Information recognition method and device for structural drawing, storage medium and electronic equipment

Abstract

This invention relates to a method, apparatus, storage medium, and electronic device for information recognition in structural drawings. The method for information recognition in structural drawings, belonging to the field of building manufacturing technology, includes: using a sliding window to cut the target drawing frame into multiple blocks of predetermined sizes; inputting the blocks into a target detection model, locating the components based on their graphic features, and outputting the bounding box positions and category labels of the components; within the detection area where the bounding box positions are located, extracting target line segments that conform to the structural features of the components based on their line segment shapes and construction patterns; and, based on the relative positional relationship between the components and the beam structure, combining the direction of the annotation lines and text information, completing the selection of target line segments and the matching and binding of target line segments with corresponding annotation information, thus improving the efficiency of drawing recognition and review.

Description

Technical Field

[0001] This invention relates to the field of building manufacturing technology, and in particular to a method, device, storage medium, and electronic device for identifying information in structural drawings. Background Technology

[0002] As construction projects become increasingly complex and larger in scale, drawing review, as a crucial part of the design phase, directly impacts the safety and compliance of project implementation. In the architectural design process, structural drawings convey key information such as the building's stress system, structural details, and reinforcement methods, serving as the core basis for ensuring structural safety and construction feasibility.

[0003] In practice, structural drawing review is typically done manually by experienced technical personnel. While this method offers some flexibility, the large number of drawings and numerous details make it prone to omissions and misinterpretations, potentially leading to hidden engineering problems. This is especially true when it comes to the layout and quantity calculation of structural components such as beams, columns, and walls. Accurate identification of component information and correct interpretation of reinforcement details, connection methods, and dimensional annotations are crucial, placing high demands on manpower. Therefore, a better method for structural drawing review is needed. Summary of the Invention

[0004] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a method, device, storage medium and electronic device for identifying information in structural drawings, which solves the technical problem of the difficulty of manually reviewing structural drawings in the prior art.

[0005] To achieve the above objectives, the main technical solutions adopted by the present invention include:

[0006] The first aspect of this invention provides a method for identifying information in structural drawings.

[0007] The method for identifying information in structural drawings proposed in this embodiment of the invention includes:

[0008] A sliding window is used to cut the target frame to obtain multiple blocks of predetermined sizes; wherein, the target frame contains beam structure information;

[0009] The image blocks are input into the target detection model, and the components are located by combining the component graphic features of the beam structure components. The bounding box position and category label of the components are then output.

[0010] Within the detection area where the bounding box is located, target line segments that conform to the structural characteristics of the component are extracted based on the line segment shape and structural rules of the component.

[0011] Based on the relative positional relationship between the components and the beam structure, and combined with the direction of the annotation lines and text information, the target line segments are filtered and matched and bound to the corresponding annotation information.

[0012] In some instances, the category labels include at least stirrups and hangers;

[0013] The process of inputting the image blocks into the target detection model, combining the component graphic features of the beam structure members, locating the components, and outputting the bounding box position and category label of the components includes:

[0014] The image blocks are input into the target detection model. Based on the component graphic features of the beam structure components, the components are located, and the first boundary box position of the stirrups and the second boundary box position of the hanger are output.

[0015] In some instances, the stirrups have a straight line segment shape and a structural pattern characterized by multiple straight line segments arranged in a parallel table.

[0016] The suspension rod has a zigzag line segment morphology and a structural rule that the zigzag line segment forms a trapezoidal shape;

[0017] Within the detection area where the bounding box is located, target line segments that conform to the structural characteristics of the component are extracted based on the line segment shape and structural rules of the component, including:

[0018] Within the detection area where the first bounding box is located, the stirrup segments that conform to the stirrup structural characteristics are extracted by combining the straight line segment shape of the stirrups and the structural rules of the parallel table of multiple straight line segments within the detection area.

[0019] Within the detection area where the second bounding box is located, based on the morphology of the broken line segments of the suspension rod and the structural rule of the broken line segments forming a trapezoidal representation, the suspension rod segments that conform to the structural characteristics of the suspension rod within the detection area are extracted.

[0020] In some instances, the process of filtering target line segments and matching and binding them with corresponding annotation information based on the relative positional relationship between the components and the beam structure, combined with the direction of the annotation lines and text information, includes:

[0021] Identify the target stirrup segments that conform to the relative positional relationship between the stirrups and the beam structure, and combine the direction of the annotation lines and text information to complete the screening of target stirrup segments and the matching and binding of target stirrup segments with corresponding annotation information;

[0022] Identify the target line segments of the suspension rods that conform to the relative positional relationship between the suspension rods and the beam structure, and combine the direction of the annotation lines and text information to complete the screening of target line segments of the suspension rods and the matching and binding of the target line segments of the suspension rods with the corresponding annotation information.

