, or it may be a model that has been trained using any data acquired from the drawing, such as text in the drawing. Furthermore, the classification unit 13 may use a large-scale language model as the drawing classification model to output the drawing type, or it may output the drawing type by combining the methods described above. Furthermore, if the score values ​​for each drawing type are below a threshold for all types, or if the scores for multiple types are close, and there is a high possibility that low-accuracy classification is being performed, the drawing creation device 1 may display the target drawing and prompt the user to input the drawing type.

[0046] <3D model generation unit 14> The three-dimensional model generation unit 14 generates a three-dimensional model of a structure based on the two-dimensional drawing data related to the acquired architectural drawings. In this embodiment, the three-dimensional model generation unit 14 identifies the structure of the structure from the acquired two-dimensional drawing data and generates a three-dimensional model. At this time, the three-dimensional model generation unit 14 acquires text information related to the text contained in the drawing and graphic information related to the shapes (polygons, lines, etc.) in the drawing based on the two-dimensional drawing data, identifies the combination of this text information and graphic information, and generates a three-dimensional model. The acquisition of graphic information and text information is not limited to the method described later, and known methods for acquiring text and graphic information may be used. A three-dimensional model is a model that represents a structure in three dimensions. In this embodiment, it is a three-dimensional steel frame model that represents the entire structure, but it may also be a model that represents a part of the structure (for example, a single steel frame or joint) in three dimensions. In this embodiment, the three-dimensional model generation unit 14 performs processing according to the type of drawing, but it may also perform processing to identify the structure of the structure, as described later, for all drawings. The AI ​​models used to generate the three-dimensional model (for example, element classification models or drawing recognition models) described later may be models composed of a single neural network or models composed of multiple neural networks. Furthermore, a single AI model (the three-dimensional model generation AI model) may perform some or all of the processing of the AI ​​models described later.

[0047] First, the three-dimensional model generation unit 14 uses a drawing recognition model to extract and classify elements such as shapes in the drawing and output drawing information. The three-dimensional model generation unit 14 first inputs two-dimensional drawing data into an element classification model to classify the shapes in the drawing by type. The element classification model is a semantic segmentation or object detection model based on deep learning, which is a model that has been trained using features related to the shape of the shape (contour, dimensions, angles) and position information (including relative position to other elements). When shape information including the features (contour, dimensions, angles) and position information of the target shape is input, it classifies the type of the shape (wall, column, beam, axis, etc.). The element classification model may also be a model that has been trained based on a drawing in which labels related to the type of shape are attached to the corresponding areas of the drawing (for example, the area showing the beam). Furthermore, if the shapes are described on separate layers for each type (for example, a beam layer and an axis layer), the element classification model may be a model that has been trained based on the layers and the types of elements described on those layers. Furthermore, to improve accuracy, the methods described above may be combined, such as making a decision based on the shape of the lines after making a decision based on the layer.

[0048] Geometric information refers to information about shapes in a drawing (polygons, lines, arcs, and axes), including information about the shape of the shape (outline, dimensions, number of corners, angle of the corners, type (e.g., quadrilateral)) and position information. Specifically, the position information of shapes may include the coordinates of the center point (centroid, etc.), the corners (vertices), the coordinates of the corners of a circumscribed rectangle, etc. Furthermore, the drawing information in this embodiment includes data indicating topological relationships such as connection relationships between shapes, inclusion relationships, intersection relationships, linking relationships (relationships between shapes connected by lines), and relative positional relationships, for the purpose of topology analysis. In addition, to determine the depth of the drawing, the geometric information may include data regarding the overlap of shapes (for example, that a quadrilateral is in front of a triangle when two triangles and quadrilaterals are overlapping).

