Intelligent recognition method and system for steel bar engineering drawing, storage medium and equipment

Abstract

The invention discloses an intelligent recognition method and system for a steel bar engineering drawing, a storage medium and equipment, and relates to the technical field of drawing recognition. The method comprises the following steps: dividing an electronic image of a steel bar engineering drawing into a plurality of identification areas; detecting a labeling relationship between the labeling line and the steel bar component in each identification area, and determining a target identification area with an ambiguous scene based on the labeling relationship; constructing a spatial feature grid corresponding to the target recognition area, and determining a target steel bar symbol pointed by the marking line and a corresponding component parameter based on the spatial feature grid; and obtaining associated steel bar information of the target steel bar symbol in an adjacent identification area of the target identification area, verifying the component parameters based on the associated steel bar information, and generating an identification result of the target identification area. By implementing the technical scheme provided by the invention, the recognition accuracy of the reinforcing steel bar engineering drawings with ambiguous scenes can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of drawing recognition, and in particular to an intelligent recognition method and system for steel reinforcement engineering drawings, a storage medium and equipment. BACKGROUND

[0002] With the continuous expansion of construction projects, the number and complexity of steel reinforcement engineering drawings also increase. In order to improve design efficiency and construction accuracy, digitizing paper drawings and achieving intelligent recognition have become an important development direction in the current engineering field. In steel reinforcement engineering drawings, the annotation line is used to indicate the position and parameter information of the steel reinforcement component, and accurately understanding the correspondence between the annotation line and the steel reinforcement component is the key to realizing intelligent recognition of the drawing.

[0003] At present, the commonly used steel reinforcement engineering drawing recognition method is mainly based on image processing technology, which obtains an electronic image by scanning the drawing, and then uses a feature recognition algorithm to extract the annotation line and steel reinforcement symbol information in the drawing. However, in actual engineering applications, due to the dense arrangement of steel reinforcement components and the complex distribution of annotation lines, ambiguous scenes often occur, such as multiple annotation lines crossing and pointing, ambiguous steel reinforcement symbols, or overlapping components. These complex scenes can cause ambiguity in the recognition of the annotation relationship, reducing the accuracy of the recognition and affecting subsequent engineering design and construction. SUMMARY

[0004] The present application provides an intelligent recognition method, system, storage medium and equipment for steel reinforcement engineering drawings, which can improve the recognition accuracy of steel reinforcement engineering drawings with ambiguous scenes.

[0005] In a first aspect, the present application provides an intelligent recognition method for steel reinforcement engineering drawings, the method comprising: obtaining an electronic image corresponding to a steel reinforcement engineering drawing, and dividing the electronic image into a plurality of recognition regions; detecting the annotation relationship between the annotation line and the steel reinforcement component in each recognition region, and determining a target recognition region with an ambiguous scene based on the annotation relationship, the ambiguous scene including at least one of annotation pointing confusion, low steel reinforcement symbol recognition, and component overlap; constructing a spatial feature grid corresponding to the target recognition region, and obtaining the spatial distribution features of the annotation line, the position features of the steel reinforcement symbol, and the component contour features in the target recognition region based on the spatial feature grid; determining the target steel reinforcement symbol pointed by the annotation line according to the spatial distribution features of the annotation line and the position features of the steel reinforcement symbol, and extracting the component parameters corresponding to the target steel reinforcement symbol based on the component contour features; The recognition result generation module is configured to acquire associated steel bar information of the target steel bar symbol in an adjacent recognition area of the target recognition area, check the component parameter based on the associated steel bar information, and generate the recognition result of the target recognition area.

[0006] By adopting the technical scheme, the electronic image is divided into multiple recognition areas, and the target recognition area with an ambiguous scene is determined by detecting the marking relationship in each recognition area, so that the ambiguous area needing to be processed can be effectively located. For the determined target recognition area, the spatial feature grid is constructed, the spatial distribution feature of the marking line, the steel bar symbol position feature and the component contour feature are extracted, the corresponding relationship between the marking line and the target steel bar symbol is determined based on the feature information, and the component parameter is extracted based on the component contour feature, so that the accurate recognition of the ambiguous scene is realized. In addition, the component parameter is checked by acquiring the associated steel bar information in the adjacent recognition area, so that the accuracy and consistency of the recognition result can be further ensured. The scheme can effectively solve the recognition difficulty problem caused by the ambiguous scene such as marking direction confusion, low steel bar symbol recognition degree and component overlap in the steel bar engineering drawing, and improves the intelligent recognition accuracy of the steel bar engineering drawing.

