A method of reviewing architectural drawings, a storage medium and an electronic device

Abstract

The present application relates to a kind of methods for examining architectural drawings, which comprises: obtaining the architectural drawings to be examined;Frame detection is carried out on the architectural drawings to be examined to obtain the drawings of roof frame;The drawings of roof frame are input into the pre-trained drawing roof component information extraction model to obtain the output result;Wherein, the output result includes roof building topological relationship;Based on the roof building topological relationship, the space mapping from roof to top plane is carried out to obtain the space mapping result;Based on the output result and the space mapping result, the roof frame drawings are subjected to roof specification examination to obtain the roof specification examination result, so that not only roof drawing examination can be realized, but also drawing roof identification, positioning accuracy can be improved.

Description

Technical Field

[0001] This invention relates to the field of architectural drawing review technology, and in particular to a method, storage medium, and electronic device for reviewing architectural drawings. Background Technology

[0002] Roofing plans are a major part of residential building construction. They include the materials and equipment used in the project, as well as the technical activities such as design, construction, and maintenance. They also refer to the object of the project and its functional protection role. Specifically, in addition to safely bearing various loads, roofing projects need to be able to resist temperature, wind, rain, snow, and even earthquake damage, as well as withstand deformation caused by temperature differences and the expansion, contraction, and cracking of the base structure.

[0003] Therefore, roofing plays a crucial role in creating a safe, environmentally friendly building that meets both practical and aesthetic requirements. Roofing is a sub-project of building construction and a large engineering field. It encompasses the roof slab and all structural layers above it, including vapor barriers, ventilation and moisture-proof layers, thermal insulation layers, waterproofing layers, and protective layers. It is a comprehensive system engineering project reflecting the multifunctional functions of the roof. National standards for roof waterproofing and insulation are reviewed, and the analysis of roof architectural drawings provides a foundation for drainage and professional roof drainage design.

[0004] Therefore, there is an urgent need for a method to review roof drawings. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a method, storage medium and electronic device for reviewing architectural drawings, which can realize the review of roof drawings.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the main technical solutions adopted by the present invention include:

[0009] In a first aspect, embodiments of the present invention provide a method for reviewing architectural drawings, the method comprising: acquiring architectural drawings to be reviewed; performing frame detection on the architectural drawings to be reviewed to obtain a roof frame drawing; inputting the roof frame drawing into a pre-trained roof component information extraction model to obtain an output result; wherein the output result includes roof building topology; performing spatial mapping from the roof to the top floor plane based on the roof building topology to obtain a spatial mapping result; and performing roof specification review on the roof frame drawing based on the output result and the spatial mapping result to obtain a roof specification review result.

[0010] In one possible embodiment, the model architecture of the drawing house component information extraction model includes an image processing layer, a text feature parsing layer, and a first convolutional layer;

[0011] The training process of the roof component information extraction model includes: acquiring training sample data; the training sample data includes training sample drawings and training sample text describing the roof's topological relationships in the training sample drawings; processing the training sample drawings through an image processing layer to obtain information decoding results; processing the training sample text through a text feature parsing layer to obtain ordered text feature vectors; and convolving the information decoding results and ordered text feature vectors through a first convolutional layer to obtain training output results; the training output results include training results for the roof space mask image, training results for basic component information, and training results for roof topological relationships.

[0012] In one possible embodiment, the image processing layer includes multiple convolutional layers, a first feature fusion layer, a second convolutional layer, an abstract feature extraction layer, and a first feature pyramid layer;

[0013] In this multi-convolutional layer, each convolutional layer is used to perform corresponding convolution processing on the training sample drawings; the first feature fusion layer is used to perform feature fusion on the convolution results output by each convolutional layer; the second convolutional layer is used to perform convolution processing on the feature fusion results output by the first feature fusion layer; the abstract feature extraction layer is used to remove redundant lines in the roof and extract the roof abstract features based on the convolution results of the second convolutional layer; and the first feature pyramid layer is used to perform information decoding operations on the roof abstract features to obtain the information decoding operation results.