[0049] The drawing recognition model is a model for recognizing elements in a drawing, and in this embodiment, it is a semantic segmentation or object detection model based on deep learning. In this embodiment, the drawing recognition model is a model that has been trained using data related to figures and their shapes (such as contours, dimensions, and corner features) and feature quantities related to figures such as location information and area as training data. The location information at this time may include the coordinates and center point of the bounding box that identifies the figure. The drawing recognition model may also acquire data related to topological relationships, such as the overlapping of drawings and text, as feature quantities related to figures or text. In addition, feature quantities obtained by convolving the drawing and information such as color at the pixel level may be acquired as feature quantities related to the drawing. At this time, the drawing information and text information include topological relationship information related to topological relationships. In this embodiment, the three-dimensional model generation unit 14 uses the drawing recognition model to acquire figure information and text information of figures classified as figures indicating members such as columns and beams in the element classification. The figure information and text information acquired from the drawing may be configured to be identifiable by being assigned identification information (such as an ID).

[0050] Furthermore, in the processing related to the output of drawing information, the three-dimensional model generation unit 14 performs a filtering process for shapes that represent elements other than structural members (for example, shapes that represent spaces such as rooms). In this embodiment, the AI ​​drawing creation system 0 is used to create steel frame construction drawings, so it judges shapes that represent elements other than steel frames as "noise," but the objects judged as noise are not limited to these. The three-dimensional model generation unit 14 also inputs two-dimensional drawing data into the filtering model and determines whether it is a "noise" shape. The filtering model is a model for identifying and filtering "noise" shapes, and in this embodiment, it is a model that uses data related to drawings in which areas corresponding to shapes are labeled as "noise" or "valid" as training data. Note that graphic information related to shapes may also be used in the training and filtering processes of the filtering model.

[0051] Furthermore, the three-dimensional model generation unit 14 recognizes text in the drawing and acquires text information based on the acquired two-dimensional drawing data. The text information is information about the text in the drawing and includes text acquired from the drawing (e.g., dimensions, model number, floor number) and text position information. At this time, the three-dimensional model generation unit 14 may acquire text information using a model such as an OCR (Optical Character Recognition) model that can recognize characters from images using a neural network, or if the two-dimensional drawing data contains information about the text in the form of metadata, it may acquire the information about the text as text information. In addition, the three-dimensional model generation unit 14 may perform processing using a model that has learned the correspondence between text and figures representing members, or a model that has learned based on rules regarding that correspondence (for example, the number placed between the auxiliary lines passing through the ends of a rectangle representing a column is the length of that column). If information about members (e.g., dimensions, parameters, etc.) is registered in advance as metadata, such metadata may be acquired as information about members instead of text information. Furthermore, if information about the shape is registered as metadata, information about the component representing the shape may be obtained from the metadata and information such as the scale of the drawing.

[0052] Furthermore, the three-dimensional model generation unit 14 may perform processing to read the distance between grid lines (the interval between lines that serve as the basis for the building's plan) and the floor height (the interval between lines that serve as the basis for the building's height) from the drawing. The floor height may also be determined by processing related to the division of floors in the side view, which will be described later. Furthermore, the dimensions of the target member may be determined based on the ratio of the size of a figure representing the target member to the ratio of the length of a figure whose length is known, such as the ratio of the length of the reference lines such as the distance between grid lines and the floor height to the length of the side of the figure representing the member. Note that these reference distances between grid lines and floor heights may be obtained based on text written in the drawing, or they may be obtained by comparing them with figures representing other members whose size is known.

[0053] The three-dimensional model generation unit 14 performs the above-described processing related to the recognition of shapes and acquisition of text in two-dimensional drawing data relating to a drawing of type "material drawing," and generates member information related to the member. The member information is information relating to the members that constitute a structure, and in this embodiment, it includes the model number, information indicating the shape (contour, member parameters (features related to height, width, depth, and angles), etc.). Based on the position of the shapes and / or text, the three-dimensional model generation unit 14 groups shapes and / or text that indicate the same member, for example, belonging to the same row, the same column, or the same table, and generates member information based on the shape information and text information relating to the shapes and text belonging to a specific group. Furthermore, in this embodiment, the three-dimensional model generation unit 14 also performs processing related to the grouping of shapes and / or text based on data relating to topology.

