A method for identifying staircase components in architectural plan staircase detail drawings

Abstract

The present invention discloses a method for identifying staircase components in a planar staircase drawing of a building professional, and the method comprises the following steps: S1, correctly recommending the corresponding layers of the basic graphics in the drawing, and printing them on a png picture, and finding the closed space composed of the graphics on the picture by the contour search operation in the traditional image algorithm; S2, cutting out all the graphics after the staircase components are merged from the picture, performing category prediction, obtaining the confidence of the stairs, and screening out whether the predicted component is a staircase and the position of the component by the confidence threshold and the size of the component; S3, finally judging the text information next to the staircase component to classify the stairs. The present invention can effectively solve the problem of inaccurate staircase recognition and position determination by using an object classification method based on deep learning to identify stairs in a building professional residential plan and a staircase drawing; and can effectively improve the recognition accuracy of staircase components.

Description

Technical Field

[0001] The present invention relates to the field of intelligent recognition of stair components in residential floor plans and stair detail drawings of architectural design professionals, and in particular to a method for recognizing stair components based on deep learning object classification technology. Background Art

[0002] CAD construction drawings are drawings that use AutoCAD software to show the overall layout of the project, the building's exterior shape, interior layout, structural structure, interior and exterior decoration, material methods, equipment, and construction. CAD construction drawings are complete, accurate, and specific. They are the basis for engineering construction, construction drawing budgets, and construction organization design. They are also important technical documents for technical management. Before construction, the construction drawings need to be carefully reviewed before entering the construction phase. The purpose is to ensure the smooth progress of the construction and avoid the impact of errors in the drawings on the use phase after the completion of the construction.

[0003] Stairs often appear in CAD construction drawings. Stairs are important components in floor plans or large-scale drawings. It is necessary to judge whether the space where the stair is located is a stairwell and whether the design of the stairs is reasonable based on the stair components. So far, there is no technology to solve the problem of accurately identifying stair components after converting CAD drawings into image formats. Manual review is extremely inconvenient and will cause waste of manpower and material resources. In addition, architectural drawings are often large in size. The use of general computer recognition methods will not only consume a lot of time and have poor results, but also lead to deviations between the identification of stair components and the actual situation, and the position of the frame is also inaccurate, affecting the subsequent compliance review of the residential plan. Summary of the invention

[0004] In view of the above problems, the present invention provides a method for identifying stair components in a plan staircase drawing of a professional building.

[0005] To solve the above technical problems, the present invention is implemented by the following technical solutions: The present invention provides a method for identifying stair components in a plan staircase drawing of a building professional, and the method comprises the following steps:

[0006] S1. First, you need to recommend the layer where the basic elements in the architectural drawing are located to the correct component layer where the stairs may appear. Then, use the CAD drawing printing service to print all the constructed elements onto a png image. Use the contour search operation in the traditional image algorithm to find the closed space composed of the elements on the image.

[0007] S2. All the graphics elements after the staircase components are merged are cut out from the image, and then sent to the deep learning model for object classification to predict the category of the component. The result returns the confidence level of the component being classified as a staircase. The confidence threshold and the component size are used to screen and finally determine whether the predicted component is a staircase and the location of the component.

[0008] S3. Finally, the stairs are subdivided by judging the text information next to the stair components.

[0009] Preferably, the specific method steps include:

[0010] S11, primitive merging;

[0011] S12, cropping thumbnails;

[0012] S13, model training;

[0013] S14, model deployment;

[0014] S15, model calling.

[0015] Preferably, the step S11 specifically includes: correctly recommending the basic graphics elements in the drawing to the corresponding layers, recommending the stair-related graphics elements to the stair layer, and printing them on a png picture, merging the graphics elements that are interconnected, and also using the contour search in the traditional image algorithm to find the closed space composed of the graphics elements.

[0016] Preferably, the step S12 specifically includes: cropping the obtained image composed of the closed space of the staircase to obtain a small image, and changing the image size to a size acceptable to the model through padding and resizing operations.