[0023] In some instances, the process of combining the direction of the annotation line with text information to complete the selection of stirrup target segments and the matching and binding of stirrup target segments with corresponding annotation information includes:

[0024] Remove target line segments with incorrect labeling direction and incorrect text information from the target line segments of stirrups, and complete the screening of target line segments of stirrups and the matching and binding of target line segments of stirrups with corresponding labeling information;

[0025] The process of combining the direction of the annotation line with the text information to complete the selection of target line segments for the suspension rod and the matching and binding of the target line segments with the corresponding annotation information includes:

[0026] Remove target line segments with incorrect label direction and incorrect text information from the target line segments of the suspension rod, and complete the screening of target line segments of the suspension rod and the matching and binding of target line segments of the suspension rod with corresponding label information.

[0027] In some instances, the target detection model is an improved YOLOv5 model;

[0028] The improved YOLOv5 model consists of one shallow feature layer and three deep feature layers, and is constructed based on the BiFPN structure.

[0029] The three deep feature layers are sequentially increased in level, consisting of the second feature layer, the third feature layer, and the fourth feature layer.

[0030] In some instances, the text information includes layer attribute information;

[0031] The process of combining the direction of the annotation line with the text information to complete the selection of stirrup target segments and the matching and binding of stirrup target segments with corresponding annotation information includes:

[0032] By combining the direction of the annotation line with the distribution of the stirrup layer, the target layer category with a layer information ratio exceeding a predetermined proportion is selected, and the stirrup target line segments that do not belong to the target layer category are removed, thus completing the stirrup target line segment selection and the matching and binding of the stirrup target line segments with the corresponding annotation information.

[0033] The process of combining the direction of the annotation line with the text information to complete the screening of target line segments for stirrups and the matching and binding of target line segments for hoop bars with corresponding annotation information includes:

[0034] By combining the direction of the annotation lines with the layer distribution of the stirrups, the target layer categories whose layer information accounts for more than a predetermined proportion are selected, and the target line segments of the stirrups that do not belong to the target layer categories are removed, thus completing the selection of target line segments of the stirrups and the matching and binding of target line segments of the stirrups with the corresponding annotation information.

[0035] A second aspect of the present invention provides an information identification device in structural drawings, comprising:

[0036] The frame cutting unit is used to cut a target frame using a sliding window to obtain multiple blocks of predetermined sizes; wherein the target frame has beam structure information;

[0037] The component positioning unit is used to input the block into the target detection model, combine the component graphic features of the beam structure component, perform component positioning, and output the bounding box position and category label of the component.

[0038] The line segment extraction unit is used to extract target line segments within the detection area where the bounding box is located, based on the line segment shape and structural rules of the component, to meet the structural characteristics of the component.

[0039] The line segment matching unit is used to filter target line segments and match and bind target line segments with corresponding annotation information based on the relative positional relationship between components and beam structures, combined with the direction of annotation lines and text information.

[0040] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps of the method described in the first aspect above.

[0041] A fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described in the first aspect above.

[0042] The present invention discloses a method for identifying information in structural drawings, comprising: cutting a target drawing frame using a sliding window to obtain multiple blocks of predetermined sizes; wherein the target drawing frame contains beam structure information; inputting the blocks into a target detection model, locating the components by combining the component graphic features of the beam structure components, and outputting the bounding box position and category label of the components; within the detection area where the bounding box position is located, extracting target line segments within the detection area that conform to the component's structural features by combining the line segment shape and construction rules of the components; and completing the selection of target line segments and the matching and binding of target line segments with corresponding annotation information based on the relative positional relationship between the components and the beam structure, combined with the direction of the annotation lines and text information. In this application, a target detection model is used to locate components and output the bounding box position and category label of the components. Within the detection area where the bounding box position is located, target line segments that conform to the structural characteristics of the components are extracted based on the line segment shape and construction rules of the components. Based on the relative positional relationship between the components and the beam structure, combined with the direction of the annotation lines and text information, the target line segments are filtered and matched and bound with the corresponding annotation information. This is beneficial for identifying the corresponding component line segments according to the different line segment shapes and construction rules of the components, and for completing the target line segment filtering and matching and binding of the target line segments with the corresponding annotation information, thereby improving the efficiency of drawing recognition and review. Attached Figure Description