[0054] In the material drawing, diagrams showing components and text describing them (part numbers, symbols, etc.) are presented in a table format, as shown in Figure 3. Therefore, the three-dimensional model generation unit 14 first recognizes items in the drawing (e.g., titles) based on the recognized text information, identifies the positions of the figures and text in the table, and groups the figures belonging to the same table column based on these positions. In this embodiment, the three-dimensional model generation unit 14 identifies combinations of symbols and part numbers written as text in the material drawing. Based on these symbols, the three-dimensional model generation unit 14 combines data obtained from the material drawing (part numbers, etc.) with data obtained from other drawings such as plan views and side views (dimensions such as length) to generate component information. In order to combine descriptions of the same component written in different drawings, methods such as using text information such as symbols and part numbers, as described later, or methods using positions in the drawings (such as using the positions of components in side views and plan views of the same hierarchical level to identify corresponding components) are used.

[0055] In this embodiment, the model number is indicated in the drawing (material drawing, etc.) as a string of characters such as H-400×200×8×12. In this example, the initial symbol H indicates the type of steel material (H for H-beam, □ for square steel pipe, etc.), 400 indicates the height (400 mm), 200 indicates the flange width (200 mm), 8 (8 mm) indicates the web plate thickness, and 12 (12 mm) indicates the flange plate width. Therefore, in this embodiment, the shape of the target member (such as the cross-sectional shape in the example above) is determined by the type of steel material (H-beam, square steel pipe, etc.) and the string of characters indicating the model number that follows. Thus, the model number includes numerical values ​​related to the shape of the member according to predetermined notation rules, and the three-dimensional model generation unit 14 performs processing related to the acquisition of feature quantities related to the shape of the member according to the model number notation rules described above.

[0056] For the processing described above, the drawing creation device 1 may store feature acquisition rules corresponding to the description rules of the model number, such as which number in the string represents the length of which part (e.g., the first number in the string is the height), for each type of member (e.g., H-beams, square steel pipes, etc.). In this case, the three-dimensional model generation unit 14 may decide which member acquisition rule to adopt based on the text information regarding the text of the model number (e.g., if the string starts with "H", it may decide to adopt the acquisition rule for H-beams), and use this determined acquisition rule to generate member information (especially acquisition of feature quantities indicating the type and shape of the member). Note that the description rules for each model number may also be used to further determine which acquisition rule to adopt. Furthermore, the three-dimensional model generation unit 14 may use a trained model, which has been learned based on combinations of the member's model number string and the member's feature quantities (and / or three-dimensional model of the member), to determine which acquisition rule to adopt and acquire feature quantities indicating the type and shape of the target member from the string (text information) indicating the model number. Furthermore, a language model that has been pre-trained on combinations of part number notation rules and corresponding acquisition rules, or that has been pre-inputted with this data, may be used to determine the corresponding acquisition rules and acquire features.

[0057] In this embodiment, since some feature quantities, such as the length of a component, are not determined from the string indicating the model number, the remaining feature quantities indicating the shape of the component, such as the length of the component, are obtained from drawings other than the material drawing (such as a plan view or side view). At this time, the three-dimensional model generation unit 14 obtains numerical values ​​placed near the figure indicating the target component, or numerical values ​​written between the auxiliary lines enclosing the figure indicating the component, based on the positional information of the figure and text. The three-dimensional model generation unit 14 may also identify the text information indicating the feature quantities of the target component by performing the matching process between the figure information and text information described above, or it may obtain the feature quantities of the target component based on feature quantity acquisition rules corresponding to the drawing notation rules (for example, the length of the component is written between the auxiliary lines enclosing the figure indicating the target component). Furthermore, if there are no numerical values ​​that can be directly used as feature quantities of the target component described in the drawing, the three-dimensional model generation unit 14 may obtain the feature quantities of the target component based on the ratio of the length of other figures that can specify its dimensions, such as a figure indicating a reference length like a ruler, or a component whose length is known and placed near the target component, and the dimensions of the other figures.

[0058] In this embodiment, the component information obtained from the material drawing is used as a master of components (a master of components for each part number) in other drawings (plan view and side view). In the process described later, the component information for each part number obtained from this component list is combined with the component information from other drawings (plan view and side view) to generate component information related to the three-dimensional shape of each component and generate a three-dimensional model. Note that if the part number is directly indicated in the plan view or side view, the material drawing does not need to be included in the set of drawings input into the system.

