Architectural drawing plane wall identification method, system and equipment

Abstract

The invention discloses a plane wall body identification method, system and equipment of an architectural drawing. The plane wall body identification method comprises the following steps: reading primitive data in the given architectural drawing; identifying a filler wall and a non-filler wall based on the primitive data, and acquiring first image data corresponding to the filler wall and second image data corresponding to the non-filler wall; and combining the first image data and the second image data to obtain the plane wall in the architectural drawing, so as to improve the recognition accuracy of the plane wall.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a planar wall identification method, system and device for architectural drawings. BACKGROUND

[0002] In the design process of existing architectural design construction drawings, computer software is mainly used for drawing. The plan view of a residential building is an important content for safety specification review in the field of architecture.

[0003] At present, the identification of the wall of a residential building mainly relies on semantic segmentation technology. The semantic segmentation technology exports all lines and filled elements in a single plan view into an image, and then uses a semantic segmentation network to segment the wall. However, due to the complexity of the wall, such as the wall of a residential building may be composed of multiple colors, and may be composed of filled blocks of different colors and different layers; the plan view of a residential building contains many interference components such as lead lines, spaces and rulers. Therefore, the accuracy of the method using semantic segmentation for identifying the wall of a residential building is very low. SUMMARY

[0004] In order to solve the above technical problems, the present application discloses a planar wall identification method, system and device for architectural drawings.

[0005] In order to achieve the above purpose, in a first aspect, the present application discloses a planar wall identification method for architectural drawings, comprising:

[0006] reading the graphic element data in a given architectural drawing;

[0007] identifying the filled wall and the non-filled wall based on the graphic element data, respectively, to obtain first image data corresponding to the filled wall and second image data corresponding to the non-filled wall;

[0008] merging the first image data and the second image data to obtain the planar wall in the architectural drawing.

[0009] The planar wall identification method for architectural drawings disclosed by the present application first reads the graphic element data in the architectural drawing to accurately identify the wall based on the graphic element data. Then, the wall is divided into filled walls and non-filled walls for identification according to the graphic element data, which simplifies the complexity of the wall and improves the accuracy of wall identification. The first image data corresponding to the filled wall and the second image data corresponding to the non-filled wall are generated by identifying the filled wall and the non-filled wall, respectively. Then, the first image data and the second image data are merged to finally identify the wall in the architectural drawing and ensure the accuracy of the identification.

[0010] As a preferred example, the image data includes a plurality of filled image elements, a plurality of line image elements, first attribute information corresponding to each of the filled image elements, and second attribute information corresponding to each of the line image elements; wherein, the attribute information includes coordinates, color, and layer.

[0011] This invention uses the image data to intuitively express the information of the walls in the architectural drawings. By reading the image data, subsequent wall recognition can be performed, which reduces the difficulty of wall recognition, improves the efficiency of wall recognition, and ensures the accuracy of wall recognition by utilizing the image data.

[0012] As a preferred example, the identification of infilled walls and non-infilled walls based on the map data includes:

[0013] Obtain the set of fill points corresponding to each fill element, and convert the set of fill points into several lines connected one after the other;

[0014] Based on the lines and the first attribute information, several first filling elements for filling wall identification are selected from the several filling elements according to preset line filtering conditions.

[0015] This invention is based on the fact that infilled walls are visually represented as filled areas on drawings. Therefore, the data of the filled graphic elements allows for intuitive identification of infilled walls, improving identification efficiency. Simultaneously, to ensure the accuracy of infilled wall identification, the filled graphic elements are simplified into lines, and then filtered using preset line filtering conditions, guaranteeing the accuracy of infilled wall identification.

[0016] As a preferred example, the step of selecting several first filling elements for infill wall identification from several filling elements based on preset line filtering conditions includes:

[0017] Match each of the remaining second lines in the plurality of lines with a single first line from the plurality of lines;

[0018] Based on the coordinates in the first attribute information, calculate the angle, spacing, and relative direction between the first line and each of the second lines;

[0019] If the angle, spacing, and relative direction all meet the preset filling conditions, then the individual filling elements corresponding to the several lines are determined as the first filling element.

[0020] This invention uses double lines, parallel line spacing, and line direction for preliminary screening of infill walls to remove elements that are not infill walls, ensuring the accuracy of recognition. At the same time, it uses the lines of the elements to reduce the difficulty of recognition and improve the efficiency of recognition.

