A highway engineering-oriented engineering drawing cost information extraction and analysis method

Abstract

This invention relates to the field of engineering drawing cost information extraction technology, specifically to a method for extracting and analyzing engineering drawing cost information for highway engineering. The invention first processes the engineering drawing image to extract a comprehensive annotation image of each drawing annotation. The center point of each comprehensive annotation image is extracted as a cost information node. Feature vectors are constructed based on the coordinate positions of the cost information nodes and engineering semantic information. The feature vectors of all cost information nodes are arranged in clockwise order to form an input cost feature sequence. This input cost feature sequence is then input into a pre-trained cost analysis model to obtain the cost assessment result of the current highway engineering project. This method can effectively analyze and obtain cost assessment results that meet actual needs.

Description

Technical Field

[0001] This invention relates to the field of engineering drawing cost information extraction technology, specifically to a method for extracting and analyzing engineering drawing cost information for highway engineering. Background Technology

[0002] Highway construction projects typically involve large investments and long construction periods. Costs need to be controlled in advance during the engineering drawing design stage. Adjusting the construction plan due to cost control during the actual construction phase will not only delay the construction period but also cause unnecessary property losses.

[0003] Traditional techniques largely rely on human experience, and due to the numerous factors involved in cost assessment, manual operation is prone to misjudgment. Existing technologies include automated solutions to assist cost assessment, such as exporting all annotation text from engineering drawings and presenting it in tabular form to aid manual evaluation. However, this approach overlooks the fact that the importance of the same annotation text varies across different engineering drawings, and its impact on cost assessment also differs. Therefore, such methods struggle to obtain accurate cost assessment results that reflect reality. Summary of the Invention

[0004] The purpose of this invention is to provide a method for extracting and analyzing cost information from engineering drawings for highway engineering projects. The feature vector of each cost information node in this invention includes coordinate position and engineering semantic information of the corresponding comprehensive labeled image. The coordinate position represents the relative position information of the cost information node in the engineering drawing image, and the engineering semantic information represents the potential cost information contained in the cost information node. The cost analysis model can capture the correlation between various cost information nodes, thereby determining the cost evaluation result of the current engineering drawing image through comprehensive evaluation.

[0005] This invention provides a method for extracting and analyzing cost information from engineering drawings for highway engineering projects, comprising: Obtain engineering drawings and images of the highway project that needs to be evaluated; Based on a pre-built library of drawing annotation leader formats, the engineering drawing image is automatically matched, and the position of each successfully matched drawing annotation is marked and the leader type is recorded. For each drawing annotation that is successfully matched, the selection rules are determined according to the leader line type. Based on the selection rules, a comprehensive annotation image is selected around the position marked by the drawing annotation. The comprehensive annotation image is subjected to text recognition to record the annotation text. The engineering semantic information is determined based on the annotation text. Extract the center point of each comprehensive annotation image as a cost information node, and record the coordinate position of each cost information node in the engineering drawing image; For each cost information node, a feature vector is constructed based on the coordinate position and the engineering semantic information of the corresponding comprehensive annotation image; the feature vectors of all cost information nodes are arranged in clockwise order to form an input cost feature sequence, and the input cost feature sequence is input into a pre-trained cost analysis model to obtain the cost evaluation result of the current highway project.

[0006] As a preferred embodiment, before automatically matching the engineering drawing image, it is necessary to perform a preprocessing operation on the engineering drawing image. The preprocessing operation includes scaling the engineering drawing image to a preset standard resolution and performing image enhancement.

[0007] As a preferred embodiment, the process of constructing the drawing annotation leader format library includes: Acquire multiple historical engineering drawings of highway projects, manually annotate the drawing annotations in the historical engineering drawings, and record the leader line type; Image segmentation models are used to segment each manually annotated drawing annotation in historical engineering drawings to obtain line structure templates for various leader types; A library of drawing annotation leader formats is built based on the line structure templates of various leader types.