[0014] In one possible embodiment, the multiple convolutional layers include a third convolutional layer, a first dilated convolutional layer, a second dilated convolutional layer, and a third dilated convolutional layer arranged in parallel.

[0015] In one possible embodiment, the hole coefficient of the first dilated convolutional layer is less than that of the second dilated convolutional layer, and the hole coefficient of the second dilated convolutional layer is less than that of the third dilated convolutional layer, and the hole coefficients of the first dilated convolutional layer, the second dilated convolutional layer, and the third dilated convolutional layer are all less than 64.

[0016] In one possible embodiment, the hole coefficients of the first, second, and third dilated convolutional layers are all even numbers.

[0017] In one possible embodiment, the abstract feature extraction layer includes a second feature pyramid layer, n parallel feature layers, and a second feature fusion layer for fusing the output of the second feature pyramid layer and the output of each of the n parallel feature layers. Each of the n feature layers includes a fourth convolutional layer, a batch normalization layer, and an activation function layer.

[0018] In one possible embodiment, the text feature parsing layer includes an input layer, a cue learning layer, and a text decoding layer;

[0019] The input layer is used to obtain the input training sample text; the cue learning layer is used to extract the text feature vector from the training sample text; and the text decoding layer is used to decode and compile the text feature vector to obtain the ordered text feature vector.

[0020] Secondly, embodiments of this application provide a storage medium storing a computer program, which, when executed by a processor, performs the method described in the first aspect or any optional implementation thereof.

[0021] Thirdly, embodiments of this application provide an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, they perform the method described in the first aspect or any optional implementation of the first aspect.

[0022] Fourthly, this application provides a computer program product that, when run on a computer, causes the computer to perform the method in the first aspect or any possible implementation thereof.

[0023] (III) Beneficial Effects

[0024] The beneficial effects of this invention are:

[0025] This application provides a method, storage medium, and electronic device for reviewing architectural drawings. The method involves acquiring the architectural drawings to be reviewed, performing frame detection on the drawings to obtain the roof frame, and inputting the roof frame into a pre-trained model for extracting roof component information to obtain output results. The output results include the roof's architectural topology, a spatial mapping from the roof to the top floor plane based on the roof's topology, and a roof specification review based on the output results and the spatial mapping results. This not only enables roof drawing review but also improves the accuracy of roof identification and positioning in the drawings.

[0026] To make the above-mentioned objectives, features and advantages to be achieved by the embodiments of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A flowchart illustrating a method for reviewing architectural drawings provided in an embodiment of this application is shown;

[0029] Figure 2 This illustration shows a schematic diagram of a drawing house component information extraction model provided in an embodiment of this application. Detailed Implementation

[0030] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] The numerous and complex lines on the roof drawings result in a low accuracy rate for identifying roof components and understanding spatial topological relationships.

[0032] Currently, classic object detection models (e.g., YOLOv2 and Solov2), semantic segmentation models (e.g., UNet and DeepLabv3+), and instance segmentation models (e.g., SAM and YOLOv2) are used to understand objects in the visual scene of architectural drawings. Understanding these objects has always been a major driving force for architectural business logic. However, as mentioned above, these classic algorithms cannot learn abstract architectural business logic through prompts; they can only be trained using semantic labels defined by single location coordinates. For example, a roof might have elevators, stairwells, and doors. Although classification and recognition using detection and segmentation algorithms are effective, the task definition of a roof drawing significantly simplifies the understanding of architectural logic itself, because a roof object, in addition to its semantic category, also... Architectural logic can be described from many other aspects. For example, a door can lead to a "staircase" and an "elevator," and a stairwell can have "multiple steps" and "platforms," ​​etc. Therefore, learning architectural logic space and construction attributes can supplement category-level recognition and topological relationships, thereby obtaining a more comprehensive and granular visual perception of roof data. However, relying on current technology to identify space using segmentation techniques, identify components using object detection methods, and then process the topological relationships between components and space using traditional methods cannot achieve end-to-end model parsing (i.e., existing technologies require at least two models, while the "end-to-end" in this context refers to extracting a model from the roof component information in the drawings), resulting in slow product deployment.

