A method for bridge structure identification and parameter extraction from bridge PDF design drawings

Abstract

This invention relates to a method for bridge structure recognition and parameter extraction from bridge PDF design drawings, comprising: acquiring and preprocessing bridge design drawings saved in PDF format; constructing and training a bridge structure detection model; inputting the bridge PDF design drawings to be detected into a PaddleOCR model for parameter recognition; inputting the preprocessed bridge PDF design drawings to be detected into the trained bridge structure detection model to recognize the bridge structure and obtain the bridge structure in different views. This invention achieves bridge structure detection in drawings by combining an improved YOLOv10 model and the K-Net semantic segmentation algorithm, thereby drawing the bridge structure outline; using the PaddleOCR model to perform text detection on the bridge design drawings to extract bridge structure parameters; and matching the structure with parameters through a series of logical constraints, ultimately realizing the structural recognition and parameter extraction of the bridge in the bridge design drawings, effectively detecting different elements in the drawings.

Description

Technical Field

[0001] This invention relates to the field of PDF image recognition technology, and in particular to a method for identifying bridge structures and extracting parameters from PDF design drawings of bridges. Background Technology

[0002] With the rapid development of transportation design and construction, bridge design has become the most technically challenging and safety-critical aspect of the entire transportation engineering design process. During the bridge construction drawing design phase, designers typically refer to existing bridge project drawings, drawing on their parameter information to reduce trial calculations and preserve design experience. Traditional methods often involve manually reviewing drawings and documents to retrieve information, but these drawings and documents are highly complex and prone to errors. In the architectural field, especially in bridge drawing recognition, domestic and international research demonstrates a dual focus on interdisciplinary integration and technological innovation. The rapid development of computer vision and artificial intelligence provides strong technical support for the automated recognition of bridge drawings. The geometric and semantic information of bridge structures is crucial for the reconstruction of Building Information Modeling (BIM). However, many bridge design drawings stored in PDF format lack BIM, and manually reconstructing BIM is extremely tedious and time-consuming. Therefore, to more effectively manage the structured data in architectural drawings, an efficient method is needed to extract the structure and parameters from bridge design drawings.

[0003] In practical applications, bridge design drawings in PDF format are usually in the form of images or vector graphics. However, these drawings often lack crucial semantic information, making it challenging for computers to understand the precise meaning of the content. First, accurately extracting key parameters from bridge drawings and performing effective rule matching is a technical challenge. Although OCR technology can recognize text and location information in drawings, this information often lacks deep semantic meaning. This makes it difficult to directly apply the identified parameters and establish connections between different views. Second, multi-view parameter matching is a crucial step in interpreting bridge architectural drawings. Bridge drawings typically include plan views, elevation views, and sections, each showing the structural details of the bridge from a different angle. These views together constitute a complete image of the bridge; for example, a plan view provides a horizontal view, while elevation and section views provide vertical and sectional views, respectively. Although these views are projected from different angles of the same entity, the transformations and relationships between them are not always intuitive. For a complete bridge structure, its parameters are scattered across multiple views. Parameters obtained from only one view cannot support the reconstruction of the entire bridge structure's information. Therefore, different views must be linked according to the characteristics of the drawings to obtain complete bridge structural parameters. Summary of the Invention

[0004] To address the shortcomings of target detection technologies in capturing the detailed structure of bridge structural drawings and extracting parameters, the present invention aims to provide a method for identifying and extracting bridge structures and parameters from bridge design drawings, effectively detecting different elements in the drawings.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for bridge structure identification and parameter extraction from bridge PDF design drawings, the method comprising the following sequential steps:

[0006] (1) Obtain the bridge design drawings saved in PDF format, preprocess them to obtain the preprocessed bridge design drawings, and form a dataset;

[0007] (2) Combine the improved YOLOv10 model and contour extraction module to construct a bridge structure detection model;

[0008] (3) The bridge structure detection model is trained using the dataset to obtain the trained bridge structure detection model;

[0009] (4) Obtain the PDF design drawings of the bridge to be detected, input the PDF design drawings of the bridge to be detected into the PaddleOCR model for parameter recognition, and obtain the necessary parameters and their coordinate information.

[0010] (5) Preprocess the PDF design drawings of the bridge to be detected, input the preprocessed PDF design drawings of the bridge to be detected into the trained bridge structure detection model, identify the bridge structure, and obtain the bridge structure in different views.