[0021] As a preferred example, the identification of infilled walls and non-infilled walls based on the map data includes:

[0022] Based on the set of fill points corresponding to each of the first fill elements, obtain the size of the first fill element and the number of fill boundary points;

[0023] Based on the size and the number of fill boundary points, the first fill element identified as an interfering fill element is deleted;

[0024] A de-dangling operation is performed on each of the first filling primitives after the interference filling primitives have been deleted.

[0025] In the process of identifying infilled walls, this invention eliminates interference from small infilled elements such as pipes. It uses the size of the infill point set and the number of boundary points of the elements to delete small infilled elements, and performs a de-dangling operation on the remaining infilled elements after deletion to eliminate the construction in the element data and improve the accuracy of infilled wall identification.

[0026] As a preferred example, the identification of infilled walls and non-infilled walls based on the map data includes:

[0027] Each line element is classified according to its color, and the line elements are converted according to their classification to obtain a first image corresponding to each color category.

[0028] Obtain the connected components of the first image, and crop the first image according to the connected components to obtain the second image corresponding to each first image;

[0029] The second image is classified according to a pre-trained lightweight classification network to obtain a non-filled classification result for each second image; wherein, the non-filled classification result includes a first classification result when the classification is successful and a second classification result when the classification fails.

[0030] This invention is based on the use of line primitives of various colors corresponding to non-filled walls in residential buildings. By constructing corresponding image data from line primitives of the same color, and classifying the image data according to a pre-built lightweight classification network, the accuracy of non-filled wall recognition is improved.

[0031] As a preferred example, the step of identifying infilled walls and non-infilled walls based on the map data further includes:

[0032] The non-filled classification result is determined to be the second classification result. The first image is divided according to the layer to obtain the single layer corresponding to the first image.

[0033] The lightweight classification network is used to classify a single layer to obtain the non-filled classification result for each of the second images.

[0034] When the non-filled classification result of the first image is a classification failure, the present invention takes into account the possible interference from other graphic elements during the recognition process, and therefore performs a second recognition on the first image with the classification failure result, thereby improving the accuracy of non-filled image recognition.

[0035] As a preferred example, obtaining the first image data corresponding to the infilled wall and the second image data corresponding to the non-infilled wall includes:

[0036] The images of several first filling primitives are transformed to obtain the first mask image corresponding to the filling wall;

[0037] Based on the first classification result and the third classification result, several first images are merged to obtain the second mask image corresponding to the non-filled wall.

[0038] This invention merges the recognition results of infilled walls and non-infilled walls to finally generate a mask image corresponding to the planar walls in the building image, thus ensuring the accuracy of the planar wall recognition.

[0039] Secondly, the present invention discloses a planar wall recognition system for architectural drawings, including a graphic element reading module, a fill recognition module, and a wall recognition module;

[0040] The graphic element reading module is used to read graphic element data from a given architectural drawing;

[0041] The filling recognition module is used to identify filled walls and non-filled walls based on the image data, and to obtain the first image data corresponding to the filled wall and the second image data corresponding to the non-filled wall.

[0042] The wall recognition module is used to merge the first image data and the second image data to obtain the planar wall in the architectural drawings.

[0043] This invention discloses a planar wall recognition system for architectural drawings. First, it reads the graphic data from the architectural drawings to accurately identify the walls based on this data. Next, it categorizes the walls into infilled walls and non-infilled walls based on the graphic data, simplifying the wall complexity and improving recognition accuracy. By separately identifying infilled and non-infilled walls, it generates first image data corresponding to the infilled walls and second image data corresponding to the non-infilled walls. Finally, it merges the first and second image data to achieve wall recognition in the architectural drawings, ensuring high accuracy.