[0007] In a second aspect of the present application, an intelligent recognition system for a steel bar engineering drawing is provided, and the system comprises: A recognition area division module is configured to acquire an electronic image corresponding to the steel bar engineering drawing, and divide the electronic image into multiple recognition areas. An ambiguous area identification module is configured to detect the marking relationship between the marking line and the steel bar component in each recognition area, and determine the target recognition area with an ambiguous scene based on the marking relationship. The ambiguous scene includes at least one of marking direction confusion, low steel bar symbol recognition degree and component overlap. A feature parameter extraction module is configured to construct a spatial feature grid corresponding to the target recognition area, and acquire the spatial distribution feature of the marking line, the steel bar symbol position feature and the component contour feature in the target recognition area based on the spatial feature grid. A corresponding relationship determination module is configured to determine the target steel bar symbol pointed by the marking line according to the spatial distribution feature of the marking line and the steel bar symbol position feature, and extract the component parameter corresponding to the target steel bar symbol based on the component contour feature. A recognition result generation module is configured to acquire associated steel bar information of the target steel bar symbol in an adjacent recognition area of the target recognition area, check the component parameter based on the associated steel bar information, and generate the recognition result of the target recognition area.

[0008] In a third aspect of the present application, a computer storage medium is provided, and the computer storage medium stores a plurality of instructions. The instructions are suitable for being loaded and executed by a processor to perform the method steps described above.

[0009] In a fourth aspect of the present application, an electronic device is provided, comprising: a processor and a memory; wherein the memory stores a computer program, and the computer program is adapted to be loaded and executed by the processor to perform the method steps described above.

[0010] To sum up, the one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: The present application can effectively locate the ambiguous region that needs to be processed by dividing the electronic image into multiple recognition regions and detecting the annotation relationship in each recognition region. For the determined target recognition region, the spatial feature grid is constructed, and the spatial distribution feature of the annotation line, the reinforcement symbol position feature, and the member contour feature are extracted. The corresponding relationship between the annotation line and the target reinforcement symbol is determined based on these feature information, and the member parameters are extracted based on the member contour feature, so as to realize accurate recognition of the ambiguous scene. In addition, the member parameters are verified by obtaining the associated reinforcement information in the adjacent recognition region, which can further ensure the accuracy and consistency of the recognition result. The scheme can effectively solve the recognition difficulty problem caused by the ambiguous scene such as annotation direction confusion, low recognition degree of reinforcement symbol, and member overlap in the reinforcement engineering drawing, and improves the intelligent recognition accuracy of the reinforcement engineering drawing. BRIEF DESCRIPTION OF DRAWINGS

[0011] Figure 1 is a flowchart of a method for intelligent recognition of a reinforcement engineering drawing provided by an embodiment of the present application; Figure 2 is a module schematic diagram of an intelligent recognition system for a reinforcement engineering drawing provided by an embodiment of the present application; Figure 3 is a structural schematic diagram of an electronic device provided by an embodiment of the present application.

[0012] Explanation of reference signs: 300, electronic device; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. DETAILED DESCRIPTION

[0013] In order for those skilled in the art to better understand the technical solutions in the present specification, the technical solutions in the embodiments of the present specification will be described clearly and completely in conjunction with the drawings in the embodiments of the present specification. Obviously, the described embodiments are only a part of the embodiments of the present application, not all embodiments.

[0014] In the description of the embodiments of the present application, the words "for example" or "for instance" are used to indicate an example, an instance, or an illustration. Any embodiment or design presented as "for example" or "for instance" in the embodiments of the present application should not be construed as being more preferred or advantageous than other embodiments or designs. Rather, the use of the words "for example" or "for instance" is intended to present concepts in a particular manner. The words "for example" or "for instance" are used in the embodiments of the present application to present concepts in a particular manner.

