Method, device and electronic device for reconstructing a three-dimensional model from three views of a cabinet

Abstract

The present invention provides a method, device and electronic device for reconstructing a three-dimensional model from a two-dimensional design drawing of a cabinet. The method comprises: obtaining three views of the cabinet; analyzing the three views using a line detection deep neural network to obtain the position of the centerline of the side of the panel; correcting the position of the centerline of the side of the panel using primitive vector data, and obtaining the thickness information of the panel at the same time; and reconstructing the three-dimensional model of the cabinet through geometric matching based on the centerline of the panel. The present invention implements a method that supports fully automated import, and can automatically identify the corresponding three-view area based on the drawing file uploaded by the user, and automatically generate the corresponding three-dimensional model with accurate dimensions; the present invention can also read the spatial dimensions of the cabinet and the occlusion relationship between the panels from the three views, and infer the dimensions of unknown panels, so as to reconstruct a three-dimensional model of a cabinet with a more complex structure.

Description

Technical Field

[0001] The present invention belongs to the field of three-dimensional reconstruction in computer vision, and in particular relates to a method, a device and an electronic device for reconstructing a three-dimensional model from three views of a cabinet. Background Art

[0002] In the design process of custom cabinets, designers are accustomed to using AutoCAD software to draw three views to design the structure and size of the cabinet. In the actual production and sales process, it is often necessary to generate a corresponding three-dimensional cabinet model. The three-dimensional cabinet model is very suitable for displaying the actual style of the cabinet, and has great advantages in factory processing, production, sales and promotion of products. However, converting the three views of the two-dimensional design into the corresponding three-dimensional model is a tedious task. The usual practice is to use three-dimensional modeling software to manually model by comparing the three views. There are also some automatic import tools. The disadvantage of these tools is that they can often only import some cabinets with relatively simple structures by using the front view plus manually setting the depth. Complex cabinets that require multiple views to determine the three-dimensional structure cannot be imported. Summary of the Invention

[0003] In view of this, in order to overcome the defects of the current method of generating a three-dimensional model based on the AutoCAD drawings of a custom cabinet, which is too dependent on manual labor and cannot import complex cabinets, the present invention provides a method, device and electronic equipment for reconstructing a three-dimensional cabinet model from three views.

[0004] In a first aspect, the present application provides a method for reconstructing a three-dimensional model from three views of a cabinet, comprising:

[0005] Get the three views of the cabinet;

[0006] The three-view image is processed using a line detection deep neural network to obtain the position of the panel side.

[0007] Use the primitive vector data to correct the position of the side centerline of the plate and obtain the thickness information of the plate;

[0008] According to the center line of the panel, the 3D model of the panel is reconstructed through geometric matching.

[0009] After completing the reconstruction of the 3D models of all panels, the 3D model of the cabinet is obtained.

[0010] In an optional embodiment, obtaining the three views of the cabinet includes:

[0011] Parse and preprocess the cabinet design drawing file to extract the three views of the cabinet.

[0012] In an optional embodiment, parsing and preprocessing include:

[0013] Parse the cabinet design file. If the cabinet design file contains multiple cabinets, split the current cabinet design file into multiple separate cabinet design files.

[0014] Extract three views of each cabinet, including front view, side view and top view;

[0015] For each view, primitive correction and / or enhancement is performed.

[0016] In an optional embodiment, splitting a cabinet design drawing file containing multiple cabinets into multiple separate cabinet design drawing files includes:

[0017] Clustering is performed based on the distance between cabinet elements. All elements of the same cabinet are clustered together, thereby separating the design drawings of different cabinets.

[0018] In an optional embodiment, extracting the three views of each cabinet includes:

[0019] Cluster the graphics elements in a single cabinet design drawing according to the distance between them and extract the three-view drawings;

[0020] Determine the view category based on the up, down, left, and right relationships and text annotations.

[0021] In an optional embodiment, the processing of the three views using a line detection deep neural network to obtain the position of the side surface of the panel includes:

[0022] Convert the three views into bitmap images;

[0023] Normalize the pixel values ​​of the converted bitmap image and scale the image to a fixed size;

[0024] The processed image is fed into the line detection neural network;

[0025] Get the position of the side center lines of all panels in the image coordinates.