[0043] Figure 1 A flowchart of a method for identifying information in structural drawings provided by an embodiment of the present invention;

[0044] Figure 2 A schematic diagram of an improved YOLOv5 model structure provided in an embodiment of the present invention;

[0045] Figure 3 This is a flowchart of information identification in structural drawings provided in an embodiment of the present invention;

[0046] Figure 4 The structural drawing provided in the embodiment of the present invention is a schematic diagram of the information recognition device. Detailed Implementation

[0047] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] The structural drawing information identification method proposed in this invention addresses the technical problem of the difficulty of manual review of structural drawings. It uses a target detection model to locate components and outputs the bounding box position and category label of each component. Within the detection area where the bounding box is located, target line segments that conform to the structural characteristics of the component are extracted based on the line segment shape and construction rules of the component. Based on the relative positional relationship between the component and the beam structure, combined with the direction of the annotation lines and text information, target line segment filtering and matching / binding with corresponding annotation information are completed. This facilitates the identification of corresponding component line segments based on different component line segment shapes and construction rules, thereby improving the efficiency of drawing recognition and review.

[0049] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be understood more clearly and thoroughly, and that the scope of the present invention can be fully conveyed to those skilled in the art.

[0050] Figure 1 This is a flowchart illustrating a method for identifying information in structural drawings, provided as an embodiment of the present invention. Figure 1 As shown, the information identification method in structural drawings proposed in this embodiment of the invention includes:

[0051] Step 100: Use a sliding window to cut the target frame to obtain multiple blocks of predetermined size; wherein, the target frame has beam structure information;

[0052] Step 110: Input the image block into the target detection model, combine it with the component graphic features of the beam structure component, locate the component, and output the bounding box position and category label of the component;

[0053] Step 120: Within the detection area where the bounding box is located, extract the target line segments within the detection area that conform to the structural characteristics of the component, based on the line segment shape and structural rules of the component.

[0054] Step 130: Based on the relative positional relationship between the components and the beam structure, and combined with the direction of the annotation lines and text information, complete the selection of target line segments and the matching and binding of target line segments with corresponding annotation information.

[0055] In this exemplary embodiment, a sliding window is used to cut the target frame to obtain multiple blocks of predetermined sizes, including:

[0056] During the process of cutting the target drawing frame using a sliding window, the size of the sliding window is adjusted according to the density of the components within the target drawing frame.

[0057] The target frame is cut using a sliding window that adjusts in real time to obtain multiple blocks of predetermined sizes.

[0058] The target frame is cut using a sliding window that adjusts in real time to obtain multiple blocks of predetermined sizes, including:

[0059] The priority of the region to be segmented is determined based on the density of component distribution within the target drawing frame.

[0060] Based on the priority of the regions to be segmented, the target frame is cut into multiple blocks of predetermined sizes, ordered from high to low density.

[0061] In this exemplary embodiment, the structural drawing can be a CAD drawing, and the drawing frame can be a sub-drawing frame of the CAD drawing. A sliding window method is used to cut the sub-drawing frame, allowing the model to effectively adapt to drawing frame inputs of different sizes. Next, a YOLOv5-based object detection method is used to detect and identify the main components (additional stirrups and hangers) in the beam quantity calculation drawing. Then, based on the existing drawing results, the line segments of the entire sub-drawing frame are filtered, and the existing beam results are used to determine the positions of the additional stirrups and hangers. Finally, after obtaining the specific line segments and positions of the additional stirrups and hangers, the annotation information of the hangers is matched.

[0062] This application can not only accurately identify additional stirrups, hangers, and other components, but also match them with the corresponding annotation information, greatly improving work efficiency and accuracy.

[0063] In this exemplary embodiment, since the drawing size of the beam sub-frame is large, directly inputting it into the deep learning model for processing results in insufficient accuracy and high computational load. Therefore, this application uses a sliding window method to cut the sub-frame image and generate several blocks of predetermined size so as to send them into the target detection model for processing more efficiently.

[0064] Secondly, based on the trained deep learning object detection model, the additional stirrups and hangers in the segmented images are initially located. This stage utilizes the typical graphic features of the additional stirrups and hangers in the structural diagram to accurately identify their location regions.

[0065] Subsequently, based on the above-mentioned detection area, the line segment elements within the area are finely screened, and at the same time, other line elements with similar characteristics in the whole picture are analyzed to further extract possible additional stirrup and hanger line segments.