[0008] As a preferred embodiment, automatic matching of the engineering drawing image includes: A sliding window is used to extract multiple local image blocks on the engineering drawing image at preset step sizes; For each local image block, a matching standard is constructed based on the line structure template. If there is a line structure in the local image block that meets the matching standard, it is considered a successful match. After a successful match, the position is marked and the lead type is determined according to the matching line structure template.

[0009] As a preferred embodiment, determining the selection rules based on the leader type, and selecting the composite annotation image around the marked position on the drawing according to the selection rules includes: If the drawing annotation is drawn from a single-ended leader, the main line is obtained by extending it from the beginning of the leader along the direction of the leader to the end; the two ends of the main line are extended outward by a first preset length along the direction of the leader to obtain the center line; the center line is extended to both sides by a first preset width to obtain the selected area; the comprehensive annotation image is selected around the position marked by the drawing annotation according to the selected area. If the drawing annotation is led out by two lead lines, extend both ends of the two lead lines outward by a second preset length to obtain two parallel baselines; after moving the two baselines away from each other by a second preset width, use the area between the two baselines as the selection area, and select the comprehensive annotation image around the position marked by the drawing annotation according to the selection area.

[0010] In a preferred embodiment, performing text recognition on the aggregated labeled image to record the labeled text includes: Each character in the composite labeled image is selected using a dynamic anchor box; The anchor frame images of each character are arranged in the order of reading to obtain the text recognition vector; The text recognition vector is input into a pre-trained text recognition model to output labeled text.

[0011] As a preferred embodiment, determining the engineering semantic information based on the labeled text includes: For the labeled text, consult the pre-constructed engineering labeling knowledge graph to determine the labeling type that the labeled text belongs to; If the annotation text contains numerical values ​​and unit information, the annotation object of the annotation text is determined to be a non-standardized object, and engineering semantic information is constructed based on the annotation type and the numerical values ​​and unit information extracted from the annotation text. If the annotation text does not contain numerical or unit information, the annotation object of the annotation text is determined to be a standardized object, and the corresponding standard parameter information is queried in the pre-built standard library. Based on the annotation type and the queried standard parameter information, engineering semantic information is constructed.

[0012] As a preferred embodiment, the annotation types include dimension annotation, material component annotation, pavement structure layer annotation, and bridge component annotation.

[0013] As a preferred embodiment, when recording the coordinate position of each cost information node in the engineering drawing image, the horizontal and vertical coordinates of the coordinate positions are normalized according to the length and width of the engineering drawing image, respectively.

[0014] As a preferred embodiment, the cost analysis model is built based on the Transformer architecture.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This invention first acquires engineering drawing images of the highway project to be evaluated; based on a pre-built library of drawing annotation leader formats, the engineering drawing images are automatically matched, and the location of each successfully matched drawing annotation is marked and the leader type is recorded; for each successfully matched drawing annotation, a selection rule is determined according to the leader type, and a comprehensive annotation image is selected around the marked location of the drawing annotation according to the selection rule; text recognition is performed on the comprehensive annotation image to record the annotation text, and engineering semantic information is determined based on the annotation text; the center point of each comprehensive annotation image is extracted as a cost information node, and the coordinate position of each cost information node in the engineering drawing image is recorded; for each cost information node, a feature vector is constructed based on the coordinate position and the engineering semantic information of the corresponding comprehensive annotation image; the feature vectors of all cost information nodes are arranged in clockwise order to form an input cost feature sequence, and the input cost feature sequence is input into a pre-trained cost analysis model to obtain the cost evaluation result of the current highway project. In this invention, the feature vector of each cost information node includes coordinate position and engineering semantic information of the corresponding comprehensive annotation image; the coordinate position represents the relative position information of the cost information node in the engineering drawing image, and the engineering semantic information represents the potential cost information contained in the cost information node; the cost analysis model can capture the correlation between each cost information node, thereby determining, through comprehensive evaluation, that the current engineering drawing image fits the actual cost evaluation result. Attached Figure Description