[0015] In the description of the embodiments of the present application, the term "a plurality of" means two or more. For example, a plurality of systems means two or more systems, and a plurality of screen terminals means two or more screen terminals. In addition, the terms "first" and "second" are used only for descriptive purposes and should not be construed as indicating or implying relative importance or implicitly indicating the technical features indicated. Therefore, the features defined with "first" and "second" can explicitly or implicitly include one or more features. The terms "include", "contain", "have" and their variants mean "include but are not limited to", unless otherwise specifically emphasized.

[0016] The technical solutions in the embodiments of the present application will be described clearly and completely in the embodiments of the present application in combination with the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments.

[0017] Please refer to Figure 1 , a flowchart of an intelligent identification method for steel reinforcement engineering drawings is specifically proposed. The method can be realized by relying on a computer program, can be realized by relying on a single-chip microcomputer, and can also run on an intelligent identification system for steel reinforcement engineering drawings. The computer program can be integrated in a computer device or can run as an independent tool application. Specifically, the method includes steps 10 to 50, and the steps are as follows: Step 10: Obtain an electronic image corresponding to the steel reinforcement engineering drawing, and divide the electronic image into a plurality of identification regions.

[0018] In the embodiments of the present application, the electronic image refers to an image file obtained by digitizing a paper steel reinforcement engineering drawing by a scanning device. The electronic image contains graphic information such as marking lines, steel reinforcement symbols, and component contours in the steel reinforcement engineering drawing.

[0019] The identification region refers to a plurality of sub-regions obtained by dividing the entire electronic image according to the content features of the drawing. Each identification region contains a certain number of marking lines and steel reinforcement components. By dividing the large-format electronic image into a plurality of identification regions, the complex identification task can be divided into a plurality of local identification tasks, which facilitates accurate identification and processing of the marking relationship.

[0020] Specifically, the steel reinforcement engineering drawings are first scanned at a resolution of 400 dpi using a paper scanner and saved as electronic images. Edge detection algorithms are used to extract the outlines of the steel reinforcement components in the images, and connected component analysis is employed to identify the boundary ranges of the components. For each component, the area and perimeter of its circumscribed rectangle are calculated, and the components are classified according to their area size. When the density of detected components reaches a preset value (e.g., more than 5 components per unit area), a 400×400 pixel recognition region is established centered on that area. For large components (area exceeding 10,000 square pixels), they are evenly divided into multiple sub-regions. A 100-pixel overlap band is set between adjacent recognition regions to handle cross-region component connections. This division method based on component distribution characteristics ensures a balance in the number and size of components within the recognition region, which helps improve recognition accuracy.

[0021] Based on the above embodiments, as another optional embodiment, the step of dividing the electronic image into multiple recognition regions may further include steps 101-103: Step 101: Obtain the distribution density of the annotation lines in the electronic image and generate a density heatmap based on the distribution density.

[0022] Specifically, the Canny edge detection algorithm is first used to extract the annotation line features in the electronic image, with dual threshold parameters of 50 and 150 to effectively distinguish the annotation lines from the background. Then, the electronic image is divided into 100×100 pixel grid cells, and the annotation line density is calculated for each grid cell. Specifically, the number of annotation line pixels in each grid cell is counted, and the annotation line density value for that cell is calculated by combining the length features of the annotation lines (obtained through Hough transform). To generate a continuous density distribution representation, a bilinear interpolation algorithm is used to interpolate the density values ​​of the grid cells, generating a resolution equal to the original. Figure 1 The density matrix is ​​1 / 10. Finally, the density matrix is ​​mapped to a heatmap and visualized using rainbow colors, where red represents the area with the highest density (density value greater than 0.8) and blue represents the area with the lowest density (density value less than 0.2), thus providing an intuitive display of the density distribution of the labeled lines.

[0023] Step 102: Determine the clustering area of ​​the marked lines based on the density heat map.