[0066] Next, by matching and judging the graphic features such as line segment shape, positional relationship, and structural rules, the target line segments that truly conform to the structural features of additional stirrups and hangers are selected.

[0067] Finally, for the identified suspension rod areas, the direction of the annotation lines and the text position in the figure are further combined to complete the matching and binding between the suspension rods and their corresponding annotation information.

[0068] In this application, a target detection model is used to locate components and output the bounding box position and category label of the components. Within the detection area where the bounding box position is located, target line segments that conform to the structural characteristics of the components are extracted based on the line segment shape and construction rules of the components. Based on the relative positional relationship between the components and the beam structure, combined with the direction of the annotation lines and text information, the target line segments are filtered and matched and bound with the corresponding annotation information. This is beneficial for identifying the corresponding component line segments according to the different line segment shapes and construction rules of the components, and for completing the target line segment filtering and matching and binding of the target line segments with the corresponding annotation information, thereby improving the efficiency of drawing recognition and review.

[0069] In some instances, the category labels include at least stirrups and hangers;

[0070] The process of inputting the image blocks into the target detection model, combining the component graphic features of the beam structure members, locating the components, and outputting the bounding box position and category label of the components includes:

[0071] The image blocks are input into the target detection model. Based on the component graphic features of the beam structure components, the components are located, and the first boundary box position of the stirrups and the second boundary box position of the hanger are output.

[0072] In this exemplary embodiment, both stirrups and hangers are beam structural members. The component graphic features of stirrups and hangers are different. During frame recognition, the blocks can be input into the target detection model, and combined with the component graphic features of the beam structural members, the components are located, and the first bounding box position of the stirrups and the second bounding box position of the hangers are output, thereby achieving preliminary positioning of the additional stirrups and hangers in the sliced ​​image.

[0073] In some instances, the stirrups have a straight line segment shape and a structural pattern characterized by multiple straight line segments arranged in a parallel table.

[0074] The suspension rod has a zigzag line segment morphology and a structural rule that the zigzag line segment forms a trapezoidal shape;

[0075] Within the detection area where the bounding box is located, target line segments that conform to the structural characteristics of the component are extracted based on the line segment shape and structural rules of the component, including:

[0076] Within the detection area where the first bounding box is located, the stirrup segments that conform to the stirrup structural characteristics are extracted by combining the straight line segment shape of the stirrups and the structural rules of the parallel table of multiple straight line segments within the detection area.

[0077] Within the detection area where the second bounding box is located, based on the morphology of the broken line segments of the suspension rod and the structural rule of the broken line segments forming a trapezoidal representation, the suspension rod segments that conform to the structural characteristics of the suspension rod within the detection area are extracted.

[0078] In this exemplary embodiment, since the line segment shapes and structural patterns of stirrups and hangers are different, stirrup line segments that conform to the structural characteristics of stirrups and hanger line segments that conform to the structural characteristics of hangers can be extracted within the detection area based on their line segment shapes and structural patterns. In this way, the extraction of stirrup line segments and hanger line segments can be achieved.

[0079] In some instances, the process of filtering target line segments and matching and binding them with corresponding annotation information based on the relative positional relationship between the components and the beam structure, combined with the direction of the annotation lines and text information, includes:

[0080] Identify the target stirrup segments that conform to the relative positional relationship between the stirrups and the beam structure, and combine the direction of the annotation lines and text information to complete the screening of target stirrup segments and the matching and binding of target stirrup segments with corresponding annotation information;

[0081] Identify the target line segments of the suspension rods that conform to the relative positional relationship between the suspension rods and the beam structure, and combine the direction of the annotation lines and text information to complete the screening of target line segments of the suspension rods and the matching and binding of the target line segments of the suspension rods with the corresponding annotation information.

[0082] In this exemplary embodiment, the text information may include component number, reinforcement description, dimensioning, text notes, and structural details.

[0083] The process of combining the direction of the annotation line with the text information to complete the selection of stirrup target segments and the matching and binding of stirrup target segments with corresponding annotation information includes:

[0084] Obtain the dimensions and scale of the target stirrup line segments in the drawing frame, and match them with the dimension annotations in the text information to complete the filtering of target stirrup line segments and the matching and binding of target stirrup line segments with corresponding annotation information.