[0059] Furthermore, information regarding ancillary construction work may be obtained from the design drawings. In this embodiment, information regarding the base materials for roofs and walls, described in text, is obtained from a drawing called a finishing schedule included in the design drawings and used in the process related to the generation of a three-dimensional model, which will be described later. At this time, the three-dimensional model generation unit 14 generates a three-dimensional model by placing members such as steel materials for fixing the base materials at the positions where the base materials are to be placed, based on the information about the base materials obtained from the text on the drawings (for example, "a base material will be installed here").

[0060] In this embodiment, component information is generated using graphic and text information (for example, dimensions and model numbers of figures indicating components) obtained from plan and side views included in the design drawings. In this embodiment, the three-dimensional model generation unit 14 obtains parameters such as the model number of each column and beam, the cross-section of the component (column and beam), and the length of the beam based on the text written in the plan view, such as the floor plan. The three-dimensional model generation unit 14 also obtains parameters such as the cross-sectional size of the component (column and beam), and the height and position of the component based on the text written in the side view, such as the frame drawing. The three-dimensional model generation unit 14 obtains the dimensions and joining method of each component (steel material) for each model number from the component list as component information. The three-dimensional model generation unit 14 may also obtain dimensions and shapes of specific parts as component information from text written in the detailed drawings. The AI ​​model used to obtain component information from drawings may be a model that obtains component information according to rules in architectural drawings, such as text written between auxiliary lines indicating the length of the figure between those auxiliary lines. Graphics indicating the shape of a component, such as the cross-sectional shape, may also be obtained as component information.

[0061] The three-dimensional model generation unit 14 matches shapes and text based on two-dimensional drawing data relating to drawings of type "plan view" or "side view," and identifies the part number of the component represented by the shape from the corresponding text. At this time, processing related to the identification of part numbers is performed based on the text corresponding to the shape whose type is a component (for example, a column, beam, bolt, etc.). In the processing related to the generation of the three-dimensional model described later, the component information corresponding to this identified part number is used as information about the shape of the component represented by the target shape. At this time, text that is highly likely to be a part number may also be identified in advance, and other text may be filtered out. At this time, a trained model that has been learned based on the part number text may be used, or a large-scale language model may be used.

[0062] In this embodiment, the three-dimensional model generation unit 14 uses text indicating the part number and / or code of a part to identify the part number and / or code of the part indicated by the figure and performs a process to acquire feature quantities related to the shape of the part. However, other text (e.g., text indicating dimensions) may also be used for matching to acquire feature quantities of the part indicated by the figure (e.g., height, width, etc.). In this embodiment, text containing alphanumeric characters is used to determine what kind of content the text indicates, such as text indicating a part number or hierarchy, and text containing only numbers or numbers followed by units is used to determine parameters indicating dimensions. This determination may be made by a rule-based AI model, by an AI model trained on the type of text and the content it indicates (part number, dimensions, etc.), or by using a pre-trained large-scale language model. Since text describing a specific figure is likely to be placed near that figure in the drawing, the three-dimensional model generation unit 14 performs matching based on the positional information of the figure and text in the drawing. In this case, figures recognized as indicating a part are used for matching, and figures recognized as auxiliary lines or noise are not used. In this embodiment, the three-dimensional model generation unit 14 determines the relationship between a figure and text by constructing a bipartite graph in which each figure and text are nodes and their corresponding relationships are edges, and solving minimum weight matching. However, the matching method is not limited to this method. In this embodiment, the weight of an edge is calculated based on distance and intersection with an axis. In the matching process described above, if a figure is matched with text indicating multiple part numbers, or if the matching cannot be performed properly, the three-dimensional model generation unit 14 may identify the corresponding part and / or part number by inputting the three-dimensional data of the part and neighboring parts, the text of part numbers and / or codes with high weights, and part information extracted from material drawings into a large-scale language model.