[0011] In step (1), the preprocessing refers to performing grayscale processing to obtain a binary image, and then labeling and data augmenting the binary image. The augmented bridge design drawings form a dataset.

[0012] Step (2) specifically refers to: improving the YOLOv10 model, which includes a backbone network, a neck network and a head network, adding the CGA attention mechanism module to the layer after the C2f module in the backbone network, and adding the CGA attention mechanism module to the layer after the C2f module in the neck network to obtain the improved YOLOv10 model.

[0013] The output of the improved YOLOv10 model is used as the input of the contour extraction module, and the output of the contour extraction module is used as the output of the bridge structure detection model. The improved YOLOv10 model and the contour extraction module together form the bridge structure detection model.

[0014] The contour extraction module uses the K-Net semantic segmentation algorithm to obtain the foreground mask of the structure to be detected, and then uses the edge pixel detection method and polygon fitting in image processing methods to finally obtain the specific contour of the structure to be detected.

[0015] An edge pixel detection method is used to detect the contour of the foreground in the mask. This method employs the Canny detection operator, which includes four steps: noise reduction, gradient magnitude and direction calculation, non-maximum suppression, and hysteresis thresholding. The two-dimensional Gaussian formula for noise reduction is:

[0016]

[0017] In the formula, G(x,y) represents the value of the Gaussian function at the point (x,y), which is the probability density at the point (x,y) in two-dimensional space; x and y are coordinate variables in two-dimensional space; σ is the standard deviation.

[0018] Step (3) specifically includes the following steps in sequence:

[0019] (3a) Set the training parameters, i.e. the training parameters of the improved YOLOv10 model. The training parameters include the number of iterations, the batch size, and the initial learning rate. The number of iterations is 300, the batch size is 8, and the initial learning rate of the model is set to 0.01.

[0020] (3b) Input the dataset into the improved YOLOv10 model;

[0021] The dataset is fed into the backbone network, which extracts and outputs the feature information of the training images.

[0022] The feature information is input into the neck network, multi-scale feature fusion is performed to enhance the feature representation capability, and the fused feature information is output.

[0023] The fused feature information is input into the head network, which outputs key information about the target.

[0024] (3c) The image output by the improved YOLOv10 model is sent to the contour extraction module. The foreground of the structure is identified by the cropped bridge structure image, the mask is extracted and the polygon is fitted. The resulting polygon is the contour map of the bridge structure.

[0025] Step (4) specifically includes the following steps in sequence:

[0026] (4a) Input the PDF design drawings of the bridge to be detected into the PaddleOCR model, set the direction of manual reading of the drawings as the standard direction, and extract the text in the standard direction normally. For the text perpendicular to the standard direction, first rotate the image by 90° and then recognize the text in the vertical direction. Then, through a coordinate transformation, normalize and save the recognized text, and output the detected text information and position information.

[0027] (4b) Set logical constraints: Use the text information detected by the PaddleOCR model, combined with the annotation area identified by the improved YOLOv10 model, and the structure identified by the bridge structure detection model to jointly filter out the text and coordinate information of the necessary parameters.

[0028] At this point, the required parameters are not assigned any attributes. Then, the required parameters are matched with the obtained bridge structure. Specifically, the Euclidean distance between the coordinates of the required parameters and the coordinates of the midpoint of the bridge edge is calculated. When the distance is the smallest, the bridge edge is matched with the required parameters. This completes the parameter and coordinate matching for one view. Then, based on the inherent relationship between different views, the parameters of multiple views are summarized into an Excel spreadsheet to achieve the final parameter extraction.

[0029] As can be seen from the above technical solution, the beneficial effects of the present invention are as follows: First, compared with the prior art, the present invention achieves the detection of bridge structures in drawings by combining the improved YOLOv10 model and the K-Net semantic segmentation algorithm, thereby drawing the outline of the bridge structure; it uses the PaddleOCR model to perform text detection on the bridge design drawings and extracts the parameters of the bridge structure; through a series of logical constraints, the structure and parameters are matched, and finally the structure recognition and parameter extraction of the bridge in the bridge design drawings are realized; Second, the present invention adds the CGA attention mechanism module to the YOLOv10 model and combines it with the K-Net semantic segmentation algorithm, which solves the problem of inaccurate outline detection in bridge design drawings; at the same time, in response to the problem that the text information detected by the PaddleOCR model lacks semantic information, semantic information is given to the text and parameters in the drawings; Third, experimental results show that the bridge structure detection model can achieve a recall rate of 96.1% and an average precision of 96.2%, which are 14.6% and 13.1% higher than the original YOLOv10 model, respectively, and can effectively detect different elements in the drawings. Attached Figure Description