[0085] The process of combining the direction of the annotation line with the text information to complete the selection of target line segments for the suspension rod and the matching and binding of the target line segments with the corresponding annotation information includes:

[0086] Obtain the dimensions and scale of the target line segments of the suspension rods in the drawing frame, and match them with the dimension annotations in the text information to complete the filtering of the target line segments of the suspension rods and the matching and binding of the target line segments of the suspension rods with the corresponding annotation information.

[0087] In this exemplary embodiment, since the stirrups, hangers, and beam structure all have relative positions, and the stirrups and hangers are all located at predetermined positions near the beam structure, target line segments for stirrups and hangers that conform to the relative positional relationship between the stirrups and the beam structure can be determined, and line segments that do not conform to the relative positional relationship can be eliminated. Then, the line segments that conform to the relative positional relationship are combined with the direction of the annotation line and the text information to complete the screening of target line segments for hangers and the matching and binding of the target line segments for hangers with the corresponding annotation information.

[0088] In some instances, the process of combining the direction of the annotation line with text information to complete the selection of stirrup target segments and the matching and binding of stirrup target segments with corresponding annotation information includes:

[0089] Remove target line segments with incorrect labeling direction and incorrect text information from the target line segments of stirrups, and complete the screening of target line segments of stirrups and the matching and binding of target line segments of stirrups with corresponding labeling information;

[0090] The process of combining the direction of the annotation line with the text information to complete the selection of target line segments for the suspension rod and the matching and binding of the target line segments with the corresponding annotation information includes:

[0091] Remove target line segments with incorrect label direction and incorrect text information from the target line segments of the suspension rod, and complete the screening of target line segments of the suspension rod and the matching and binding of target line segments of the suspension rod with corresponding label information.

[0092] In this exemplary embodiment, target line segments with incorrect labeling direction and incorrect text information can be removed from the stirrup target line segments. Then, based on the remaining stirrup target line segments and hanger target line segments, text information correction is performed to complete the matching and binding of the corresponding labeling information of the stirrup target line segments and hanger target line segments.

[0093] In some instances, the target detection model is an improved YOLOv5 model;

[0094] The improved YOLOv5 model consists of one shallow feature layer and three deep feature layers, and is constructed based on the BiFPN structure.

[0095] The three deep feature layers are sequentially increased in level, consisting of the second feature layer, the third feature layer, and the fourth feature layer.

[0096] In this application, the improved YOLOv5 model includes:

[0097] A new shallow feature extraction channel is added, and the feature fusion between shallow and deep layers is enhanced through the BiFPN (Bidirectional Feature Pyramid Network) structure.

[0098] This application adds a shallow feature extraction channel (P2 layer) to the YOLOv5 network structure and enhances the feature fusion mechanism between shallow and deep layers (introducing a BiFPN structure) to preserve fine line texture information in the image, thereby improving the detection capability of small linear targets such as hanging rods and additional stirrups. This structural optimization not only improves the recall rate of small targets but also maintains detection efficiency.

[0099] Based on the size distribution of small targets in the training set, the default Anchor settings of YOLOv5 are reconfigured using K-means clustering to better match the geometric scale of the additional stirrups and hanger symbols.

[0100] In this exemplary embodiment, Figure 2 This is a schematic diagram of an improved YOLOv5 model structure provided in an embodiment of the present invention. Figure 2 As shown, in the YOLOv5 + BiFPN architecture, the improved YOLOv5 model includes: a first feature layer P2, a second feature layer P3, a third feature layer P4, and a fourth feature layer P5. Here, the first feature layer P2, the second feature layer P3, the third feature layer P4, and the fourth feature layer P5 are feature maps representing different resolutions and semantic levels, respectively. P2, P3, P4, and P5 represent feature maps at different resolutions and semantic levels, supporting the detection of multi-scale objects ranging from small to large targets. Specifically:

[0101] P2 is a shallow, high-resolution feature map with a downsampling rate of 4 (input ÷ 4), typically 160 × 160 in size (when the input is 640 × 640). It preserves the most spatial detail and is well-suited for small object detection. By introducing BiFPN, this feature map is fused with other layers in the neck, thereby enhancing the representation of small objects.

[0102] P3 is a feature map with a downsampling rate of 8 (640 ÷ 8 = 80). As the first default detection layer, it integrates semantic and some spatial information and is an important node in small and medium-scale detection.

[0103] P4 corresponds to a downsampling rate of 16 (output is 40×40). It has stronger semantic representation, but lower spatial resolution compared to P3, and is a key layer for medium-scale object recognition.