[0033] To improve the accuracy of roof identification and positioning in drawings, this application combines a visual language model pre-trained on massive multimodal roof drawing data with prompt-based learning using available component detection and spatial attribute topological relationship recognition data. Furthermore, the forward propagation of the network training process uses continuous functions of an integral neural network to represent weights, replacing the traditional discrete weight tensor representation of the forward propagation network layer weights with continuous integral operations. This significantly improves the real-time performance of roof drawing parsing. The combination of training prompt-based learning to understand component and spatial attributes greatly promotes the learning of roof spatial topological relationships. The abstract logic of roof building business greatly improves the identification of small target components and spaces in the roof drawing frame, achieving the establishment of an end-to-end roof information parsing topological map. This network structure is not only applicable to roofs but also to computer rooms and attics drawing information parsing, and also provides an information database for the review of building roof specifications, facilitating roof specification review and information verification.

[0034] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be understood more clearly and thoroughly, and that the scope of the present invention can be fully conveyed to those skilled in the art.

[0035] Please see Figure 1 , Figure 1 A flowchart illustrating a method for reviewing architectural drawings according to an embodiment of this application is shown. It should be understood that this method for reviewing architectural drawings can be executed by an electronic device, and the specific device of the electronic device can be configured according to actual needs; this embodiment is not limited thereto. For example, the electronic device can be a computer or a server, etc. Specifically, the method for reviewing architectural drawings includes:

[0036] Step S110: Obtain the architectural drawings to be reviewed. These architectural drawings do not refer to paper images; they can be electronic drawings such as CAD drawings. Furthermore, the architectural drawings can be electronic drawings of a specific building, and the architectural image includes the roof area.

[0037] Step S120: Perform a drawing frame check on the architectural drawings to be reviewed to obtain the roof drawing frame.

[0038] It should be understood that the specific method for performing drawing frame detection on the architectural drawings to be reviewed can be set according to actual needs, and the embodiments of this application are not limited thereto.

[0039] For example, a drawing can be input for frame detection. Frame detection requires text recognition to extract the coordinates of the text in the title bar, and to traverse line segments of the title bar that conform to the drawing standards by referring to the layer names in the drawing software to determine the coordinates of the frame position. Then, the position of the sub-frame is obtained through digital image processing of the contour edges and connected components, and the sub-frame is identified by the drawing name to determine whether it is a roof.

[0040] In other words, the input architectural drawings can be checked for drawing frames. If a drawing frame contains the roof and its floor plan (e.g., standard floor and first floor), it can be split to obtain the actual roof plan drawing.

[0041] Step S130: Input the roof drawing frame into the pre-trained roof component information extraction model to obtain the output results. The output results include the roof building topology (e.g., whether the door is on the elevator or in the room; how many steps are on the stairs, etc.), basic component information (e.g., if the basic components include doors, windows and walls, the basic component information may include the position coordinates of the doors, windows and walls, etc.), and a mask image of the roof space (e.g., if the basic components include doors, windows and walls, the space formed by the basic components, such as the space formed by the doors, windows and walls).

[0042] It should be understood that the specific model structure of the drawing house component information extraction model can be set according to actual needs, and the embodiments of this application are not limited thereto.

[0043] Optionally, such as Figure 2 As shown, Figure 2 This illustration shows a schematic diagram of a drawing house component information extraction model provided in an embodiment of this application. For example... Figure 2 As shown, it includes an image processing layer, a text feature parsing layer, and a first convolutional layer.

[0044] The training process for the model for extracting information about building components from blueprints includes:

[0045] Obtain training sample data; wherein, the training sample data includes training sample drawings and training sample text used to describe the roof building topology relationship in the training sample drawings, and the training sample drawings can be drawings of roof frames.

[0046] The training sample drawings are processed by an image processing layer to obtain the information decoding results; the information decoding results may include feature vectors and segmentation mask images.

[0047] The training sample text is processed by a text feature parsing layer to obtain ordered text feature vectors. The ordered text feature vectors are obtained by decoding the text feature vectors and can include category words, attribute words, and noun phrases.