[0024] Specifically, first, set the density threshold to 0.7 (density value range 0-1), and perform binaryzation processing on the heat map. Then, use the region growing algorithm to perform connected region analysis on the binaryzation image, and set the minimum connected area to 1000 square pixels to exclude the influence of noise. For each detected connected region, calculate its centroid coordinates and area size. In order to avoid the interference of adjacent aggregation regions, set the minimum distance threshold to 200 pixels, and when the centroid distance of two aggregation regions is less than the threshold, merge the smaller region into the larger region.

[0025] Step 103: Take the aggregation region as a reference point for region division, and divide a plurality of recognition regions according to a preset size range centered on the reference point.

[0026] Specifically, the centroid coordinates of the aggregation region are taken as the reference point to construct the division scheme of the recognition region. First, set the reference size of the recognition region to 300x300 pixels, and establish a rectangular region of this size around each reference point. In order to adapt to different density distribution situations, an adaptive size adjustment strategy is adopted: when the density value at the reference point exceeds 0.9, the recognition region is enlarged to 400x400 pixels; when the density value is less than 0.5, it is reduced to 200x200 pixels. Considering the continuity of the annotation relationship, a 50-pixel overlap region is set between adjacent recognition regions. For the recognition regions at the edge of the drawing, adjust their shape and size according to the actual situation to ensure that the edge information is not lost. Finally, assign a unique number to each recognition region, and record its boundary coordinates, area, number of contained annotation lines, etc. to establish an index table of the recognition region, facilitating subsequent recognition processing.

[0027] Step 20: Detect the annotation relationship between the annotation lines and the steel reinforcement members in each recognition region, and determine the target recognition region with ambiguous scenarios based on the annotation relationship, the ambiguous scenarios including at least one of annotation pointing confusion, low steel symbol recognition degree, and member overlap.

[0028] In the embodiments of the present application, the annotation line refers to an indication line drawn from the steel reinforcement member to identify the parameters such as the steel reinforcement specification and size, which is composed of two straight lines: one is a diagonal line drawn from the steel reinforcement member, and the other is a horizontal annotation parameter line. The end of the annotation line is marked with the diameter specification of the steel reinforcement or other member parameters.

[0029] The target recognition region refers to a region in the electronic image that has ambiguous scenarios such as annotation pointing confusion, low steel symbol recognition degree, or member overlap. Due to reasons such as dense intersection of annotation lines, multiple annotation lines pointing to the same region, unclear steel symbol, or overlap of multiple members, the annotation relationship is difficult to accurately identify in such regions.

[0030] Specifically, first, the labeling relationship of each recognition region is detected, the labeling line features are extracted by an edge detection algorithm, and the starting point, ending point and direction vector of the labeling line are calculated. For each labeling line, the steel bar components within a 30-pixel range around the starting point are analyzed, and a preliminary correspondence between the labeling line and the steel bar component is established. When the following conditions are detected, the region is marked as a target recognition region with an ambiguous scene: 1) the starting points of multiple labeling lines are located within a 10-pixel range of the same steel bar component, causing labeling direction confusion; 2) the pixel gray value standard deviation of the steel symbol at the ending point of the labeling line is less than 0.1, or the edge contour incompleteness exceeds 20%, indicating low steel symbol recognition; 3) through contour overlap calculation, it is found that the overlapping area of two or more steel components exceeds 30% of their respective areas, i.e., there is component overlap. This multi-feature-based detection method can accurately locate complex regions that need to be processed, providing a target range for subsequent accurate recognition.

[0031] On the basis of the above-mentioned embodiments, as another optional embodiment, the step of determining the target recognition region with an ambiguous scene based on the labeling relationship can further include steps 201-203: Step 201: Obtain the graphical feature parameters of each recognition region, wherein the graphical feature parameters include labeling line features, steel symbol features and component features.