[0044] Thirdly, the present invention discloses an electronic device comprising: at least one processor, and a memory communicatively connected to at least one of the processors; wherein the memory stores instructions which are executed by at least one of the processors to enable the at least one processor to implement a planar wall recognition method for architectural drawings as described in the first aspect when executing the instructions. Attached Figure Description

[0045] Figure 1 : A schematic flowchart of a method for identifying planar walls in architectural drawings provided by an embodiment of the present invention;

[0046] Figure 2 : A schematic diagram of the structure of a planar wall recognition system for architectural drawings provided in an embodiment of the present invention;

[0047] Figure 3 : A schematic flowchart of a method for identifying planar walls in architectural drawings, provided in another embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of a filling detection process provided in another embodiment of the present invention;

[0049] Figure 5 This is a schematic diagram of a non-filling detection process provided in another embodiment of the present invention;

[0050] Figure 6 : A schematic diagram of the structure of a lightweight classification network provided in another embodiment of the present invention;

[0051] Figure 7 : A schematic diagram illustrating how a layer interferes with non-fill detection, according to another embodiment of the present invention;

[0052] Figure 8 This is a schematic diagram of a filling detection result provided in another embodiment of the present invention;

[0053] Figure 9 This is a schematic diagram of a non-filling detection result provided in another embodiment of the present invention;

[0054] Figure 10 This is a schematic diagram of a planar wall recognition result provided in another embodiment of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0056] Example 1

[0057] This embodiment discloses a method for identifying planar walls in architectural drawings. For the specific implementation process of the identification method, please refer to [reference needed]. Figure 1 It mainly includes steps 101 to 103, and the steps are mainly as follows:

[0058] Step 101: Read the drawing data from the given architectural drawings.

[0059] In this embodiment, the main steps are as follows: the image data includes several filled image elements, several line image elements, first attribute information corresponding to each filled image element, and second attribute information corresponding to each line image element; wherein, the attribute information includes coordinates, color, and layer.

[0060] In this embodiment, this step is based on the fact that the infilled wall is a visually represented infill on the drawing. Therefore, the data of the infilled elements can be used to intuitively identify the infilled wall, improving the efficiency of identification. At the same time, in order to ensure the accuracy of infilled wall identification, the infilled elements are simplified into lines, and then the infilled elements are filtered using preset line filtering conditions to ensure the accuracy of infilled wall identification.

[0061] Step 102: Based on the image data, identify the filled walls and non-filled walls respectively, and obtain the first image data corresponding to the filled walls and the second image data corresponding to the non-filled walls.

[0062] In this embodiment, the main steps are as follows: First, the process of identifying the infill wall is as follows: obtaining the set of filling points corresponding to each filling element and converting the set of filling points into several lines connected one after the other; according to the several lines and the first attribute information, selecting several first filling elements for infill wall identification from the several filling elements based on preset line filtering conditions; wherein, each of the remaining second lines in the several lines is matched with a single first line in the several lines; calculating the angle, spacing and relative direction between the first line and each second line according to the coordinates in the first attribute information; if the angle, spacing and relative direction all meet the preset filling conditions, then the single filling element corresponding to the several lines is determined as the first filling element; further, obtaining the size and number of filling boundary points of the first filling element according to the set of filling points corresponding to each first filling element; deleting the first filling elements determined to be interfering filling elements according to the size and the number of filling boundary points; and performing a de-hanging operation on each first filling element after deleting the interfering filling elements.

[0063] Secondly, the process of identifying the non-filling is as follows: each line primitive is classified according to its color, and the line primitive is transformed according to the classification to obtain a first image corresponding to each color class; connected components of the first image are obtained, and the first image is cropped according to the connected components to obtain a second image corresponding to each first image; the second image is classified according to a pre-trained lightweight classification network to obtain a non-filling classification result corresponding to each second image; wherein, the non-filling classification result includes a first classification result corresponding to successful classification and a second classification result corresponding to unsuccessful classification; further, the non-filling classification result is determined to be the second classification result, the first image is divided according to layers to obtain a single layer corresponding to the first image; the single layer is classified according to the lightweight classification network to obtain the non-filling classification result corresponding to each second image.

[0064] Finally, the process of constructing the first image data and the second image data is as follows: converting several first filling primitives into images to obtain the first mask image corresponding to the filling wall; merging several first images according to the first classification result to obtain the second mask image corresponding to the non-filling wall.

[0065] In this embodiment, for infill walls, this step uses double lines, parallel line spacing, and line direction for initial screening to remove non-infill wall elements, ensuring recognition accuracy. Simultaneously, the lines of the elements reduce recognition difficulty and improve efficiency. Further, based on the various colors of line elements corresponding to non-infill walls in residential buildings, image data corresponding to line elements of the same color is constructed. This image data is then classified according to a pre-built lightweight classification network, improving the accuracy of non-infill wall recognition. Finally, the recognition results of infill and non-infill walls are merged to generate a mask image corresponding to the planar walls in the building image, ensuring the accuracy of planar wall recognition.