[0030] Figure 1 This is a sequence diagram of the method of the present invention;

[0031] Figure 2 This is a network structure diagram of the improved YOLOv10 model in this invention;

[0032] Figure 3This is a network diagram of the CGA attention mechanism involved in this invention;

[0033] Figure 4 This is a schematic diagram of the cross-section of a cast-in-place box girder;

[0034] Figure 5 This is a schematic diagram of the elevation of a cast-in-place box girder;

[0035] Figure 6 This is a schematic diagram of the top surface of the cast-in-place box girder;

[0036] Figure 7 This is a schematic diagram of the bottom surface of the cast-in-place box girder;

[0037] Figure 8 The PDF design drawings of the bridge to be inspected;

[0038] Figure 9 This is a schematic diagram of the semantic segmentation results;

[0039] Figure 10 The result is the contour fitting.

[0040] Figure 11 This is a diagram showing the parameter extraction results involved in this invention. Detailed Implementation

[0041] like Figure 1 As shown, a method for bridge structure identification and parameter extraction from bridge PDF design drawings is presented. This method includes the following steps in sequence:

[0042] (1) Obtain the bridge design drawings saved in PDF format, preprocess them to obtain the preprocessed bridge design drawings, and form a dataset;

[0043] (2) Combine the improved YOLOv10 model and contour extraction module to construct a bridge structure detection model;

[0044] (3) The bridge structure detection model is trained using the dataset to obtain the trained bridge structure detection model;

[0045] (4) Obtain the PDF design drawings of the bridge to be inspected, such as Figure 8 As shown, the PDF design drawings of the bridge to be detected are input into the PaddleOCR model for parameter recognition to obtain the necessary parameters and their coordinate information;

[0046] (5) Preprocess the PDF design drawings of the bridge to be detected, input the preprocessed PDF design drawings of the bridge to be detected into the trained bridge structure detection model, identify the bridge structure, and obtain the bridge structure in different views.

[0047] In step (1), the preprocessing refers to performing grayscale processing to obtain a binary image, and then labeling and data augmenting the binary image. The augmented bridge design drawings form a dataset.

[0048] Step (2) specifically refers to: such as Figure 2 As shown, the YOLOv10 model is improved, which includes a backbone network, a neck network, and a head network, as follows: Figure 3 As shown, adding the CGA attention mechanism module to the layer after the C2f module in the backbone network and adding the CGA attention mechanism module to the layer after the C2f module in the neck network can improve the model's ability to detect bridge structures; thus, the improved YOLOv10 model is obtained.

[0049] The output of the improved YOLOv10 model is used as the input of the contour extraction module, and the output of the contour extraction module is used as the output of the bridge structure detection model. The improved YOLOv10 model and the contour extraction module together form the bridge structure detection model.

[0050] The contour extraction module uses the K-Net semantic segmentation algorithm to obtain the semantic segmentation result, i.e., the foreground mask of the structure to be detected, such as... Figure 9 As shown, the black area is the background and the red area is the foreground.

[0051] Then, edge pixel detection and polygon fitting methods from image processing are used to finally obtain the specific outline of the structure to be detected.

[0052] An edge pixel detection method is used to detect the contour of the foreground in a mask. This method employs the Canny detection operator, a multi-stage edge detection operator that integrates filtering, enhancement, and detection, aiming to find an optimal edge contour. The detection results are fused using the Canny detection operator to improve accuracy. Canny detection includes four steps: noise reduction, gradient magnitude and direction calculation, non-maximum suppression, and hysteresis thresholding. The two-dimensional Gaussian formula for noise reduction is:

[0053]

[0054] In the formula, G(x,y) represents the value of the Gaussian function at the point (x,y), which is the probability density at the point (x,y) in two-dimensional space; x and y are coordinate variables in two-dimensional space; σ is the standard deviation.