[0048] The first convolutional layer performs convolution processing on the information decoding operation results and the ordered feature vector of the text to obtain the training output results; among which, the training output results include the training results of the mask image of the roof space, the training results of the basic component information, and the training results of the roof building topology.

[0049] It should be understood that the specific layers included in the image processing layer, the specific layers included in the first convolutional layer, and the specific layers included in the text feature parsing layer can all be set according to actual needs, and the embodiments of this application are not limited thereto.

[0050] Optionally, such as Figure 2 As shown, the image processing layer includes multiple convolutional layers, a first feature fusion layer, a second convolutional layer, an abstract feature extraction layer, and a first feature pyramid layer (i.e., the first FPN layer);

[0051] In this multi-convolutional layer, each convolutional layer performs corresponding convolution processing on the training sample drawings; the first feature fusion layer performs feature fusion on the convolution results output by each convolutional layer; the second convolutional layer performs convolution processing on the feature fusion results output by the first feature fusion layer; the abstract feature extraction layer removes redundant lines from the roof and extracts abstract roof features (also known as higher-order roof features) based on the convolution results of the second convolutional layer. Redundant lines mainly include parapet wall lines, drainage pipe lines, etc., while abstract roof features can include whether the roof is an upward-facing roof, and abstract roof attribute features such as expansion joints; the first feature pyramid layer performs information decoding operations on the abstract roof features to obtain the information decoding results.

[0052] It should also be understood that the specific layer structures of the multi-convolutional layers, the first feature fusion layer, the second convolutional layer, the abstract feature extraction layer, and the first feature pyramid layer can all be set according to actual needs, and the embodiments of this application are not limited thereto.

[0053] Optionally, such as Figure 2 As shown, the multi-convolutional layer includes a third convolutional layer, a first dilated convolutional layer, a second dilated convolutional layer, and a third dilated convolutional layer arranged in parallel.

[0054] Among them, the void coefficient of the first void convolutional layer is less than that of the second void convolutional layer, and the void coefficient of the second void convolutional layer is less than that of the third void convolutional layer. Furthermore, the void coefficients of the first void convolutional layer, the second void convolutional layer, and the third void convolutional layer are all less than 64.

[0055] Preferably, the void coefficients of the first voided convolutional layer, the second voided convolutional layer, and the third voided convolutional layer are all even numbers.

[0056] For example, the third convolutional layer can be conv1*1, the first dilated convolutional layer can be conv3*3 with a dilation rate of 4, the second dilated convolutional layer can be conv3*3 with a dilation rate of 8, and the third dilated convolutional layer can be conv3*3 with a dilation rate of 16.

[0057] Optionally, the first feature fusion layer can be a concat layer.

[0058] Optionally, the second convolutional layer can be conv1*1.

[0059] Optionally, the abstract feature extraction layer includes a second feature pyramid layer, n parallel feature layers, and a second feature fusion layer for fusing the output of the second feature pyramid layer (i.e., the second FPN layer) with the output of each of the n parallel feature layers. Each of the n feature layers includes a fourth convolutional layer, a batch normalization layer (i.e., BN layer), and an activation function layer (i.e., ReLU layer) connected in sequence.

[0060] It should also be understood that the specific layers and their layer structures of each layer contained in the abstract feature extraction layer can be set according to actual needs, and the embodiments of this application are not limited thereto.

[0061] For example, the fourth convolutional layer can be conv3*3.

[0062] Optionally, the first convolutional layer can be conv3*3.

[0063] Therefore, based on the above, multiple convolutional layers, including the first, second, and third dilated convolutional layers, are merged. Compared to ordinary convolution, this increases the receptive field of the roof drawing's outer contour boundary, preventing the loss of boundary features. Subsequently, the second convolutional layer leads to the abstract feature extraction layer for N layers of operations, and the second convolutional layer performs FPN operations to obtain the roof space and components. Channel concatenation is then performed to obtain the basic features such as the backbone geometry, texture, and color of the roof components and spatial visual extraction.