[0063] Furthermore, the three-dimensional model generation unit 14 identifies the hierarchy of the plan view and the side view in order to align the members in the side view and the plan view. In this embodiment, the three-dimensional model generation unit 14 identifies the hierarchy indicated by the target plan view based on the two-dimensional drawing data of the drawing type "plan view". At this time, the three-dimensional model generation unit 14 makes a determination based on the text in the drawing, for example, determining that it is a plan view of the first floor if the text "1F" is included in the drawing. If it is not possible to determine from the text in the drawing, the three-dimensional model generation unit 14 outputs the hierarchy by inputting the text information obtained from the two-dimensional drawing data, topology space features (for example, the relative positions of the plan view and the side view), building hierarchy command rules and knowledge graphs related to those command rules into a large-scale language model. Alternatively, the three-dimensional model generation unit 14 may identify the hierarchy by inputting the two-dimensional drawing data into a large-scale language model. Furthermore, if the identified hierarchy is inappropriate, such as when the same hierarchy (e.g., the first floor) is output for multiple floor plans, the AI ​​drawing creation system 0 may consider the hierarchy identification to have failed, display a hierarchy input screen, and prompt the user to input information about the hierarchy for each drawing.

[0064] Furthermore, the three-dimensional model generation unit 14 divides the drawing type "side view" into layers. At this time, the three-dimensional model generation unit 14 obtains layer elevation lines for dividing the side view into layers based on the two-dimensional drawing data, and divides the side view using these layer elevation lines, thereby dividing it into layers and generating side sub-views, which are side views of each layer (for example, the side view of the first floor). At this time, the layer may be specified for each side sub-view, such as the side sub-view of the first floor from the bottom of the side sub-view.

[0065] In this embodiment, the three-dimensional model generation unit 14 inputs line elements that do not constitute a figure into the hierarchical elevation line determination model and identifies line elements that are likely to be hierarchical elevation lines, which are line elements that divide the hierarchies. By inputting information about these line elements that are likely to be hierarchical elevation lines into a hierarchical elevation line selection model that has been pre-trained using two-dimensional drawing data labeled with line elements indicating hierarchical elevation lines and the number of floors k of the building as training data, k or k+1 hierarchical elevation lines are obtained. In this embodiment, the number of floors k of the structure is data obtained from the floor plan (for example, if there are 14 floor plans, it is 14 floors), but it may also be data entered by the user.

[0066] If the hierarchical elevation lines cannot be obtained using the method described above, the three-dimensional model generation unit 14 may use a clustering algorithm such as the K-Means algorithm to group the figures in the side view and determine the boundaries of each cluster as the hierarchical elevation lines. Furthermore, if the hierarchical elevation lines cannot be obtained even after the clustering process described above, the AI ​​drawing creation system 0 will display the side view and prompt the user to input the hierarchical elevation lines.

[0067] The three-dimensional model generation unit 14 identifies the correspondence between figures in the plan view and the side view based on the hierarchy identified in the above-described process. In this embodiment, the three-dimensional model generation unit 14 performs a process to identify the correspondence between figures belonging to the same hierarchy in the plan view and the side view (side sub-view in this embodiment). At this time, the drawing information and text information related to the corresponding figures are integrated as information indicating the same member. In addition, in this embodiment, the three-dimensional model generation unit 14 identifies the correspondence between figures in the plan view and figures in the side view by constructing a bipartite graph of the figures in the plan view and the figures in the side view and performing matching. At this time, the three-dimensional model generation unit 14 uses a bipartite graph in which the polygons of the plan view and the polygons of the side view are each nodes, and the weight of the edges is calculated based on the distance in the relative coordinates to perform matching. In this embodiment, by obtaining information on the corresponding figures in the side view, information on the height of the member indicated by the figure in the plan view and the relative floor height is obtained.

[0068] Furthermore, if there are members in the drawing (a plan view in this embodiment) whose correspondence relationship has not been identified even after the above-described process is completed, the three-dimensional model generation unit 14 may obtain information regarding the height of the member and the floor height relative to it by inputting information about other members that have already been matched with the topology relationship (positional relationship, etc.) of the target member into the member height determination model. In addition, if there are members whose part numbers cannot be identified, the three-dimensional model generation unit 14 may identify the part number to which it belongs by inputting member information generated based on data about the shape of the member and neighboring members, text information about the member, material drawings, etc. into a large-scale language model.