[0026] In an optional embodiment, the method of correcting the position of the side centerline of the plate using the primitive vector data and obtaining the thickness information of the plate includes:

[0027] Each of the three views is processed as follows:

[0028] Convert the side center lines of all panels from the image coordinate system to the design drawing coordinate system;

[0029] Add all geometric elements in the design drawing to the R-tree;

[0030] For each panel side centerline:

[0031] Search the R-tree for primitives whose distance is within a certain threshold;

[0032] Find all closed and legal loops in the primitive;

[0033] The rings were converted to form of major axis and thickness.

[0034] In an optional embodiment, the ring simultaneously satisfies:

[0035] (a) is a rectangle;

[0036] (b) the major axis is parallel to the line segment;

[0037] (c) The ratio of the projection length of the major axis on the midline of the side of the plate to the length of the major axis is greater than a certain threshold.

[0038] In an optional embodiment, the line detection deep neural network includes:

[0039] Image feature extraction backbone ResNet50, two Transformer encoders and two Transformer decoders.

[0040] In a second aspect, the present application provides a device for reconstructing a three-dimensional model from three views of a cabinet, comprising:

[0041] An acquisition unit, configured to acquire three views of the cabinet;

[0042] A side position determination unit, configured to process the three-view image using a line detection deep neural network to obtain the side position of the panel;

[0043] a thickness information determining unit, configured to use the primitive vector data to correct the position of the center line of the side surface of the plate and obtain the thickness information of the plate;

[0044] The panel 3D reconstruction unit is used to reconstruct the 3D model of the panel through geometric matching based on the center line of the panel.

[0045] In an optional embodiment, the acquisition unit is further configured to parse and pre-process the cabinet design drawing file to extract the three views of the cabinet.

[0046] In a third aspect, the present application provides an electronic device, comprising:

[0047] The system comprises a processor and a memory, wherein the memory stores a computer program. When the computer program is executed by the processor, the processor executes any one of the steps of the above-mentioned method for reconstructing a three-dimensional model from three views of a cabinet.

[0048] In a fourth aspect, the present application provides a computer-readable storage medium comprising a computer program. When the computer program is run on an electronic device, the computer program is used to enable the electronic device to execute any one of the steps of the above-mentioned method for reconstructing a three-dimensional model from three views of a cabinet.

[0049] In a fifth aspect, the present application also provides a computer program product, which includes a computer program, and the computer program is stored in a computer-readable storage medium; when the processor of an electronic device reads the computer program from the computer-readable storage medium, the processor executes the computer program, so that the electronic device performs any step of the above-mentioned method of reconstructing a three-dimensional model from three views of a cabinet.

[0050] By adopting the above technical solution, the present invention realizes a method that supports fully automatic import, which can automatically identify the corresponding three-view area according to the drawing file uploaded by the user, and automatically generate the corresponding three-dimensional model with accurate size; the present invention can also read the spatial dimensions of the cabinet and the occlusion relationship between the panels from the three views, and infer the dimensions of unknown panels, so as to reconstruct a three-dimensional model of a cabinet with a more complex structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic diagram of a custom cabinet design drawing file provided by an embodiment of the present invention;

[0052] Figure 2 is a schematic diagram of a three-dimensional model of a cabinet involved in an embodiment of the present invention;

[0053] Figure 3 1 is a flow chart of a method for reconstructing a three-dimensional cabinet model provided by an embodiment of the present invention;

[0054] Figure 4 2 is a schematic diagram of panel side centerline detection according to an embodiment of the present invention;

[0055] Figure 5 Schematic diagram of the line detection neural network structure provided by an embodiment of the present invention;

[0056] Figure 6 is a schematic diagram of coordinate system definition provided by an embodiment of the present invention;

[0057] Figure 7 It is a schematic diagram of an application scenario of an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0059] Unless otherwise specified, the terms (including technical terms) used in this embodiment have the meanings commonly understood by those skilled in the art. In addition, it is understood that terms defined in commonly used dictionaries should be understood to have the same meanings as those in the context of the relevant fields, and should not be understood as idealized or overly formal.

[0060] This embodiment provides a method for reconstructing a three-dimensional cabinet model based on a three-view drawing designed by CAD software for a customized cabinet, which is used to realize the Figure 1 The custom cabinet DWG or DXF drawing (also called design drawing in this article) file is generated as shown below Figure 2 The 3D model of the cabinet is shown.

[0061] The overall method flow is as follows Figure 3 Shown, including:

[0062] Step S1: Obtain three views of the cabinet.