[0055] Step (3) specifically includes the following steps in sequence:

[0056] (3a) Set the training parameters, i.e. the training parameters of the improved YOLOv10 model. The training parameters include the number of iterations, the batch size, and the initial learning rate. The number of iterations is 300, the batch size is 8, and the initial learning rate of the model is set to 0.01.

[0057] (3b) Input the dataset into the improved YOLOv10 model;

[0058] The dataset is fed into the backbone network, which extracts and outputs the feature information of the training images.

[0059] The feature information is input into the neck network, multi-scale feature fusion is performed to enhance the feature representation capability, and the fused feature information is output.

[0060] The fused feature information is input into the head network, which outputs key information about the target.

[0061] (3c) The image output by the improved YOLOv10 model is sent to the contour extraction module. The foreground of the structure is identified by the cropped bridge structure image, the mask is extracted and the polygon is fitted. The resulting polygon is the contour map of the bridge structure.

[0062] Step (4) specifically includes the following steps in sequence:

[0063] (4a) Input the PDF design drawings of the bridge to be detected into the PaddleOCR model, set the direction of manual reading of the drawings as the standard direction, and extract the text in the standard direction normally. For the text perpendicular to the standard direction, first rotate the image by 90° and then recognize the text in the vertical direction. Then, through a coordinate transformation, normalize and save the recognized text, and output the detected text information and position information.

[0064] (4b) Set logical constraints: Use the text information detected by the PaddleOCR model, combined with the annotation area identified by the improved YOLOv10 model, and the structure identified by the bridge structure detection model to jointly filter out the text and coordinate information of the necessary parameters.

[0065] At this point, the required parameters are not assigned any attributes. Then, the required parameters are matched with the obtained bridge structure. Specifically, the Euclidean distance between the coordinates of the required parameters and the coordinates of the midpoint of the bridge edge is calculated. When the distance is the smallest, the bridge edge is matched with the required parameters. This completes the parameter and coordinate matching for one view. Then, based on the inherent relationship between different views, the parameters of multiple views are summarized into an Excel spreadsheet to achieve the final parameter extraction.

[0066] Figure 4The required parameters are marked in the diagram. Among them, "8-8" represents the serial number, "160" represents the beam height, "20" represents the thickness of the cantilever end, "1686.8" represents the width of the bridge top surface, and "186.8" represents the width of the bridge bottom surface.

[0067] Figure 5 Several required parameters are marked in the figure, among which "2495" represents the span length, "|8" and "|11" represent the section number, and "Z12" and "Z13" represent the span positioning number.

[0068] Figure 6 The required parameters are marked in the code, among which "2495" represents the span length, "1675" represents the width of the bridge top surface, and "Z12" and "Z13" represent the span positioning numbers.

[0069] Figure 7 Several required parameters are marked in the code, among which "2495" represents the span length, "1275" represents the width of the bridge bottom surface, and "Z12" and "Z13" represent the span positioning numbers.

[0070] like Figure 10 As shown, the bridge structure detection model can accurately extract the outline structure of the bridge from the bridge design drawings.

[0071] like Figure 11 As shown, this invention can detect the type of bridge structure in bridge design drawings, obtain the necessary parameter information, and simultaneously incorporate all structural dimensions into a library.

[0072] In summary, this invention achieves bridge structure detection in drawings by combining an improved YOLOv10 model and the K-Net semantic segmentation algorithm, thereby drawing the bridge structure outline. It utilizes the PaddleOCR model to perform text detection on the bridge design drawings, extracting the bridge structure parameters. Through a series of logical constraints, the structure and parameters are matched, ultimately achieving bridge structure recognition and parameter extraction in the bridge design drawings. This invention incorporates the CGA attention mechanism module into the YOLOv10 model and combines it with the K-Net semantic segmentation algorithm, solving the problem of inaccurate outline detection in bridge design drawings. Simultaneously, addressing the lack of semantic information in the text detected by the PaddleOCR model, semantic information is assigned to the text and parameters in the drawings. Experimental results show that the bridge structure detection model achieves a recall rate of 96.1% and an average precision of 96.2%, which are improvements of 14.6% and 13.1% respectively compared to the original YOLOv10 model, effectively detecting different elements in the drawings.