[0104] P5 is a feature map with a downsampling rate of 32 (output is 20×20), which has the strongest semantic information, but less spatial detail, and is suitable for detecting larger targets.

[0105] Through the above optimization measures, the model significantly improves the detection accuracy and robustness of small target components such as additional stirrups and hangers while maintaining a high detection speed.

[0106] In some instances, the text information includes layer attribute information;

[0107] The process of combining the direction of the annotation line with the text information to complete the selection of stirrup target segments and the matching and binding of stirrup target segments with corresponding annotation information includes:

[0108] By combining the direction of the annotation line with the distribution of the stirrup layer, the target layer category with a layer information ratio exceeding a predetermined proportion is selected, and the stirrup target line segments that do not belong to the target layer category are removed, thus completing the stirrup target line segment selection and the matching and binding of the stirrup target line segments with the corresponding annotation information.

[0109] The process of combining the direction of the annotation line with the text information to complete the screening of target line segments for stirrups and the matching and binding of target line segments for hoop bars with corresponding annotation information includes:

[0110] By combining the direction of the annotation lines with the layer distribution of the stirrups, the target layer categories whose layer information accounts for more than a predetermined proportion are selected, and the target line segments of the stirrups that do not belong to the target layer categories are removed, thus completing the selection of target line segments of the stirrups and the matching and binding of target line segments of the stirrups with the corresponding annotation information.

[0111] In this exemplary embodiment, each line segment has layer attributes. In this application, some line segments with layer attributes different from the majority can be removed. Then, based on the remaining line segments, the target line segments for the stirrups are filtered, and the target line segments for the hoop bars are matched and bound to the corresponding annotation information. The predetermined ratio can be less than 20%.

[0112] In some instances, the annotation line includes an extended annotation line; the step of combining the annotation line direction and text information to complete the selection of stirrup target segments and the matching and binding of stirrup target segments with corresponding annotation information includes:

[0113] Determine whether there are extended annotation lines based on geometric relationships;

[0114] If extended label lines exist, analyze whether the extended label lines are related to the text elements of stirrup category or specification;

[0115] If the extended annotation line is associated with the text element of the stirrup category or specification, then based on the association relationship between the extended annotation line and the text element of the stirrup category or specification, the stirrup target line segment is filtered and the stirrup target line segment is matched and bound with the corresponding annotation information.

[0116] In this exemplary embodiment, the annotation line includes an extended annotation line; the step of combining the annotation line direction and text information to complete the screening of target line segments for the suspension rod and the matching and binding of the target line segments for the suspension rod with the corresponding annotation information includes:

[0117] Determine whether there are extended annotation lines based on geometric relationships;

[0118] If extended label lines exist, analyze whether the extended label lines are related to the text elements of the hanging rod category or specification;

[0119] If the extended annotation line is associated with the text elements of the suspension rod category or specification, then based on the association between the extended annotation line and the text elements of the suspension rod category or specification, the target line segment of the suspension rod is filtered and the target line segment of the suspension rod is matched and bound with the corresponding annotation information.

[0120] In this exemplary embodiment, analyzing whether the extended annotation line is associated with a text element of the stirrup category or specification includes:

[0121] Analyze whether the extended annotation lines contain text elements indicating the type or specifications of stirrups;

[0122] If the extended annotation line contains text elements indicating the type or specification of stirrups, then the extended annotation line is considered to be associated with these text elements, indicating a relationship. Based on this relationship, the target stirrup segments are then selected, and the target stirrup segments are matched and bound to their corresponding annotation information.

[0123] Figure 3 This is a flowchart illustrating the information identification process in structural drawings provided for embodiments of the present invention. For example... Figure 3 As shown, the information identification process in the structural drawings includes:

[0124] Step 30: Begin;

[0125] Step 31: Cutting the sub-frame sliding window and detecting additional stirrups / hanging bars;

[0126] Step 32: Filtering detection results, filtering original image line segments, and filtering image features;

[0127] Step 33: Match the hanging rod marking lines;

[0128] Step 34, End.

[0129] This invention provides a device for identifying information in structural drawings. Figure 4 This is a structural diagram of the information recognition device provided in the structural drawings for an embodiment of the present invention. For example... Figure 4 As shown in the structural drawing, the information recognition device includes:

[0130] The frame cutting unit 40 is used to cut the target frame using a sliding window to obtain multiple blocks of predetermined size; wherein the target frame has beam structure information;

[0131] The component positioning unit 41 is used to input the block into the target detection model, combine the component graphic features of the beam structure component, perform component positioning, and output the bounding box position and category label of the component.