[0069] The three-dimensional model generation unit 14 generates member information based on various data such as member information generated in the above-described process (for example, positional relationships, shapes, parameters, etc. of members), and generates a three-dimensional model by placing the member models based on the generated member information in appropriate positions. Specifically, the three-dimensional model generation unit 14 in this embodiment generates a three-dimensional model based on data such as the shape of a member identified based on one or more of the member's model number, member description text, and graphic information, positional information of the graphic representing the member, and positional relationships of members identified based on the positional relationships between the graphic representing the member on the drawing. At this time, the three-dimensional model generation unit 14 generates a three-dimensional model of the structure by placing the three-dimensional model for each member, which is composed of member parameters identified from the model number and parameters based on text obtained from the plan view and side view, in corresponding positions in the coordinate system from the plan view and side view. At this time, the coordinate system when placing the members is generated based on the axes in the plan view and / or side view, the dimensions attached to the axes (for example, the width of a room), the grid line numbers and their positional relationships, the distance between column centers, etc. A three-dimensional model of the structure is generated by placing members at corresponding positions in this coordinate system. The generated three-dimensional model may also be stored in the storage unit 102 in association with the user information of the user who created the three-dimensional model (registering two-dimensional drawing data and instructing the creation of the three-dimensional model). If parameters cannot be obtained from text, the parameters of the member represented by the figure may be obtained from the drawing scale and the characteristic quantities of the figure itself (e.g., side lengths). In addition, some or all of the parameters of the member may be obtained from a plan view or side view instead of a material drawing.

[0070] Furthermore, in the processing described above, such as the classification of drawing types, the classification of shapes, and the extraction of member parameters, the three-dimensional model generation unit 14 may generate output by inputting two-dimensional drawing data and the target (e.g., graphic information, text information, etc.) into a large-scale language model.

[0071] Figure 6 shows an example of a generated three-dimensional model. As shown in Figure 6, the three-dimensional model generation unit 14 outputs a three-dimensional model that shows the structure of the structure.

[0072] <Price Calculation Section 15> In this embodiment, the user of the AI ​​drawing creation system 0 is charged according to the three-dimensional model generated. Therefore, the fee calculation unit 15 calculates the fee based on the generated three-dimensional model.

[0073] In this embodiment, the fee is calculated based on the weight of the steel frame included in the generated three-dimensional model. Therefore, the fee calculation unit 15 calculates the fee based on the volume obtained from the dimensions of the members included in the three-dimensional model. If different materials are used for each member, the weight may be calculated for each member. The fee calculation unit 15 may also calculate the fee based on the total floor area of ​​the target structure. In this embodiment, since the fee is calculated based on a three-dimensional model created by the user, the fee is calculated based on the three-dimensional model registered in association with the user information of the target user. The fee calculation unit 15 may also calculate the usage fee for the system during a specific period based on the three-dimensional model generated during that period. In this case, the three-dimensional model is stored in the storage unit 102 along with information regarding the date and time the three-dimensional model was created.

[0074] Furthermore, the fee may be calculated on an organizational basis to which the user belongs. In this case, the user information about the user may include organizational information about the organization to which the user belongs, and the fee calculation unit 15 may calculate the fee for the three-dimensional model generated by the user for each organization based on this organizational information.

[0075] <Display Processing Unit 16> The display processing unit 16 performs processing to display a screen on a computer device such as the user terminal 2 based on the various necessary data, such as the generated three-dimensional model. In addition, if the data output by the AI ​​drawing creation system 0 is unsuitable, the display processing unit 16 may display a screen for inputting alternative data.

[0076] <Processing flow> The following describes the processing flow in one embodiment using Figure 7.

[0077] The acquisition unit 11 acquires two-dimensional drawing data relating to the drawings of the structure to be designed (S101). The inspection unit 12 inspects whether the drawings are appropriate based on the acquired two-dimensional drawing data (S102). If the inspection by the inspection unit 12 determines that the drawings are acceptable, the classification unit 13 classifies the drawings by type based on the two-dimensional drawing data of the drawings (S103). In this case, for drawings whose type could not be classified or whose classification accuracy was not high, the display processing unit 16 may display a drawing type input screen to prompt the user to input the drawing type for each drawing, and the acquisition unit 11 may acquire the drawing type entered by the user.