Claims

1. A method for bridge structure identification and parameter extraction from bridge PDF design drawings, characterized in that: The method includes the following steps in sequence: (1) Obtain the bridge design drawings saved in PDF format, preprocess them to obtain the preprocessed bridge design drawings, and form a dataset; (2) Combine the improved YOLOv10 model and contour extraction module to construct a bridge structure detection model; (3) The bridge structure detection model is trained using the dataset to obtain the trained bridge structure detection model; (4) Obtain the PDF design drawings of the bridge to be inspected, input the PDF design drawings of the bridge to be inspected into the PaddleOCR model for parameter recognition, and obtain the necessary parameters and their coordinate information; (5) Preprocess the PDF design drawings of the bridge to be detected, input the preprocessed PDF design drawings of the bridge to be detected into the trained bridge structure detection model, identify the bridge structure, and obtain the bridge structure in different views. Step (2) specifically refers to: improving the YOLOv10 model, which includes a backbone network, a neck network and a head network, adding the CGA attention mechanism module to the layer after the C2f module in the backbone network, and adding the CGA attention mechanism module to the layer after the C2f module in the neck network to obtain the improved YOLOv10 model. The output of the improved YOLOv10 model is used as the input of the contour extraction module, and the output of the contour extraction module is used as the output of the bridge structure detection model. The improved YOLOv10 model and the contour extraction module together form the bridge structure detection model. The contour extraction module uses the K-Net semantic segmentation algorithm to obtain the foreground mask of the structure to be detected, and then uses the edge pixel detection method and polygon fitting in image processing methods to finally obtain the specific contour of the structure to be detected. An edge pixel detection method is used to detect the contour of the foreground in the mask. This method employs the Canny detection operator, which includes four steps: noise reduction, gradient magnitude and direction calculation, non-maximum suppression, and hysteresis thresholding. The two-dimensional Gaussian formula for noise reduction is: ； In the formula, Indicates at point The value of the Gaussian function at a point in two-dimensional space, i.e., the value of the Gaussian function at that point. The probability density at that location; and It is a coordinate variable in two-dimensional space; That is the standard deviation.

2. The method for bridge structure identification and parameter extraction from bridge PDF design drawings according to claim 1, characterized in that: In step (1), the preprocessing refers to performing grayscale processing to obtain a binary image, and then labeling and data augmenting the binary image. The augmented bridge design drawings form a dataset.

3. The method for bridge structure identification and parameter extraction from bridge PDF design drawings according to claim 1, characterized in that: Step (3) specifically includes the following steps in sequence: (3a) Set the training parameters, i.e. the training parameters of the improved YOLOv10 model. The training parameters include the number of iterations, the batch size, and the initial learning rate. The number of iterations is 300 epochs, the batch size is 8, and the initial learning rate of the model is set to 0.

01. (3b) Input the dataset into the improved YOLOv10 model; The dataset is fed into the backbone network, which extracts and outputs the feature information of the training images. The feature information is input into the neck network, multi-scale feature fusion is performed to enhance the feature representation capability, and the fused feature information is output. The fused feature information is input into the head network, which outputs key information about the target. (3c) The image output by the improved YOLOv10 model is sent to the contour extraction module. The foreground of the structure is identified by the cropped bridge structure image, the mask is extracted and the polygon is fitted. The resulting polygon is the contour map of the bridge structure.

4. The method for bridge structure identification and parameter extraction from bridge PDF design drawings according to claim 1, characterized in that: Step (4) specifically includes the following steps in sequence: (4a) Input the PDF design drawings of the bridge to be detected into the PaddleOCR model, set the direction of manual reading of the drawings as the standard direction, and extract the text in the standard direction normally. For the text perpendicular to the standard direction, first rotate the image by 90° and then recognize the text in the vertical direction. Then, through a coordinate transformation, normalize and save the recognized text, and output the detected text information and position information. (4b) Set logical constraints: Use the text information detected by the PaddleOCR model, combined with the annotation area identified by the improved YOLOv10 model, and the structure identified by the bridge structure detection model to jointly filter out the text and coordinate information of the necessary parameters; At this point, the required parameters are not assigned any attributes. Then, the required parameters are matched with the obtained bridge structure. Specifically, the Euclidean distance between the coordinates of the required parameters and the coordinates of the midpoint of the bridge edge is calculated. When the distance is the smallest, the bridge edge is matched with the required parameters. This completes the parameter and coordinate matching for one view. Then, based on the inherent relationship between different views, the parameters of multiple views are summarized into an Excel spreadsheet to achieve the final parameter extraction.

Images