[0132] The line segment extraction unit 42 is used to extract target line segments within the detection area where the bounding box is located, based on the line segment shape and structural rules of the component, to meet the structural characteristics of the component.

[0133] The line segment matching unit 43 is used to complete the selection of target line segments and the matching and binding of target line segments with corresponding annotation information based on the relative positional relationship between the component and the beam structure, combined with the direction of the annotation line and text information.

[0134] In this exemplary embodiment, a sliding window approach is used to segment the sub-frame, allowing the model to effectively adapt to input frames of different sizes. Next, a YOLOv5-based object detection method is used to detect and identify the main components (additional stirrups and hangers) in the beam quantity calculation drawing. Then, based on the existing drawing results, the line segments of the entire sub-frame are filtered, and the positions of the additional stirrups and hangers are determined using existing beam results. Finally, after obtaining the specific line segments and positions of the additional stirrups and hangers, the annotation information of the hangers is matched.

[0135] This application can not only accurately identify additional stirrups, hangers, and other components, but also match them with the corresponding annotation information, greatly improving work efficiency and accuracy.

[0136] In this exemplary embodiment, since the drawing size of the beam sub-frame is large, directly inputting it into the deep learning model for processing results in insufficient accuracy and high computational load. Therefore, this application uses a sliding window method to cut the sub-frame image and generate several blocks of predetermined size so as to send them into the target detection model for processing more efficiently.

[0137] Secondly, based on the trained deep learning object detection model, the additional stirrups and hangers in the segmented images are initially located. This stage utilizes the typical graphic features of the additional stirrups and hangers in the structural diagram to accurately identify their location regions.

[0138] Subsequently, based on the above-mentioned detection area, the line segment elements within the area are finely screened, and at the same time, other line elements with similar characteristics in the whole picture are analyzed to further extract possible additional stirrup and hanger line segments.

[0139] Next, by matching and judging the graphic features such as line segment shape, positional relationship, and structural rules, the target line segments that truly conform to the structural features of additional stirrups and hangers are selected.

[0140] Finally, for the identified suspension rod areas, the direction of the annotation lines and the text position in the figure are further combined to complete the matching and binding between the suspension rods and their corresponding annotation information.

[0141] In this application, a target detection model is used to locate components and output the bounding box position and category label of the components. Within the detection area where the bounding box position is located, target line segments that conform to the structural characteristics of the components are extracted based on the line segment shape and construction rules of the components. Based on the relative positional relationship between the components and the beam structure, combined with the direction of the annotation lines and text information, the target line segments are filtered and matched and bound with the corresponding annotation information. This is beneficial for identifying the corresponding component line segments according to the different line segment shapes and construction rules of the components, and for completing the target line segment filtering and matching and binding of the target line segments with the corresponding annotation information, thereby improving the efficiency of drawing recognition and review.

[0142] Since the systems / devices described in the above embodiments of the present invention are systems / devices used to implement the methods of the above embodiments of the present invention, those skilled in the art can understand the specific structure and modifications of the systems / devices based on the methods described in the above embodiments of the present invention, and therefore will not be repeated here. All systems / devices used in the methods of the above embodiments of the present invention fall within the scope of protection of the present invention.

[0143] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps of the method described in the first aspect above.

[0144] A fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described in the first aspect above.

[0145] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0146] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0147] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first and second features are in direct contact, or that they are in indirect contact through an intermediate medium. Furthermore, "above," "over," or "on top" the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," or "beneath" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0148] In the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0149] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make modifications, alterations, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for identifying information in structural drawings, characterized in that, include: A sliding window is used to cut the target frame to obtain multiple blocks of predetermined sizes; wherein, the target frame contains beam structure information; The image blocks are input into the target detection model, and the components are located by combining the component graphic features of the beam structure components. The bounding box position and category label of the components are then output. Within the detection area where the bounding box is located, target line segments that conform to the structural characteristics of the component are extracted based on the line segment shape and structural rules of the component. Based on the relative positional relationship between the components and the beam structure, and combined with the direction of the annotation lines and text information, the target line segments are filtered and matched and bound to the corresponding annotation information.

2. The method according to claim 1, characterized in that, The category labels include at least stirrups and hangers; The process of inputting the image blocks into the target detection model, combining the component graphic features of the beam structure members, locating the components, and outputting the bounding box position and category label of the components includes: The image blocks are input into the target detection model. Based on the component graphic features of the beam structure components, the components are located, and the first boundary box position of the stirrups and the second boundary box position of the hanger are output.