[0016] Figure 1 The flowchart illustrates a method for extracting and analyzing cost information from engineering drawings for highway engineering, as provided in this embodiment of the invention. Detailed Implementation

[0017] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0018] The term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0019] Combination Figure 1 This embodiment provides a method for extracting and analyzing cost information from engineering drawings for highway engineering projects, which includes: Step S1: Obtain the engineering drawings of the highway project to be evaluated; Step S2: Based on a pre-built library of drawing annotation leader formats, automatically match the engineering drawing images, mark the location of each successfully matched drawing annotation, and record the leader type. Before automatically matching the engineering drawing images, preprocessing is required. This preprocessing includes scaling the engineering drawing images to a preset standard resolution and enhancing the image using adaptive histogram equalization. Since various engineering drawing images for highway engineering come from different sources, this embodiment eliminates image differences caused by different scanning resolutions by standardizing and enhancing the drawing images, effectively improving the accuracy of subsequent drawing annotation matching.

[0020] The construction process of the drawing annotation leader format library includes: acquiring multiple historical engineering drawing images of highway engineering; manually annotating the drawing annotations in the historical engineering drawing images and recording the leader type; using an image segmentation model to segment each manually annotated drawing annotation in the historical engineering drawing images to obtain the line structure template of each leader type; and constructing the drawing annotation leader format library based on the line structure template of each leader type. In this embodiment, the image segmentation model adopts the U-Net image segmentation network.

[0021] The automatic matching of the engineering drawing image includes: using a sliding window to extract multiple local image blocks on the engineering drawing image with a preset step size; for each local image block, constructing a matching standard based on a line structure template; if there is a line structure in the local image block that meets the matching standard, it is considered a successful match; after a successful match, the position is marked and the leader type is determined according to the matching line structure template.

[0022] The matching criteria in this embodiment include: if the similarity between the line structure and the line structure template within a local image patch at the same scale is greater than a preset similarity threshold, the match is considered successful. This embodiment uses the line structure template as the matching criterion, which can effectively identify drawing annotations at different scales and under rotational changes, resulting in a high accuracy rate for drawing annotation detection.

[0023] Step S3: For each drawing annotation that is successfully matched, determine the selection rules according to the leader line type, select the comprehensive annotation image around the marked position of the drawing annotation according to the selection rules, perform text recognition on the comprehensive annotation image to record the annotation text, and determine the engineering semantic information based on the annotation text.

[0024] Specifically, determining the selection rules based on the type of leader line, and selecting the comprehensive annotation image around the marked position on the drawing according to the selection rules includes: if the drawing annotation is led out by a single-ended leader line, extending from the beginning of the leader line along the direction of the leader line to the end to obtain the main line; extending both ends of the main line outward along the direction of the leader line by a first preset length to obtain the center line; expanding the center line to both sides by a first preset width to obtain the selection area; selecting the comprehensive annotation image around the marked position on the drawing according to the selection area; if the drawing annotation is led out by two-ended leaders line, extending both ends of both leaders line outward by a second preset length to obtain two parallel baselines; moving the two baselines away from each other by a second preset width, using the area between the two baselines as the selection area, and selecting the comprehensive annotation image around the marked position on the drawing according to the selection area.

[0025] This embodiment determines the selection rules based on the type of the leader, which can adapt to different types of drawing annotations and select as much effective information as possible.

[0026] The process of performing text recognition on the comprehensive labeled image to record the labeled text includes: using dynamic anchor boxes to select each character (including numbers, letters, Chinese characters, and special symbols) in the comprehensive labeled image; in a specific embodiment, after selection, the characters within the anchor boxes need to be tilted and corrected; arranging the anchor box images of each character in sequence to obtain a text recognition vector; and inputting the text recognition vector into a pre-trained text recognition model to output the labeled text. The text recognition model here is built based on the Transformer architecture, which can effectively capture the contextual dependencies between characters to output semantically accurate labeled text.