[0017] Preferably, the step S13 specifically includes the following steps:

[0018] S131. Data preparation: Obtain the required graphic element information on about 30 drawings through CAD parsing, and generate 100 PNG images of residential floor plans through CAD drawing printing service;

[0019] S132, check the 100 residential floor plans in step 1 to ensure that the stair component elements are printed correctly;

[0020] S133, using traditional image processing technology to obtain the closed space outline in the residential plane;

[0021] S134, capture all small images of stair components in the residential floor plan, 200×350, and mark them as stair;

[0022] S135. Send the labeled data set to the classification model for training, and select the parameters with the highest average accuracy of the validation set during the iteration process as the parameters of the subsequent deployment model.

[0023] Preferably, the step S14 specifically includes: deploying the trained model to ai-service for calling by the AI ​​image recognition engine.

[0024] Preferably, the step S15 specifically includes: sending the cropped small picture to the model for prediction to obtain the correct component category, which is stairs in this case.

[0025] It can be seen from the above description of the present invention that, compared with the prior art, the present invention has the following advantages:

[0026] The present invention discloses a method for identifying stair components in a professional architectural plan staircase drawing. By using an object classification method based on deep learning to identify stairs in professional architectural residential plan drawings and staircase drawings, the problem of inaccurate stair identification and position determination can be effectively solved; and the recognition accuracy of stair components can be effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flow chart of the steps of the present invention;

[0028] Figure 2 The specific steps of the embodiment of the present invention are as follows: Figure 1 ;

[0029] Figure 3 The specific steps of the embodiment of the present invention are as follows: Figure 2 ;

[0030] Figure 4 It is a flow chart of specific steps of model training of the present invention;

[0031] Figure 5 It is a CAD plane schematic diagram of the present invention incorporating staircase graphic elements and cutting small graphics;

[0032] Figure 6 It is a CAD plane diagram of the model training of the present invention

[0033] Figure 7 It is a CAD plan schematic diagram of the category prediction of the present invention. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with 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.

[0035] like Figure 1-7As shown, the present invention provides a method for identifying stair components in a plan staircase drawing of a building professional, and the method comprises the following steps:

[0036] S1. First, you need to recommend the layer where the basic primitives (the primitives refer to the visible basic elements that make up the graphics, which correspond to the visible entities on the drawing interface. For example, straight lines, arcs, circles, etc., these basic elements form meaningful components, such as doors and windows) in the architectural drawings to the correct component layer where the stairs may appear, and then use the CAD drawing printing service to print all the constructed primitives to a png image (the width of the image is 11890 and the height is 8410). Through the contour search operation in the traditional image algorithm, find the closed space composed of the primitives on the image;

[0037] S2. All the graphics elements after the staircase components are merged are cut out from the image, and then sent to the deep learning model for object classification to predict the category of the component. The result returns the confidence level of the component being classified as a staircase. The confidence threshold and the component size are used to screen and finally determine whether the predicted component is a staircase and the location of the component.

[0038] S3. Finally, the stairs are subdivided by judging the text information next to the stair components.

[0039] The specific method steps include:

[0040] S11, primitive merging;

[0041] S12, cropping thumbnails;

[0042] S13, model training;

[0043] S14, model deployment;

[0044] S15, model calling.

[0045] Among them, step S11 specifically includes: correctly recommending the corresponding layers of the basic graphics in the drawing, recommending the stair-related graphics to the stair layer, and printing them on the png picture, merging the graphics that are interconnected, and using the contour search in the traditional image algorithm to find the closed space composed of the graphics.

[0046] Among them, step S12 specifically includes: cropping the obtained image composed of the closed space of the staircase to obtain a small image, and changing the image size to a size acceptable to the model through padding and resizing operations.