3. The method according to claim 2, characterized in that, The stirrups have a straight line segment shape and a structural pattern characterized by multiple straight line segments arranged in a parallel table. The suspension rod has a zigzag line segment morphology and a structural rule that the zigzag line segment forms a trapezoidal shape; Within the detection area where the bounding box is located, target line segments that conform to the structural characteristics of the component are extracted based on the line segment shape and structural rules of the component, including: Within the detection area where the first bounding box is located, the stirrup segments that conform to the stirrup structural characteristics are extracted by combining the straight line segment shape of the stirrups and the structural rules of the parallel table of multiple straight line segments within the detection area. Within the detection area where the second bounding box is located, based on the morphology of the broken line segments of the suspension rod and the structural rule of the broken line segments forming a trapezoidal representation, the suspension rod segments that conform to the structural characteristics of the suspension rod within the detection area are extracted.

4. The method according to claim 3, characterized in that, The process of filtering target line segments and matching and binding them with corresponding annotation information, based on the relative positional relationship between the components and the beam structure, combined with the direction of the annotation lines and text information, includes: Identify the target stirrup segments that conform to the relative positional relationship between the stirrups and the beam structure, and combine the direction of the annotation lines and text information to complete the screening of target stirrup segments and the matching and binding of target stirrup segments with corresponding annotation information; Identify the target line segments of the suspension rods that conform to the relative positional relationship between the suspension rods and the beam structure, and combine the direction of the annotation lines and text information to complete the screening of target line segments of the suspension rods and the matching and binding of the target line segments of the suspension rods with the corresponding annotation information.

5. The method according to claim 4, characterized in that, The process of combining the direction of the annotation line with the text information to complete the selection of stirrup target segments and the matching and binding of stirrup target segments with corresponding annotation information includes: Remove target line segments with incorrect labeling direction and incorrect text information from the target line segments of stirrups, and complete the screening of target line segments of stirrups and the matching and binding of target line segments of stirrups with corresponding labeling information; The process of combining the direction of the annotation line with the text information to complete the selection of target line segments for the suspension rod and the matching and binding of the target line segments with the corresponding annotation information includes: Remove target line segments with incorrect label direction and incorrect text information from the target line segments of the suspension rod, and complete the screening of target line segments of the suspension rod and the matching and binding of target line segments of the suspension rod with corresponding label information.

6. The method according to any one of claims 1-5, characterized in that, The target detection model is an improved YOLOv5 model; The improved YOLOv5 model consists of one shallow feature layer and three deep feature layers, and is constructed based on the BiFPN structure. The three deep feature layers are sequentially increased in level, consisting of the second feature layer, the third feature layer, and the fourth feature layer.

7. The method according to claim 4, characterized in that, The text information includes layer attribute information; The process of combining the direction of the annotation line with the text information to complete the selection of stirrup target segments and the matching and binding of stirrup target segments with corresponding annotation information includes: By combining the direction of the annotation line with the distribution of the stirrup layer, the target layer category with a layer information ratio exceeding a predetermined proportion is selected, and the stirrup target line segments that do not belong to the target layer category are removed, thus completing the stirrup target line segment selection and the matching and binding of the stirrup target line segments with the corresponding annotation information. The process of combining the direction of the annotation line with the text information to complete the screening of target line segments for stirrups and the matching and binding of target line segments for hoop bars with corresponding annotation information includes: By combining the direction of the annotation lines with the layer distribution of the stirrups, the target layer categories whose layer information accounts for more than a predetermined proportion are selected, and the target line segments of the stirrups that do not belong to the target layer categories are removed, thus completing the selection of target line segments of the stirrups and the matching and binding of target line segments of the stirrups with the corresponding annotation information.

8. An information recognition device in structural drawings, characterized in that, include: The frame cutting unit is used to cut a target frame using a sliding window to obtain multiple blocks of predetermined sizes; wherein the target frame has beam structure information; The component positioning unit is used to input the block into the target detection model, combine the component graphic features of the beam structure component, perform component positioning, and output the bounding box position and category label of the component. The line segment extraction unit is used to extract target line segments within the detection area where the bounding box is located, based on the line segment shape and structural rules of the component, to meet the structural characteristics of the component. The line segment matching unit is used to filter target line segments and match and bind target line segments with corresponding annotation information based on the relative positional relationship between components and beam structures, combined with the direction of annotation lines and text information.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1 to 7.

Images