[0027] Determining the semantic information of the project based on the labeled text includes: for the labeled text, consulting a pre-constructed engineering labeling knowledge graph to determine the labeling type that the labeled text conforms to; as an example, the labeling type includes dimension labeling (e.g., length 10m±0.2m, diameter φ2.5m, inclination angle 5°), material component labeling (e.g., HRB400 steel bars, asphalt concrete AC-20, M7.5 mortar-grouted rubble masonry), pavement structure layer labeling (e.g., 4cm asphalt mastic crushed stone SMA-13 ​​surface layer, AC-20C type medium-grained asphalt concrete 6cm, base course 20cm thick cement-stabilized crushed stone), and bridge component labeling (e.g., 20-meter prestressed concrete box girder, pier diameter 1.0m, rubber bearing GJZ30040047). If the annotation text contains numerical values ​​and unit information, the annotation object of the annotation text is determined to be a non-standardized object, and engineering semantic information is constructed based on the annotation type and the numerical values ​​and unit information extracted from the annotation text; if the annotation text does not contain numerical values ​​and unit information, the annotation object of the annotation text is determined to be a standardized object, and the corresponding standard parameter information is queried in the pre-built standard library, and engineering semantic information is constructed based on the annotation type and the queried standard parameter information.

[0028] Step S4: Extract the center point of each comprehensive annotation image as a cost information node, and record the coordinate position of each cost information node in the engineering drawing image. When recording the coordinate position of each cost information node in the engineering drawing image, the horizontal and vertical coordinates of the coordinate positions are normalized according to the length and width of the engineering drawing image, respectively. Since different engineering drawing images have different sizes, the coordinate positions of the cost information nodes in this embodiment need to be normalized so that engineering drawing images of various sizes can be applied to the cost analysis model.

[0029] Step S5: For each cost information node, construct a feature vector based on its coordinate position and the engineering semantic information of the corresponding comprehensive annotation image; arrange the feature vectors of all cost information nodes in clockwise order to form an input cost feature sequence, and input the input cost feature sequence into a pre-trained cost analysis model to obtain the cost assessment result of the current highway project. In a specific embodiment, the feature vectors of all cost information nodes are arranged in clockwise or counterclockwise order to form the input cost feature sequence.

[0030] The cost analysis model is built based on the Transformer architecture; the training method of the cost analysis model includes: acquiring historical engineering drawings of highway projects and corresponding actual cost information; constructing training samples with the input cost feature sequence constructed from the historical engineering drawings as input and the actual cost information as output; constructing a dataset based on the training samples of each historical engineering drawing; and training the cost analysis model using the dataset.

[0031] The cost analysis model in this embodiment is built on the Transformer architecture. The feature vector of each cost information node includes coordinate position and engineering semantic information of the corresponding comprehensive annotation image. The coordinate position represents the relative position information of the cost information node in the engineering drawing image, and the engineering semantic information represents the potential cost information contained in the cost information node. The cost analysis model can capture the long-distance dependency relationship between each cost information node through pre-training, and thus determine the cost evaluation result of the current engineering drawing image through comprehensive evaluation.

[0032] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

[0033] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0034] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0035] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0036] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0037] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A highway engineering-oriented engineering drawing cost information extraction and analysis method, characterized by, include: Obtain engineering drawings and images of the highway project that needs to be evaluated; Based on a pre-built library of drawing annotation leader formats, the engineering drawing image is automatically matched, and the position of each successfully matched drawing annotation is marked and the leader type is recorded. For each drawing annotation that is successfully matched, the selection rules are determined according to the leader line type. Based on the selection rules, a comprehensive annotation image is selected around the position marked by the drawing annotation. The comprehensive annotation image is subjected to text recognition to record the annotation text. The engineering semantic information is determined based on the annotation text. Extract the center point of each comprehensive annotation image as a cost information node, and record the coordinate position of each cost information node in the engineering drawing image; For each cost information node, a feature vector is constructed based on its coordinate position and the engineering semantic information of the corresponding comprehensive annotation image; The feature vectors of all cost information nodes are arranged in clockwise order to form an input cost feature sequence. The input cost feature sequence is then input into a pre-trained cost analysis model to obtain the cost assessment result of the current highway project.