[0066] Step 103: Merge the first image data and the second image data to obtain the planar wall in the architectural drawings.

[0067] In this embodiment, the main step is to merge the first mask image and the second mask image to construct the mask image corresponding to the planar wall.

[0068] On the other hand, this embodiment also discloses a planar wall recognition system for architectural drawings. For the specific structural composition of the recognition system, please refer to... Figure 2 It includes a graphic element reading module 201, a fill recognition module 202, and a wall recognition module 203.

[0069] The graphic element reading module 201 is used to read graphic element data in a given architectural drawing;

[0070] The filling recognition module 202 is used to identify filled walls and non-filled walls based on the image data, and to obtain the first image data corresponding to the filled wall and the second image data corresponding to the non-filled wall.

[0071] The wall recognition module 203 is used to merge the first image data and the second image data to obtain the planar wall in the architectural drawings.

[0072] In this embodiment, the graphic element reading module 201 reads graphic element data including a plurality of filled graphic elements, a plurality of line graphic elements, first attribute information corresponding to each filled graphic element, and second attribute information corresponding to each line graphic element; wherein, the attribute information includes coordinates, color, and layer.

[0073] In this embodiment, the filling recognition module 202 includes a filling wall recognition unit, a non-filling wall recognition unit, and an image construction unit;

[0074] The infill wall recognition unit is used to acquire the set of filling points corresponding to each of the filling elements, and convert the set of filling points into several lines connected one after the other; based on the several lines and the first attribute information, it selects several first filling elements for infill wall recognition from the several filling elements based on preset line filtering conditions; wherein, each of the remaining second lines in the several lines is matched with a single first line in the several lines; based on the coordinates in the first attribute information, it calculates the angle, spacing and relative direction between the first line and each second line; if the angle, spacing and relative direction all meet the preset filling conditions, then the single filling element corresponding to the several lines is determined as the first filling element; further, based on the set of filling points corresponding to each first filling element, it acquires the size of the first filling element and the number of filling boundary points; based on the size and the number of filling boundary points, the first filling elements determined to be interfering filling elements are deleted; and a de-hanging operation is performed on each first filling element after the interfering filling elements are deleted.

[0075] The non-filling recognition unit is used to classify each line primitive according to its color, and to perform image transformation on the line primitive according to the classification to obtain a first image corresponding to each color class; to obtain connected components of the first image, and to crop the first image according to the connected components to obtain a second image corresponding to each first image; to classify the second image according to a pre-trained lightweight classification network to obtain a non-filling classification result corresponding to each second image; wherein, the non-filling classification result includes a first classification result corresponding to successful classification and a second classification result corresponding to unsuccessful classification; further, the non-filling classification result is determined to be the second classification result, the first image is divided according to layers to obtain a single layer corresponding to the first image; the single layer is classified according to the lightweight classification network to obtain the non-filling classification result corresponding to each second image.

[0076] The image construction unit is used to convert several first filling primitives into images to obtain a first mask image corresponding to the filling wall; according to the first classification result, several first images are merged to obtain a second mask image corresponding to the non-filling wall.

[0077] In this embodiment, the wall recognition module 203 constructs a mask image corresponding to the planar wall by merging the first mask image and the second mask image.

[0078] In some embodiments of this example, an electronic device and a computer-readable storage medium are also disclosed. The electronic device includes at least one processor and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that are executed by the at least one processor to cause the at least one processor to implement a planar wall recognition method for architectural drawings as described in this embodiment; the computer-readable storage medium stores computer-executable instructions that cause a computer to perform the planar wall recognition method for architectural drawings as described in this embodiment.

[0079] This embodiment discloses a method, system, device, and medium for identifying planar walls in architectural drawings. First, it reads the graphic data from the architectural drawings to accurately identify the walls based on this data. Next, it categorizes the walls into infilled walls and non-infilled walls based on the graphic data, simplifying the wall complexity and improving the accuracy of wall identification. By separately identifying the infilled and non-infilled walls, it generates first image data corresponding to the infilled walls and second image data corresponding to the non-infilled walls. Finally, it merges the first and second image data to achieve wall identification in the architectural drawings, ensuring accurate identification.