[0063] In some embodiments, the cabinet three-view file imported by the user can be directly obtained.

[0064] In other embodiments, a two-dimensional cabinet design drawing file imported by a user is received, such as a DWG or DXF format file designed by CAD software. Drawings drawn by designers manually often have non-standard errors. For example, a custom cabinet drawing file drawn using CAD design software contains multiple cabinets, the viewing angle information of the view is unknown, the rectangles in the drawing are not closed, the line segments are repeated, there are coordinate errors and other problems. In order to be compatible with the drawing habits of different users and increase the robustness against drawing errors, the drawings need to be preprocessed. The specific preprocessing process is as follows:

[0065] Step S1.1: Separate the cabinet drawings, and separate the input drawings of different cabinets, processing each cabinet separately.

[0066] In this embodiment, clustering is performed based on the distance between cabinet graphics elements, and all graphics elements of the same cabinet are clustered together, thereby separating drawings of different cabinets.

[0067] Step S1.2: For a single cabinet drawing, extract drawings from different perspectives (front view, side view, top view).

[0068] In this embodiment, the graphics elements in a single cabinet drawing are clustered according to the distance between the graphics elements, and various views are extracted; the category of the view (front view, side view, top view) is determined according to the up-down, left-right relationship and text annotation.

[0069] Step S1.3 performs primitive correction and enhancement on each view.

[0070] In some embodiments, possible errors in the results automatically completed in the pre-processing process steps S1.1 and S1.2 can be corrected through an interactive process with the user.

[0071] The specific process of primitive correction and enhancement may include:

[0072] Step S1.3.1 Coordinate alignment: align all primitive coordinate points to the cluster center to eliminate small gaps in the drawing.

[0073] Step S1.3.2: Deduplication of primitives, removing duplicate primitives in the image.

[0074] Step S1.3.3: Interrupt line segments at their intersections to ensure that all loops in the drawing are closed. This step achieves the purpose of enhancement. Designers often use drawing methods that appear closed to the naked eye but do not conform to the mathematical definition of a closed loop. This method is correct in itself, but cannot be recognized by computer programs. Therefore, line segments need to be interrupted to ensure that they meet the mathematical definition of a closed loop. For example, if two line segments AB and CD intersect, find their intersection point F; and replace the original line segments with AF, FB, CF, and FD.

[0075] Step S1.3.4 removes the annotation primitives.

[0076] Step S2: Analyze the three-view image using a line detection deep neural network to obtain the position of the side surface of the panel in the three-view image.

[0077] This embodiment uses a line detection deep neural network to analyze the drawing to obtain the position of the side surfaces of the component panels in the drawing.

[0078] The specific process includes:

[0079] In step S2.1, in order to utilize the powerful capabilities of convolutional neural networks, the CAD drawings are first converted into bitmap images.

[0080] Step S2.2 normalizes the pixel values ​​of the transformed image and scales the image to a fixed size.

[0081] Step S2.3 inputs the processed image into the line detection neural network.

[0082] Step S2.4 obtains the positions of the side center lines of all panels in the image coordinates.

[0083] Schematic diagram of panel side centerline detection Figure 4 shown.

[0084] Figure 5 Schematic diagram of the line detection neural network structure used in this embodiment.

[0085] In some embodiments, the line detection neural network may employ an image line detection network (LETR).

[0086] Its structure consists of three parts: image feature extraction backbone ResNet50, two Transformer encoders, and two Transformer decoders.

[0087] The workflow of the image line detection network (LETR) is as follows:

[0088] 1. The image is input into the image feature backbone network to obtain the coarse and fine image feature codes f_img_16 and f_img_32, which are matrices of shape [c, h / 16, w / 16] and [c, h / 32, w / 32] respectively.

[0089] 2. The coarse and fine image features are fed into two Transformer encoders. These features, f_coarse and f_fine, contain more global image information. These matrices are shaped like [h / 16*w / 16,c] and [h / 32*w / 32,c], respectively.

[0090] 3. Input K randomly generated line segment codes ([K,5] matrix) together with f_coarse into the CoarseTransformer decoder to generate K line segment prediction coding intermediate results.

[0091] 4. The K line segment prediction coding intermediate results obtained in step 3 are input into the FineTransformer decoder together with f_fine. The intermediate results are optimized using f_fine to obtain the K line segment coding final results. The line segments with a confidence score greater than the threshold are selected as the final line segment detection results.