[0064] Optionally, such as Figure 2 As shown, the text feature parsing layer includes an input layer, a cue learning layer, and a text decoding layer;

[0065] The input layer is used to obtain the input training sample text; the cue learning layer is used to extract the text feature vector from the training sample text; and the text decoding layer is used to decode and compile the text feature vector to obtain the ordered text feature vector.

[0066] The text feature vector can include category words, attribute words, and noun phrases. For example, if the training sample text is a staircase space in a roof with 10 blue straight steps and an M1022 curved swing door, the text feature vector can include blue, straight lines, curves, and door, etc.

[0067] It should also be understood that the specific layers and their layer structures contained in the text feature parsing layer can be set according to actual needs, and the embodiments of this application are not limited thereto.

[0068] For example, the text-based cue learning layer used in this cue learning layer utilizes text to guide or stimulate the image recognition model to complete specific task topological relationships. It is a supervised learning method that continuously optimizes the model's performance through text cues, thereby reducing the purely visual error rate. Furthermore, this cue learning layer can consist of multiple Transform layers.

[0069] In other words, the topological connections of architectural logic require textual feature geometry. Textual feature analysis first requires parsing the text, extracting category words, attributes, and noun phrases from sentences using machine learning (e.g., Hidden Markov Models, HMMs) or deep learning methods (e.g., BERT) for textual cues learning, and then using Transform or self-attention mechanisms for text decoding and compilation to obtain ordered textual feature vectors. This application combines textual analysis of architectural roof logic with topological architectural connections, simultaneously locating targets and inferring their architectural logical semantic relationships and visual attributes to achieve a multi-model recognition model. Furthermore, during training, the forward propagation of the network uses continuous functions in an integral neural network to represent weights, replacing the traditional discrete weight tensors used in neural network models with continuous integral operations. This significantly improves the real-time performance of roof drawing parsing and increases training and testing speed. Furthermore, combining training prompts with learning to understand the attributes of components and spaces can greatly promote the learning of roof space topology relationships. The abstract logic of roof building business greatly improves the identification of small target components and spaces in the roof drawing frame, realizing the establishment of an end-to-end roof information parsing topology map. This network structure is not only applicable to roofs but also to the parsing of drawing information in computer rooms and attics.

[0070] Step S140: Perform spatial mapping from the roof to the top floor plane based on the roof building topology to obtain the spatial mapping result.

[0071] It should be understood that the specific results of spatial mapping can be set according to actual needs, and the embodiments of this application are not limited thereto.

[0072] For example, if there is a water tank on the roof, and there are relevant regulations for the water tank (e.g., the water tank cannot be built above the bedroom), it is possible to determine whether the current room is a bedroom, and to perform spatial mapping based on the roof topology and the roof to the top floor plane.

[0073] Step S150: Based on the output results and spatial mapping results, the roof drawing frame is reviewed according to roof specifications to obtain the roof specification review results.

[0074] Specifically, the output results and spatial mapping results can be stored in a relational database, and the roof drawing frame can be reviewed for roof specifications based on the output results and spatial mapping results stored in the relational database to obtain the roof specification review results.

[0075] Therefore, by means of the above technical solution, this application embodiment obtains the architectural drawings to be reviewed, performs frame detection on the architectural drawings to be reviewed to obtain the roof frame drawings, and inputs the roof frame drawings into a pre-trained drawing roof component information extraction model to obtain output results. The output results include the roof building topology relationship, and a spatial mapping from the roof to the top floor plane based on the roof building topology relationship to obtain spatial mapping results. Based on the output results and spatial mapping results, the roof frame drawings are reviewed for roof specifications to obtain roof specification review results. Thus, not only can roof drawing review be realized, but the accuracy of roof identification and positioning in drawings can also be improved.

[0076] It should be understood that the above-described method for reviewing architectural drawings is merely exemplary, and those skilled in the art can make various modifications based on the above method, and the modified solutions also fall within the protection scope of this application.

[0077] This application provides a storage medium storing a computer program, which is executed by a processor to perform the methods described in the embodiments.

[0078] This application also provides a computer program product that, when run on a computer, causes the computer to perform the method described in the method embodiment.

[0079] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the system described above can be referred to the corresponding process in the aforementioned method, and will not be elaborated further here.