[0032] Specifically, for each recognition region, the graphical feature parameters can be extracted as follows: first, the labeling line features are extracted: the Canny edge detection is used to extract the labeling line contour, and the Hough transform is used to obtain the geometric parameters of the labeling line, including the starting point coordinates (x1, y1), the ending point coordinates (x2, y2), the line segment length L and the inclination angle θ. For each labeling line, the pixel gradient direction histogram within its neighborhood range is calculated to obtain the direction feature vector D of the labeling line. Then the steel symbol features are extracted: a 20x20 pixel detection window is set near the ending point of the labeling line, the symbol region is extracted using an image segmentation algorithm, and the area S, perimeter P, gray mean μ, gray standard deviation σ and contour completeness C of the symbol are calculated. Finally, the component features are extracted: the connected component analysis is used to obtain the circumscribed rectangle parameters of each steel component, including the center coordinates (cx, cy), the width w and the height h, etc.

[0033] Step 202: Calculate the feature scores of each evaluation index according to the pre-set feature evaluation index system and image feature parameters, wherein the evaluation indexes include labeling direction confusion index, symbol recognition index and component overlap index.

[0034] Specifically, based on the preset feature evaluation index system, the extracted graphic feature parameters are quantitatively evaluated, and the feature evaluation index system includes a label pointing confusion index, a symbol recognition degree index, and a component overlap index. For the label pointing confusion index, the number n of label lines within a 15-pixel range of each steel component neighborhood is calculated, and the spatial distribution density p of the starting points of these label lines is calculated. The confusion degree score S1 = a1 n + a2 p, wherein a1 and a2 are weight coefficients. For the symbol recognition degree index, the gray scale standard deviation s, the contour completeness C, and the edge clarity E (calculated by the Sobel operator) of the symbol are comprehensively considered. The recognition degree score S2 = b1 s + b2 C + b3 E, wherein b1, b2, and b3 are weight coefficients. For the component overlap index, the overlap area ratio r and the boundary distance d between adjacent components are calculated. The overlap degree score S3 = g1 r + g2 (1 / d), wherein g1 and g2 are weight coefficients. Each weight coefficient is obtained by training a large number of samples to ensure the accuracy of the evaluation results.

[0035] Step 203: Compare each feature score with a preset evaluation criterion, and regard the recognition region in which any feature score does not satisfy the corresponding evaluation criterion as a target recognition region.

[0036] Specifically, each feature score calculated is accurately compared with a preset evaluation criterion. The preset evaluation criterion includes: the threshold T1 = 0.75 of the label pointing confusion score S1, which indicates that there is serious pointing confusion when S1 > T1; the threshold T2 = 0.65 of the symbol recognition degree score S2, which indicates that the symbol recognition degree is low when S2 < T2; and the threshold T3 = 0.8 of the component overlap score S3, which indicates that there is obvious component overlap when S3 > T3. The three scores of each recognition region are compared one by one. When any score does not satisfy the corresponding evaluation criterion (i.e., S1 > T1 or S2 < T2 or S3 > T3), the recognition region is immediately marked as a target recognition region, and the specific item and score value that do not satisfy the criterion are recorded. This multi-threshold-based determination method can accurately screen the recognition region that needs to be processed, and at the same time retains the specific reason that leads to the determination of the target region, thereby providing a clear direction for subsequent processing.

[0037] Step 30: Construct a spatial feature grid corresponding to the target recognition region, and obtain the spatial distribution features of the label lines, the position features of the steel symbols, and the component contour features in the target recognition region based on the spatial feature grid.

[0038] In the embodiments of the present application, the spatial feature grid represents a regular grid array structure formed by uniformly dividing the recognition region according to a preset grid size (such as 20x20 pixels).