[0092] Step S3: Use the primitive vector data to correct the position of the side centerline of the plate detected by the deep neural network, and obtain the thickness information of the plate at the same time.

[0093] The panel side centerline obtained in step S2 is not accurate enough and lacks thickness information. This embodiment uses the primitive vector data in the drawing to correct the position of the panel side centerline detected by the deep neural network and obtains the thickness information of the panel.

[0094] The specific processing flow for drawings at each perspective is as follows:

[0095] Step S3.1: Convert the line segments detected in step S2 from the image coordinate system to the drawing coordinate system.

[0096] Step S3.2: Add all geometric elements in the drawing to the R-tree.

[0097] Step S3.3: For each line segment detected in step S2:

[0098] Step S3.3.1: Search the R-tree for primitives whose distance from the primitive is within a certain threshold.

[0099] Step S3.3.2: Find all closed and legal loops in the primitive.

[0100] Step S3.3.3: Select the ring that meets the following conditions from the found rings:

[0101] (1) is a rectangle;

[0102] (2) The major axis is parallel to the line segment;

[0103] (3) The ratio of the projection length of the major axis on the line segment to the length of the major axis is greater than a certain threshold.

[0104] Step S3.3.4: Convert the rectangular ring into the form of major axis and thickness.

[0105] Step S4: Reconstruct the three-dimensional model of the cabinet through geometric matching based on the center line of the panel.

[0106] According to the center line of the panel obtained in step S3, the three-dimensional contour of the panel is obtained through geometric matching, that is, a three-dimensional model of the panel is generated.

[0107] When there is no obstruction on the side of a 3D plate, any plate has two side center lines in two other views that are inconsistent with the stretching direction. For example, the coordinate system is defined as Figure 6 As shown, if the plate's profile is parallel to plane XOY (i.e., stretched along the Z axis), the side of the plate is visible in the front and left views. Based on the projection relationship between the three views, the front and left views share a Z coordinate. By checking whether the center lines have the same Z coordinate value, the plate's 3D profile information can be recovered. The thickness of the plate can then be recovered using the thickness information in step S3.

[0108] The specific process is as follows:

[0109] Step S4.1: Select any two views and match the panel centerline detected in step S3 to restore the three-dimensional panel.

[0110] Step S4.2: For the center lines that are not matched in the above steps, we can assign default dimensions to these center lines to restore the three-dimensional panels. The default depth can locate the size of the panel by finding the border outline of the cabinet in the remaining views.

[0111] Step S4.3: For all reconstructed panels, unreasonable sizes can be corrected or unreasonable panels can be excluded based on the occlusion relationship between panels and the consistency of the views.

[0112] Optionally, it may also include:

[0113] Step S4.4: The user can modify the reconstruction result of the algorithm through interactive means.

[0114] Since the reconstructed panels are all in one coordinate system, the cabinet can be constructed by merging these panels together.

[0115] That is, the cabinet is composed of panels, and reconstructing the cabinet is equivalent to reconstructing all panels. In other words, completing the 3D reconstruction of all panels completes the 3D reconstruction of the entire cabinet. Thus, the technical solution provided in this embodiment implements an end-to-end panel centerline-based cabinet 3D reconstruction process, employing deep learning-based panel centerline detection and a reconstruction method based on three-view centerline matching.

[0116] By adopting the above technical solution, this embodiment can support the fully automated import of cabinet design drawings, can automatically identify the corresponding three-view area based on the drawing file uploaded by the user, and automatically generate the corresponding three-dimensional model with accurate dimensions; the present invention also realizes further reading the spatial dimensions of the cabinet and the occlusion relationship between the panels from the three views, and infers the dimensions of unknown panels, so as to reconstruct a more complex cabinet three-dimensional model.

[0117] This embodiment also provides a device for reconstructing a three-dimensional model from three views of a cabinet, which is used to implement the aforementioned method for reconstructing a three-dimensional model from three views of a cabinet.

[0118] The device comprises:

[0119] An acquisition unit, configured to acquire three views of the cabinet;

[0120] A side position determination unit, configured to process the three-view image using a line detection deep neural network to obtain the side position of the panel;

[0121] a thickness information determining unit, configured to use the primitive vector data to correct the position of the center line of the side surface of the plate and obtain the thickness information of the plate;

[0122] The panel 3D reconstruction unit is used to reconstruct the 3D model of the panel through geometric matching based on the center line of the panel.