[0078] The three-dimensional model generation unit 14 acquires graphic information and text information related to the figures in the drawing based on the two-dimensional drawing data (S104). The three-dimensional model generation unit 14 generates member information based on the two-dimensional drawing data of a drawing whose drawing type is "material drawing" (S105). The three-dimensional model generation unit 14 performs matching of figures and text based on the position information of the figures and the position information of the text (S106) and identifies the model number of the member indicated by the figure. In this embodiment, the process for identifying the model number of the member is performed for the plan view and the side view. The three-dimensional model generation unit 14 performs a process to identify the hierarchy of the plan view based on the two-dimensional drawing data of the plan view and a process to identify the hierarchy of the side view based on the two-dimensional drawing data of the side view (S107). Regarding the identification of the hierarchy of the plan view and the side view, if the hierarchy cannot be identified, the display processing unit 16 may display the plan view and / or side view to prompt input of information regarding the hierarchy, and the acquisition unit 11 may acquire the information regarding the hierarchy. The three-dimensional model generation unit 14 performs a process to identify the correspondence between figures included in the plan view and figures included in the side view based on the identified hierarchy (S108). The three-dimensional model generation unit 14 generates a three-dimensional model based on structural data of the structure obtained from the two-dimensional drawing data. The display processing unit 16 displays the generated three-dimensional model on the user terminal 2. The drawing creation device 1 may also further include a two-dimensional drawing generation unit to convert the generated three-dimensional model into two-dimensional construction drawing data. In this case, the two-dimensional drawing generation unit identifies a viewpoint in the two-dimensional drawing and performs a process to project the objects in the three-dimensional model as if viewed from that viewpoint. Any known method may be used to convert the three-dimensional model to two-dimensional. Furthermore, at this time, an AI model that has learned about the format of construction drawings, etc., by learning based on a set of three-dimensional models and construction drawings may be used for conversion to two-dimensional drawings. In addition, at this time, multiple drawings such as a top view and a side view of the structure may be generated as construction drawings.

[0079] Furthermore, data related to construction drawings, such as three-dimensional models and two-dimensional construction drawing data, may be sent via email or other means to the general contractor or other party who requested the creation of the construction drawings. Alternatively, instead of three-dimensional models or two-dimensional construction drawing data, a URL to a webpage for viewing that data may be sent via email. In this case, the AI ​​drawing creation system 0 may identify when an event has occurred, such as when the email is opened or when the user is redirected to a webpage for viewing construction drawing data via a link attached to the email. The AI ​​drawing creation system 0 may also be configured to notify or display to the user who created the data related to the target construction drawing (three-dimensional model, two-dimensional construction drawing data) that an event related to viewing this data has occurred.

[0080] The inspection results from the inspection unit 12 may be displayed on the user terminal 2. At this time, the user will be shown the reason for the failure (for example, the drawing is not layered). Furthermore, if a drawing containing a figure whose graphic information could not be read, or a drawing containing a component whose component information could not be read, is deemed unacceptable, the drawing creation device 1 may display a screen on the user terminal 2 showing the figures whose information (graphic information / component information) could not be read. At this time, the screen showing the figures whose information could not be read may distinguish between figures whose information could be read (polygons, lines, etc.) and figures whose information could not be read by coloring them in different colors.

[0081] Furthermore, the AI ​​drawing creation system 0 may include a consistency verification unit that checks the generated three-dimensional model and / or construction drawing data. In this case, the consistency verification unit may detect collisions between members in the three-dimensional model and notify the user. Any known method may be used to detect member collisions. In this case, the three-dimensional model may be displayed in a way that makes the collision locations identifiable, such as by coloring the collision locations. In addition, the consistency verification unit may identify problems and notify the user of the problems using a model that has been learned based on the correspondence between labels related to problem items in the three-dimensional model and / or two-dimensional construction drawing data and the drawing areas (or feature quantities in those areas) related to the problems.