2. The highway engineering-oriented engineering drawing cost information extraction and analysis method according to claim 1, characterized by, Before automatically matching the engineering drawing image, it is necessary to perform preprocessing operations on the engineering drawing image, including scaling the engineering drawing image to a preset standard resolution and performing image enhancement.

3. The highway engineering-oriented engineering drawing cost information extraction and analysis method according to claim 1, characterized by, The process of constructing the drawing annotation leader format library includes: Acquire multiple historical engineering drawings of highway projects, manually annotate the drawing annotations in the historical engineering drawings, and record the leader line type; Image segmentation models are used to segment each manually annotated drawing annotation in historical engineering drawings to obtain line structure templates for various leader types; A library of drawing annotation leader formats is built based on the line structure templates of various leader types.

4. The highway engineering-oriented engineering drawing cost information extraction and analysis method according to claim 3, characterized by, Automatic matching of the engineering drawing images includes: A sliding window is used to extract multiple local image blocks on the engineering drawing image at preset step sizes; For each local image block, a matching standard is constructed based on the line structure template. If there is a line structure in the local image block that meets the matching standard, it is considered a successful match. After a successful match, the position is marked and the lead type is determined according to the matching line structure template.

5. The highway engineering oriented engineering drawing cost information extraction and analysis method according to claim 1, characterized by, The selection rules are determined based on the leader type, and the composite annotation image is selected around the marked position on the drawing according to the selection rules, including: If the drawing annotation is drawn from a single-ended leader, the main line is obtained by extending it from the beginning of the leader along the direction of the leader to the end; the two ends of the main line are extended outward by a first preset length along the direction of the leader to obtain the center line; the center line is extended to both sides by a first preset width to obtain the selected area; the comprehensive annotation image is selected around the position marked by the drawing annotation according to the selected area. If the drawing annotation is led out by two lead lines, extend both ends of the two lead lines outward by a second preset length to obtain two parallel baselines; after moving the two baselines away from each other by a second preset width, use the area between the two baselines as the selection area, and select the comprehensive annotation image around the position marked by the drawing annotation according to the selection area.

6. The highway engineering oriented engineering drawing cost information extraction and analysis method according to claim 1, characterized by, Performing text recognition on the comprehensive annotated image to record the annotated text includes: Each character in the composite labeled image is selected using a dynamic anchor box; The anchor frame images of each character are arranged in the order of reading to obtain the text recognition vector; The text recognition vector is input into a pre-trained text recognition model to output labeled text.

7. The method for extracting and analyzing engineering drawing cost information for highway engineering according to claim 1, characterized in that, The semantic information of the project determined from the labeled text includes: For the labeled text, consult the pre-constructed engineering labeling knowledge graph to determine the labeling type that the labeled text belongs to; If the annotation text contains numerical values ​​and unit information, the annotation object of the annotation text is determined to be a non-standardized object, and engineering semantic information is constructed based on the annotation type and the numerical values ​​and unit information extracted from the annotation text. If the annotation text does not contain numerical or unit information, the annotation object of the annotation text is determined to be a standardized object, and the corresponding standard parameter information is queried in the pre-built standard library. Based on the annotation type and the queried standard parameter information, engineering semantic information is constructed.

8. The highway engineering-oriented engineering drawing cost information extraction and analysis method according to claim 7, characterized by, The annotation types include dimension annotation, material component annotation, pavement structure layer annotation, and bridge component annotation.

9. The highway engineering oriented engineering drawing cost information extraction and analysis method according to claim 1, characterized by, When recording the coordinate position of each cost information node in the engineering drawing image, the horizontal and vertical coordinates of the coordinate positions are normalized according to the length and width of the engineering drawing image, respectively.

10. The highway engineering oriented engineering drawing cost information extraction and analysis method according to claim 1, characterized by, The cost analysis model is built on the Transformer architecture.

Images