[0123] As an optional embodiment, the acquisition unit is further configured to parse and preprocess the cabinet design file to extract the three views of the cabinet. The parsing and preprocessing includes: parsing the cabinet design file; if the cabinet design file contains multiple cabinets, splitting the current cabinet design file into multiple separate cabinet design files; extracting the three views of each cabinet, including a front view, a side view, and a top view; and performing primitive correction and / or enhancement processing on each view.

[0124] like Figure 7 , which shows the cabinet reconstruction application scenario of this embodiment. The application scenario diagram includes three terminal devices 110 and a server 120.

[0125] In one embodiment, the terminal device 110 includes but is not limited to tablet computers, laptop computers, desktop computers, smart phones, intelligent voice interaction devices and other devices; a client of three-dimensional model reconstruction related software can be installed on the terminal device, and the client can be software (such as a browser, modeling software, etc.), or a web page, applet, etc. The server 120 is a background server corresponding to the software or web page, applet, etc., or a server specifically used for three-dimensional modeling, which is not specifically limited in this application. The server 120 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0126] It should be noted that the cabinet three-dimensional reconstruction method in each embodiment of the present application can be executed by an electronic device, which can be a server 120 or a terminal device 110. That is, the method can be executed by the server 120 or the terminal device 110 alone, or can be executed jointly by the server 120 and the terminal device 110. For example, the cabinet CAD design drawing input by the user through downloading, copying, drawing, etc. is obtained by the terminal device 110, and sent to the server 120 via the communication network, and the server 120 completes the construction of the cabinet three-dimensional model.

[0127] In one embodiment, the electronic device may include a memory, a communication module, and one or more processors.

[0128] Memory is used to store computer programs executed by the processor. It primarily consists of a program storage area and a data storage area. The program storage area can store the operating system and programs required for data transmission and communication functions, while the data storage area can store various data transmission and communication information and operating instruction sets.

[0129] The memory may be a volatile memory, such as random-access memory (RAM); a non-volatile memory, such as read-only memory, flash memory, a hard disk drive (HDD), or a solid-state drive (SSD); or any other medium that can be used to carry or store a desired computer program in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory may also be a combination of the above memories.

[0130] The processor may include one or more central processing units (CPUs) or digital processing units, etc. The processor is configured to implement the above-mentioned method for reconstructing the three-dimensional cabinet model when calling a computer program stored in a memory.

[0131] In some implementations, a circuit Wireless Fidelity (WiFi) module may be included. The WiFi module belongs to a short-range wireless transmission technology, and the electronic device can help the user send and receive data or information through the WiFi module.

[0132] The memory can be used to store software programs and data. The processor executes various functions and data processing of the terminal device by running the software programs or data stored in the memory. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage device. The memory stores an operating system that enables the terminal device to run. In the present application, the memory can store an operating system and various application programs, and can also store a computer program that executes the method of reconstructing a three-dimensional model of a cabinet using a CAD design drawing of a custom cabinet according to an embodiment of the present application.

[0133] In some possible implementations, the method for reconstructing a three-dimensional cabinet model provided in an embodiment of the present application can also be implemented in the form of a program product, which includes a computer program. When the program product is run on an electronic device, the computer program is used to enable the electronic device to execute the steps of the various methods for reconstructing a three-dimensional cabinet model according to the present application described above in this specification.

[0134] Although several units or subunits of the device are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, depending on the embodiment of the application, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided and embodied by multiple units.

[0135] It is worth noting that although the execution process of the method of the present application is described in a specific order in the accompanying drawings, this does not require or imply that the method must be executed in this specific order, or that all steps shown must be executed to achieve the intended result. Optionally, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0136] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, devices, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain a computer-usable computer program.