[0080] Example 2

[0081] This embodiment discloses a method for identifying planar walls in architectural drawings, used to extract infill and non-infill walls within a planar wall structure. Specifically, refer to... Figure 3 The diagram below shows the implementation details of the planar wall recognition method described in this embodiment, including steps 301 to 310. The main steps are as follows:

[0082] Step 301: Read the drawing data from the input architectural drawings.

[0083] In this embodiment, the main steps are: acquiring the architectural drawings to be identified, exporting the drawing data of the architectural drawings using computer-aided design software, such as CAD software, and then acquiring the graphic data contained in the architectural drawings. The graphic data includes graphic elements such as lines and fills, as well as attribute information for each graphic element such as coordinates, color, and layer.

[0084] Step 302: Obtain the data of the filling element from the data of the map data, and obtain the first filling element for filling wall identification from the data of the filling element by means of line filtering.

[0085] In this embodiment, the main steps are as follows: obtaining the set of fill points corresponding to each fill element, converting the set of fill points into several lines connected one after the other, matching each of the remaining second lines with a single first line among the several lines, calculating the angle, spacing and relative direction between the first line and each second line, and determining the single fill element corresponding to the several lines as the first fill element when the angle, spacing and relative direction all meet the preset fill conditions.

[0086] Specifically, in this embodiment, since infill walls are visually represented as filled areas on architectural drawings, and infill walls are divided into two types: solid infill and pattern infill, infill elements and their attribute information are filtered from the data. Further, the data for infill elements is represented as a point set pts[[pt1,pt2,...],[],...]. Each point in the point set represents a boundary point of the infill.

[0087] Optionally, to identify infill walls, preliminary screening can be performed using double lines, parallel line spacing, and line direction. Specifically, refer to... Figure 4 In this context, the set of points corresponding to a single filled primitive is... Figure 4 The Infill ElementData input is in Figure 4 The line filtering module, or Line Filter Module, converts the graphic data into consecutive lines and matches the remaining lines based on the single line contained in a single filled graphic element, thereby calculating the angle of the lines. Figure 4 The angle and the spacing of the lines are as follows Figure 4 The Di stance and the relative direction of the lines are as follows Figure 4 The angle of the lines can be used to determine whether two lines are parallel; the spacing between the lines represents the distance from a point to a line; and the relative direction of the lines can determine whether two lines are in the same direction.

[0088] If multiple lines after a filled element is converted are parallel, aligned in direction, and have a width within a preset range, such as 350, then the current filled element is determined to be a filled wall.

[0089] Step 303: Filter out several interfering primitives present in the first filling primitives, and perform a de-dangling operation on each of the first filling primitives.

[0090] In this embodiment, the main steps are as follows: based on the set of filling points corresponding to each first filling element, obtain the size of the first filling element and the number of filling boundary points; based on the size and the number of filling boundary points, delete the first filling elements that are determined to be interfering filling elements; and perform a de-dangling operation on each first filling element after deleting the interfering filling elements.

[0091] Specifically, in the process of identifying infill walls, to eliminate interference from graphic elements such as small pipe infills, one can, for example... Figure 4 As shown, the first filled primitives selected by the line filtering module are input into the data post-processing module, i.e., the PostProcessing Module. Interfering primitives are filtered out by the component filter unit within the data post-processing module. Specifically, the filtering unit can filter out interfering primitives based on the number and size of the filled boundary points.

[0092] Furthermore, for suspended elements such as doorways, a mid-air removal operation can be performed to improve the accuracy of wall recognition. The mid-air removal operation involves: solving for connected components in the original exported image to obtain components with dimensions within a certain range, such as 100; then, by taking the intersection with the wall mask, if a component falls within the wall's area, it is removed.

[0093] Step 304: Convert the first filled primitive that has undergone the above-mentioned de-suspended operation into an image to obtain the first image data corresponding to the filled wall.

[0094] Specifically, in this embodiment, the processed image data is converted into an image to obtain an Infilled Mask that fills the wall.

[0095] Optionally, refer to Figure 4 The first filling image data, processed by the data post-processing module, is converted into an image to obtain an Infilled Mask of the filling wall. The Infilled Mask of the filling wall can be referenced... Figure 8 .