[0039] Specifically, a uniform spatial feature grid of 20x20 pixels is constructed for the target recognition region, and each grid cell has a unique row-column index (i, j) and center coordinates (x, y). First, the spatial distribution features of the marking lines are extracted: the edge of the marking lines is extracted by the Canny operator, and the density value d is obtained by calculating the number of marking line pixels in each grid cell; the direction angle θ (0°~180°) of the marking lines is obtained by Hough transform, and the direction angle is divided into 9 intervals, and the number of marking lines in each interval is counted to form the direction histogram H; the position of the start and end points of each marking line in the grid is calculated to generate the connection relationship matrix L, which records the direction of the marking lines across the grid. Then the position features of the reinforcement symbols are extracted: symbol detection is performed in each grid cell, and when a symbol is detected, the symbol type t (such as Φ20, Φ25, etc.), the symbol center coordinates (sx, sy), the symbol occupied grid range [i1:i2, j1:j2] and the symbol orientation angle α are recorded; the complete contour of the symbol is extracted by morphological operation, and the area s and perimeter p of the symbol are calculated to evaluate the integrity of the symbol. Finally, the component contour features are extracted: the gradient amplitude and direction in the grid cell are calculated using the Sobel operator to extract the component edge points; the coverage rate r and the distribution density p of the edge points of the component pixels in each grid cell are calculated; the B-spline curve is used to fit the component contour to obtain the key control point coordinate sequence P and the curvature information k. All feature parameters are stored in a three-dimensional feature tensor T[m, n, f], where m and n are the grid dimensions, and f is the feature dimension, realizing the structured representation of the spatial features of the target recognition region.

[0040] On the basis of the above-mentioned embodiments, as an optional embodiment, according to the step of obtaining the spatial distribution features of the marking lines, the position features of the reinforcement symbols and the component contour features in the target recognition region based on the spatial feature grid, the step 301- step 304 can also be included: Step 301: Calculate the feature density value of each grid cell in the target recognition region, including the marking line density, the reinforcement symbol density and the component density.

[0041] Specifically, the feature distribution is quantified by calculating the feature density value in each grid cell (20x20 pixels). For the label line density, the label line pixels are extracted using Canny edge detection, and the ratio of the number of label line pixels in the cell to the total number of pixels in the cell is calculated as dl; the number of label line intersection points nc is also counted, and the label line density value D1 = w1 x dl + w2 x nc is obtained, where w1 and w2 are weight coefficients. For the reinforcement symbol density, the symbol region is detected using a template matching method, and the symbol pixel ratio ds and the number of symbols ns are calculated to obtain the symbol density value D2 = w3 x ds + w4 x ns. For the component density, the component region is extracted by the region growing algorithm, and the component pixel ratio dc and the number of component edge pixels ne are calculated to obtain the component density value D3 = w5 x dc + w6 x ne. Finally, the feature density vector D = [D1, D2, D3] of the grid cell is obtained, which provides basic data for subsequent key region identification.

[0042] Step 302: Determine the key feature region of the target recognition area based on the feature density value, which includes the label line aggregation area, the reinforcement symbol distribution area, and the component dense area.

[0043] Specifically, the key feature region is determined based on the feature density value D = [D1, D2, D3]. First, the label line density value D1 is spatially clustered, and when the D1 values of adjacent grid cells are all above the threshold T1 = 0.6 and the connected area is greater than 100 pixels, these grid cells are marked as the label line aggregation area A1. Then the distribution of the reinforcement symbol density value D2 is analyzed, and when the D2 value of a certain grid cell exceeds the threshold T2 = 0.4 and there are at least 2 symbols in the adjacent domain, the region is expanded to form the reinforcement symbol distribution area A2. For the component density value D3, the region growing algorithm is used to start expanding from the grid cell with the largest D3 value, and when the difference in D3 value between adjacent cells is less than the threshold T3 = 0.2, the growth continues, and finally the component dense area A3 is formed. Through this adaptive region division method, the key feature region that needs to be analyzed is accurately located.

[0044] Step 303: Extract the label line direction and length parameters of the label line aggregation area, and generate the label line spatial distribution features of the target recognition area.

[0045] Specifically, the marking lines in the marking line gathering area A1 are parameterized. First, the Hough transform is used to detect straight line segments, and the parameter equation of each marking line is obtained as ρ = xcosθ + ysinθ. The direction angle θ (0°~180°) of the marking line is calculated, and the direction histogram H is obtained to determine the main direction θm. For each marking line, the length L and the endpoint coordinates (x1, y1), (x2, y2) are calculated, and a marking line description vector V = [θ, L, x1, y1, x2, y2] is generated. Then, the spatial organization characteristics of the marking lines are analyzed, the included angle α and the shortest distance d between adjacent marking lines are calculated, and the marking line correlation matrix R is established. Finally, the marking line spatial distribution feature F1 = {H, V, R} is generated, which comprehensively describes the geometric and topological features of the marking lines.