[0080] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For apparatus embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0081] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0082] It should be understood, in the several embodiments provided in this application, that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0083] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0084] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. It should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0085] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application. It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0086] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for reviewing architectural drawings, characterized in that, include: Obtain the architectural drawings to be reviewed; The architectural drawings to be reviewed are subjected to a drawing frame detection to obtain the roof drawing frame. The roof drawing is input into a pre-trained roof component information extraction model to obtain an output result. The output result includes the roof's architectural topology. The model architecture of the roof component information extraction model includes an image processing layer, a text feature parsing layer, and a first convolutional layer. The first convolutional layer performs convolution processing on the image feature decoding result from the image processing layer and the ordered text feature vector from the text feature parsing layer to generate an output result containing the roof's architectural topology. The image processing layer includes an abstract feature extraction layer, which removes redundant lines from the roof and extracts abstract roof features based on the convolution result. The roof component information extraction model is used to achieve end-to-end roof information parsing and topology map establishment. Based on the roof building topology, a spatial mapping is performed from the roof to the top floor plane to obtain the spatial mapping result; Based on the output results and the spatial mapping results, the roof drawing frame is reviewed according to roof specifications to obtain the roof specification review results.

2. The method according to claim 1, characterized in that, The training process of the drawing house component information extraction model includes: Acquire training sample data; wherein, the training sample data includes training sample drawings and training sample text used to describe the roof building topology of the roof in the training sample drawings; The training sample drawings are processed by the image processing layer to obtain the information decoding operation results. The text feature parsing layer is used to perform text feature parsing on the training sample text to obtain an ordered text feature vector. The information decoding operation result and the ordered feature vector of the text are convolved by the first convolutional layer to obtain the training output result; wherein, the training output result includes the training result of the mask image of the roof space, the training result of the basic component information and the training result of the roof building topology relationship.

3. The method according to claim 2, characterized in that, The image processing layer includes multiple convolutional layers, a first feature fusion layer, a second convolutional layer, and a first feature pyramid layer; In this configuration, each convolutional layer is used to perform corresponding convolution processing on the training sample drawings; the first feature fusion layer is used to perform feature fusion on the convolution results output by each convolutional layer; the second convolutional layer is used to perform convolution processing on the feature fusion results output by the first feature fusion layer; and the first feature pyramid layer is used to perform information decoding operations on the roof abstract features to obtain the information decoding operation results.

4. The method according to claim 3, characterized in that, The multiple convolutional layers include a third convolutional layer, a first dilated convolutional layer, a second dilated convolutional layer, and a third dilated convolutional layer arranged in parallel.

5. The method according to claim 4, characterized in that, The hole coefficient of the first dilated convolutional layer is less than that of the second dilated convolutional layer, and the hole coefficient of the second dilated convolutional layer is less than that of the third dilated convolutional layer. Furthermore, the hole coefficients of the first dilated convolutional layer, the second dilated convolutional layer, and the third dilated convolutional layer are all less than 64.

6. The method according to claim 5, characterized in that, The hole coefficients of the first, second, and third dilated convolutional layers are all even numbers.

7. The method according to claim 3, characterized in that, The abstract feature extraction layer includes a second feature pyramid layer, n parallel feature layers, and a second feature fusion layer for fusing the output of the second feature pyramid layer and the output of each of the n parallel feature layers. Each of the n feature layers includes a fourth convolutional layer, a batch normalization layer, and an activation function layer.

8. The method according to claim 2, characterized in that, The text feature parsing layer includes an input layer, a prompting learning layer, and a text decoding layer; The input layer is used to acquire the input training sample text; the prompting learning layer is used to extract text feature vectors from the training sample text; and the text decoding layer is used to perform text decoding and compilation on the text feature vectors to obtain the ordered text feature vectors.

9. A storage medium having a computer program stored thereon, characterized in that, The computer program is executed by the processor to perform the method for reviewing architectural drawings as described in any one of claims 1-8.

10. An electronic device comprising a processor, a memory, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the method for reviewing architectural drawings as described in any one of claims 1-8.

Images