[0096] Step 305: Classify the line primitives in the image data according to color, so as to generate first image data corresponding to a single color according to the classification.

[0097] In this embodiment, the main steps are: classifying each line graphic element according to the color, and performing image conversion on the line graphic elements according to the classification to obtain a first image corresponding to each color; obtaining the connected components of the first image, and cropping the first image according to the connected components to obtain a second image corresponding to each first image;

[0098] Specifically, in this embodiment, referring to Figure 5 The line primitives are categorized by color to generate a set of line primitives corresponding to each color category. Each set of line primitives is then converted into an image to obtain an image corresponding to a single color.

[0099] Optionally, refer to Figure 5 The image of a single color is fed into the Crop Module for cropping. The Crop Module solves the connectivity problem for the input image (mask) to obtain all connected components. Based on the left-right coordinates of these connected components, it identifies the regions in the image containing only wall distributions, enlarges them by 20 pixels, and then crops them to obtain the first image data to be predicted.

[0100] Step 306: Classify each of the first image data according to the pre-trained lightweight classification network to generate a first image dataset and a second image dataset; wherein, the first image dataset includes several first images corresponding to successful classification and several first images corresponding to unsuccessful classification.

[0101] In this embodiment, the main steps are: classifying the first image according to the pre-trained lightweight classification network, and determining whether the first image belongs to the image corresponding to the non-filled wall.

[0102] Specifically, refer to Figure 5 The image data corresponding to each single color obtained above is fed into a pre-trained lightweight classification network, Tiny_classifynet, for prediction to obtain the classification result of the image for that color. Specifically, when the Tiny_classifynet network scores the image data beyond a threshold, i.e., exceeds the threshold... Figure 5 When the value is 0.9, it can be determined that the image data is the image data corresponding to the non-filled wall.

[0103] In some implementations of this embodiment, pre-training is required to ensure the accuracy of the Tiny_classify.net network. Optionally, the structure of the Tiny_classify.net network can be referred to... Figure 6 ,like Figure 6As shown, the backbone of the Tiny_classifyynet network uses a convolutional + c3 module to extract features from image features. Upsampling is used in the middle to improve the representation of features at different scales. A c2f feature fusion module is used to concatenate channels, enhancing the information transmission of gradient flow data. This network has fast training and inference speeds, can acquire high-resolution and rich semantic features, and exhibits good accuracy for small datasets such as wall classification.

[0104] Furthermore, referring to Figure 6 The training process of the Tiny_classifyynet network uses walls under a single frame as training data, and divides the training and test samples at a ratio of 9:1 to obtain the classification structure of the data: in the category names={0:'door_window',1:'edge',2:'furnace',3:'other',4:'wall'}, walls are the category with index 4.

[0105] Furthermore, in such Figure 5 As shown, after classifying images based on color, some first images may still fail to be classified or be determined to contain other components. For these unclassified first images, further image segmentation using layers can be used to avoid issues such as... Figure 7 The different layers of the shown color interfere with non-fill detection. A single layer of that color is fed into the network for prediction as a single data point. A score exceeding 0.8 is considered accurate classification. Other unclassified values ​​are placed under the "other" category.

[0106] Step 307: Obtain the second image data corresponding to the non-filled wall according to the first image dataset, and merge the first image data and the second image data to obtain the image data corresponding to the planar wall in the architectural drawing.

[0107] Specifically, in this embodiment, the first images in the first image dataset are merged to obtain the second mask corresponding to the non-filled wall. The second mask can be referenced from... Figure 9 .

[0108] The first mask corresponding to the infill wall and the second mask corresponding to the non-infill wall are merged to synthesize the mask corresponding to the planar wall in the architectural drawings. The mask corresponding to the planar wall can then be referenced... Figure 10 .

[0109] This embodiment discloses a method for identifying planar walls in architectural drawings, which divides and conquers infilled and non-filled walls within the planar wall structure. For infilled walls, the location and mask are determined through parallel line filtering, size filtering, and graphic shape filtering. For non-filled walls, a trained lightweight wall classification network is used to sequentially filter by color and layer to obtain a non-filled wall mask that meets the requirements. This identification method is adaptable to wall component identification in residential buildings and to wall identification in complex scenarios. The wall segmentation model in this invention can be applied to larger drawings, such as underground parking garages, for wall positioning. For precisely segmented walls and components, a 3D model can be accurately reconstructed for residential building model design.