[0137] The present application is described with reference to the flowcharts and / or block diagrams of the methods, apparatuses, electronic devices, and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program commands. These computer program commands can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the commands executed by the processor of the computer or other programmable data processing device generate commands for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0138] It should be understood that references to "one embodiment" throughout this specification mean that a particular feature, structure, or characteristic associated with the embodiment is included in at least one embodiment of the present invention. Therefore, appearances of "in one embodiment" throughout this specification do not necessarily refer to the same embodiment. Furthermore, these particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

Claims

1. A method for reconstructing a three-dimensional model from three views of a cabinet, characterized in that: include: Get the three views of the cabinet; The three-view image is processed using a line detection deep neural network to obtain the position of the side centerline of the panel; Use the primitive vector data to correct the position of the side centerline of the plate and obtain the thickness information of the plate; Reconstruct the 3D model of the panel through geometric matching based on the panel centerline; After the reconstruction of all the panel 3D models is completed, the 3D model of the cabinet is obtained; The method of using a line detection deep neural network to process the three views to obtain the position of the side centerline of the panel includes: Convert the three views into bitmap images; Normalize the pixel values ​​of the converted bitmap image and scale the image to a fixed size; The processed image is fed into the line detection neural network; Get the position of the side center lines of all panels in the image coordinates; The method of correcting the position of the side centerline of the plate by using the primitive vector data and obtaining the thickness information of the plate includes: Each of the three views is processed as follows: Convert the side center lines of all panels from the image coordinate system to the design drawing coordinate system; Add all geometric elements in the design drawing to the R-tree; For each panel side centerline: Search the R-tree for primitives whose distance is within a certain threshold; Find all closed and legal loops in the primitive; The rings were converted to form of major axis and thickness.

2. The method according to claim 1, wherein: The obtaining of the three views of the cabinet includes: Parse and preprocess the cabinet design drawing file to extract the three views of the cabinet.

3. The method according to claim 2, wherein: The parsing and preprocessing include: Parse the cabinet design file. If the cabinet design file contains multiple cabinets, split the current cabinet design file into multiple separate cabinet design files. Extract three views of each cabinet, including front view, side view and top view; For each view, primitive correction and / or enhancement is performed.

4. The method according to claim 3, wherein: Splitting a cabinet design file containing multiple cabinets into multiple separate cabinet design files includes: Clustering is performed based on the distance between cabinet elements. All elements of the same cabinet are clustered together, thereby separating the design drawings of different cabinets.

5. The method according to claim 3, wherein: Extract the three views of each cabinet including: Cluster the graphics elements in a single cabinet design drawing according to the distance between them and extract the three-view drawings; Determine the view category based on the up, down, left, and right relationships and text annotations.

6. The method according to claim 1, wherein: The ring satisfies simultaneously: (a) is a rectangle; (b) the major axis is parallel to the line segment; (c) The ratio of the projection length of the major axis on the midline of the side of the plate to the length of the major axis is greater than a certain threshold.

7. The method according to claim 1, wherein: The line detection deep neural network includes: Image feature extraction backbone ResNet50, two Transformer encoders and two Transformer decoders.

8. A device for reconstructing a three-dimensional model from three views of a cabinet, characterized in that: include: An acquisition unit, configured to acquire three views of the cabinet; A side position determination unit, configured to process the three-view image using a line detection deep neural network to obtain the position of the side centerline of the panel; a thickness information determining unit, configured to use the primitive vector data to correct the position of the center line of the side surface of the plate and obtain the thickness information of the plate; The panel 3D reconstruction unit is used to reconstruct the 3D model of the panel through geometric matching based on the center line of the panel. Once the reconstruction of the 3D model of all panels is completed, the 3D model of the cabinet is obtained; When the side position determination unit processes the three views using the line detection deep neural network to obtain the position of the side center line of the panel, the following operations are performed: Convert the three views into bitmap images; Normalize the pixel values ​​of the converted bitmap image and scale the image to a fixed size; The processed image is fed into the line detection neural network; Get the position of the side center lines of all panels in the image coordinates; When the thickness information determining unit uses the primitive vector data to correct the position of the center line of the side surface of the plate and obtains the thickness information of the plate, the following operations are performed: Each of the three views is processed as follows: Convert the side center lines of all panels from the image coordinate system to the design drawing coordinate system; Add all geometric elements in the design drawing to the R-tree; For each panel side centerline: Search the R-tree for primitives whose distance is within a certain threshold; Find all closed and legal loops in the primitive; The rings were converted to form of major axis and thickness.

9. The device according to claim 8, characterized in that: The acquisition unit is also used to parse and preprocess the cabinet design drawing file to extract the three views of the cabinet.

10. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor is enabled to perform the steps of the method according to any one of claims 1 to 7.

11. A computer-readable storage medium, characterized in that The method comprises a computer program, and when the computer program is run on an electronic device, the computer program is used to enable the electronic device to perform the steps of the method according to any one of claims 1 to 7.

Images