[0082] Furthermore, if there is a defect in the shape of the component, the drawing creation device 1 may accept input regarding component information constituting the three-dimensional model via the user terminal 2. [Explanation of symbols]

[0083] 0 AI drawing creation system 1. Drawing creation device 11 Acquisition Department 12. Inspection Department 13 Classification section 14. Three-dimensional model generation unit 15. Price Calculation Section 16 Display Processing Unit 2 User terminals NW Network

Claims

1. An AI drawing creation system, It comprises an acquisition unit and a three-dimensional model generation unit, The acquisition unit acquires two-dimensional drawing data relating to a two-dimensional architectural drawing, which includes figures and / or text indicating the components. The aforementioned architectural drawings include a floor plan and a side view, The three-dimensional model generation unit identifies the hierarchy of the plan view and side view included in the architectural drawing based on the two-dimensional drawing data, identifies the correspondence between the figures in the plan view and side view based on the identified hierarchy, integrates the figure information and text information of the corresponding figures to generate member information, and generates a three-dimensional model based on the member information. AI drawing creation system.

2. An AI drawing creation system, It comprises an acquisition unit, a three-dimensional model generation unit, and a two-dimensional drawing generation unit, The acquisition unit acquires two-dimensional drawing data relating to a two-dimensional architectural drawing, which includes figures and / or text indicating the components. The three-dimensional model generation unit generates member information based on the two-dimensional drawing data, A three-dimensional model is generated based on the aforementioned member information. The two-dimensional drawing generation unit generates a two-dimensional construction drawing based on the three-dimensional model. AI drawing creation system.

3. The three-dimensional model generation unit obtains text information related to text and graphic information related to shapes from the two-dimensional drawing data, and performs matching of shapes representing members with text. Based on the text information and graphic information relating to the matched shapes and text, the three-dimensional model is generated. The AI ​​drawing creation system according to claim 1 or 2.

4. The three-dimensional model generation unit performs matching of the figures and the text based on the positional information of the figures and the positional information of the text obtained from the two-dimensional drawing data, as part of the matching process. The AI ​​drawing creation system according to claim 3.

5. The aforementioned architectural drawings include material drawings, The three-dimensional model generation unit generates member information, including member parameters and part numbers, based on the two-dimensional drawing data relating to the material drawing, and generates the three-dimensional model based on the member information. The AI ​​drawing creation system according to claim 1 or 2.

6. The aforementioned architectural drawings include a floor plan and a side view, The three-dimensional model generation unit further identifies the code and / or model number of the member indicated by the figure included in the two-dimensional drawing data based on text information obtained from the two-dimensional drawing data relating to the plan view and / or side view. Based on the identified code and / or model number of the component information, the three-dimensional model is generated. The AI ​​drawing creation system according to claim 5.

7. The three-dimensional model generation unit performs a process to identify correspondences between figures belonging to the same hierarchical level in the plan view and the side view. The AI ​​drawing creation system according to claim 1.

8. The AI ​​drawing creation system further includes a fee calculation unit, The aforementioned fee calculation unit determines the volume and / or weight of the members included in the structure based on the three-dimensional model and calculates the fee. The AI ​​drawing creation system according to claim 1 or 2.

9. A computer-based AI drawing creation method, It comprises an acquisition process and a three-dimensional model generation process, In the acquisition process described above, two-dimensional drawing data relating to a two-dimensional architectural drawing is acquired, which includes figures and / or text indicating the components. The aforementioned architectural drawings include a floor plan and a side view, In the three-dimensional model generation process, the hierarchy of the plan view and side view included in the architectural drawing is identified based on the two-dimensional drawing data, the correspondence between the figures in the plan view and side view is identified based on the identified hierarchy, the figure information and text information of the corresponding figures are integrated to generate member information, and a three-dimensional model is generated based on the member information. AI drawing creation method.

10. A computer-based AI drawing creation method, It comprises an acquisition process, a three-dimensional model generation process, and a two-dimensional drawing generation process. In the acquisition process described above, two-dimensional drawing data relating to a two-dimensional architectural drawing is acquired, which includes figures and / or text indicating the components. In the three-dimensional model generation process, component information is generated based on the two-dimensional drawing data. A three-dimensional model is generated based on the aforementioned member information. In the two-dimensional drawing generation process, a two-dimensional construction drawing is generated based on the three-dimensional model. AI drawing creation method.

11. An AI drawing creation program that causes the computer to execute the AI ​​drawing creation method described in claim 9 or 10.

Images