[0110] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for identifying planar walls in architectural drawings, characterized in that, The method comprises the steps of: reading the graphic data in a given architectural drawing; based on the graphic data, identifying the filled wall and the non-filled wall respectively, obtaining the first image data corresponding to the filled wall and the second image data corresponding to the non-filled wall; merging the first image data and the second image data to obtain the planar wall in the architectural drawing.

2. The method of claim 1, wherein the planar wall is identified by a line segment in the architectural drawing. The graphic data comprises a plurality of filled graphic elements, a plurality of line graphic elements, first attribute information corresponding to each filled graphic element, and second attribute information corresponding to each line graphic element; wherein the attribute information comprises coordinates, colors, and layers.

3. The method of claim 2, wherein the planar wall is identified by: The identification of the filled wall and the non-filled wall based on the graphic data comprises: obtaining a set of filling points corresponding to each filled graphic element, and converting the set of filling points into a plurality of lines connected in front and back; based on a plurality of lines and the first attribute information, a plurality of first filled graphic elements used for filled wall identification are screened from a plurality of filled graphic elements based on a preset line screening condition.

4. The method of claim 3, wherein the planar wall is identified by: The screening of a plurality of first filled graphic elements used for filled wall identification from a plurality of filled graphic elements based on a preset line screening condition comprises: matching each second line in the remaining plurality of lines with a single first line in the plurality of lines; based on the coordinates in the first attribute information, calculating the angle, distance, and relative direction of the first line and each second line; determining that the angle, distance, and relative direction all satisfy a preset filling condition, and then determining the single filled graphic element corresponding to the plurality of lines as the first filled graphic element.

5. The method of claim 3, wherein the planar wall is identified by a line segment in the architectural drawing. The identification of the filled wall and the non-filled wall based on the graphic data comprises: based on the set of filling points corresponding to each first filled graphic element, obtaining the size and the number of filling boundary points of the first filled graphic element; based on the size and the number of filling boundary points, deleting the first filled graphic element determined as an interference filled graphic element; performing a hanging removal operation on each first filled graphic element after the interference filled graphic element is deleted.

6. The method of claim 2, wherein the planar wall identification method of an architectural drawing is characterized by, The identification of the filled wall and the non-filled wall based on the graphic data comprises: classifying each line graphic element according to the color, and performing image conversion on the line graphic element according to the classification to obtain a first image corresponding to each color category; obtaining a connected block of the first image, and performing cropping on the first image according to the connected block to obtain a second image corresponding to each first image; classifying the second image according to a pre-trained lightweight classification network to obtain a non-filled classification result corresponding to each second image; wherein the non-filled classification result comprises a first classification result corresponding to a successful classification and a second classification result corresponding to a failed classification.

7. The method of claim 6, wherein the planar wall is identified by a line segment in the architectural drawing. The identification of the filled wall and the non-filled wall based on the graphic data further comprises: determining that the non-filled classification result is the second classification result, dividing the first image according to the layer to obtain a single layer corresponding to the first image; According to the lightweight classification network, the single layer is classified to obtain the non-filled classification result corresponding to each second image.

8. A method of identifying planar walls in an architectural drawing according to any one of claims 1 to 7, wherein, The first image data corresponding to the filled wall and the second image data corresponding to the non-filled wall are obtained, including: The first mask image corresponding to the filled wall is obtained by converting the images of the first filled image elements. According to the first classification result, the second mask image corresponding to the non-filled wall is obtained by merging the first images.

9. A planar wall identification system for architectural drawings, characterized by It comprises a figure reading module, a filling identification module and a wall identification module. The figure reading module is used to read the figure data in the given architectural drawing. The filling identification module is used to identify the filled wall and the non-filled wall based on the figure data, to obtain the first image data corresponding to the filled wall and the second image data corresponding to the non-filled wall. The wall identification module is used to merge the first image data and the second image data to obtain the plane wall in the architectural drawing.

10. An electronic device, comprising: It comprises: At least one processor and a memory connected in communication with the at least one processor; wherein the memory stores instructions, and the instructions are executed by the at least one processor to make the at least one processor execute the instructions to realize the method for identifying the plane wall of the architectural drawing according to any one of claims 1 to 8.

Images