[0047] Wherein, step S13 specifically includes the following steps:

[0048] S131. Data preparation: Obtain the required graphic element information on about 30 drawings through CAD parsing, and generate 100 PNG images of residential floor plans through CAD drawing printing service;

[0049] S132, check the 100 residential floor plans in step 1 to ensure that the stair component elements are printed correctly;

[0050] S133, using traditional image processing technology to obtain the closed space outline in the residential plane;

[0051] S134, capture all small images of stair components in the residential floor plan, 200×350, and mark them as stair;

[0052] S135. Send the labeled data set to the classification model for training (the MobileNet lightweight model is used here), and select the parameters with the highest average accuracy of the validation set during the iteration process as the parameters of the subsequent deployment model.

[0053] Among them, step S14 specifically includes: deploying the trained model to ai-service for calling by the AI ​​image recognition engine.

[0054] Among them, step S15 specifically includes: sending the cropped small picture to the model for prediction to obtain the correct component category, which is stairs in this case.

[0055] The present invention provides a method for identifying stair components in architectural professional plan staircase drawings. By using an object classification method based on deep learning to identify stairs in architectural professional residential plan drawings and staircase drawings, the problems of inaccurate stair identification and position determination can be effectively solved; and the recognition accuracy of stair components can be effectively improved.

[0056] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying staircase components in a planar staircase drawing of a building, characterized in that: The method comprises the following steps: S1. First, you need to recommend the layer where the basic elements in the architectural drawing are located to the correct component layer where the stairs may appear. Then, use the CAD drawing printing service to print all the constructed elements onto a png image. Use the contour search operation in the traditional image algorithm to find the closed space composed of the elements on the image. S2. All the graphics elements after the staircase components are merged are cut out from the image, and then sent to the deep learning model for object classification to predict the category of the component. The result returns the confidence level of the component being classified as a staircase. The confidence threshold and the component size are used to screen and finally determine whether the predicted component is a staircase and the location of the component. S3. Finally, the stairs are subdivided by judging the text information next to the stair components.

2. According to the method for identifying staircase components in a planar staircase drawing of a building professional according to claim 1, it is characterized by: The specific steps include: S11, primitive merging; S12, cropping thumbnails; S13, model training; S14, model deployment; S15, model calling.

3. According to claim 2, a method for identifying staircase components in a planar staircase drawing of a building, characterized in that: The step S11 specifically includes: correctly recommending the basic graphics elements in the drawing to the corresponding layers, recommending the stair-related graphics elements to the stair layer, and printing them on a png picture, merging the graphics elements that are interconnected, and using the contour search in the traditional image algorithm to find the closed space composed of the graphics elements.

4. According to claim 2, a method for identifying staircase components in a planar staircase drawing of a building, characterized in that: The step S12 specifically includes: cropping the obtained image composed of the closed space of the staircase to obtain a small image, and changing the image size to a size acceptable to the model through padding and resizing operations.

5. According to claim 2, a method for identifying staircase components in a planar staircase drawing of a building, characterized in that: The step S13 specifically includes the following steps: S131, data preparation: 30 drawings are parsed through CAD to obtain the required element information on the drawings, and 100 PNG images of residential floor plans are generated through the CAD drawing printing service; S132, check the 100 residential floor plans in step 1 to ensure that the stair component elements are printed correctly; S133, using traditional image processing technology to obtain the closed space outline in the residential plane; S134, capture all small pictures of stair components in the residential floor plan and mark them as stair; S135. Send the labeled data set to the classification model for training, and select the parameters with the highest average accuracy of the validation set during the iteration process as the parameters of the subsequent deployment model.

6. According to claim 2, a method for identifying staircase components in a planar staircase drawing of a building, characterized in that: The step S14 specifically includes: deploying the trained model to ai-service for the AI ​​image recognition engine to call.

7. According to claim 2, a method for identifying staircase components in a planar staircase drawing of a building, characterized in that: The step S15 specifically includes: sending the cropped small picture to the model for prediction to obtain the correct component category, which is stairs in this case.

Images