[0046] Step 304: Determine the symbol coordinates and area range of the steel bar symbol distribution area, and generate the steel bar symbol position feature of the target recognition area.

[0047] Specifically, the steel bar symbol distribution area A2 is precisely positioned and analyzed. First, for each detected symbol, the symbol center coordinates (cx, cy) are determined by centroid calculation, and the bounding box parameters [x, y, w, h] of the symbol are obtained using the minimum enclosing rectangle algorithm. Then, the orientation angle α and the scale factor s of the symbol are calculated, and the symbol feature vector S = [cx, cy, w, h, α, s] is established. For the determination of the symbol area range, an adaptive expansion strategy is adopted: when the gray value of the surrounding pixels changes by less than a threshold value δ = 0.1 and the connectivity is maintained, the area is included in the symbol range, and the symbol mask matrix M is obtained. Finally, the steel bar symbol position feature F2 = {S, M} is generated, realizing the precise positioning and range division of the symbol.

[0048] Step 305: Extract the edge contour point set of the component dense area, and generate the component contour feature of the target recognition area.

[0049] Specifically, the fine contour feature of the component dense area A3 is extracted. First, the Sobel operator is used to calculate the gradient amplitude map and direction θmap, and the initial edge point set E is extracted by non-maximum suppression. The edge points are connected: when the gradient direction difference between adjacent edge points is less than 15° and the spatial distance is less than 3 pixels, they are connected to form an edge chain. The RANSAC algorithm is used to fit the geometric primitives (straight line segments, circular arcs, etc.) of the edge chain, and the basic contour description of the component is obtained. Then, the key features of the contour are calculated: curvature k, corner β and length l, and the contour feature vector C = [k, β, l] is generated. Finally, the component contour feature F3 = {E, C} is constructed, providing geometric constraints for subsequent component recognition.

[0050] Step 40: Determine the target reinforcement symbol pointed by the annotation line according to the spatial distribution characteristics of the annotation line and the position characteristics of the reinforcement symbol, and extract the component parameters corresponding to the target reinforcement symbol based on the component contour characteristics.

[0051] In the embodiments of the present application, the target reinforcement symbol refers to the graphical symbol pointed by the annotation line for indicating the reinforcement specification and parameters; and the component parameters refer to the geometric and physical characteristics of the reinforcement component corresponding to the target reinforcement symbol.

[0052] Specifically, first, the spatial distribution characteristics of the annotation line are analyzed, and the direction vector v and the end point coordinates (x1, y1) and (x2, y2) of each annotation line are extracted. A fan-shaped search area of ±15° is set in the extension direction of the annotation line, and the search radius is 1.5 times the length of the annotation line. Based on the position characteristics of the reinforcement symbol, the perpendicular distance d and the direction angle θ of each symbol center point (cx, cy) in the search area to the annotation line are calculated, and a matching degree score S = w1 / d + w2×cos(θ) is generated. The symbol with the highest score and exceeding the threshold value of 0.8 is selected as the target reinforcement symbol, and its type information (such as Φ20) and position range [x, y, w, h] are extracted. Then, the complete component contour is extracted within a range of 30 pixels around the target symbol by using the component contour characteristics, and the component parameters are calculated, including but not limited to the center coordinates (px, py), the main direction α, the length l, the width w, and other geometric parameters. This method based on multiple feature constraints can accurately identify the annotation relationship and extract complete component information.

[0053] On the basis of the above-mentioned embodiments, as another optional embodiment, according to the spatial distribution characteristics of the annotation line and the position characteristics of the reinforcement symbol, the target reinforcement symbol pointed by the annotation line is determined, and based on the component contour characteristics, the component parameters corresponding to the target reinforcement symbol are extracted. This step can further include steps 401-404: Step 401: Extract the extension direction and end point position of the annotation line according to the spatial distribution characteristics of the annotation line, and establish the feature parameters pointed by each annotation line, including the extension distance, the direction angle, and the end point coordinates.

[0054] Specifically, the geometric features of each annotation line are extracted by analyzing the spatial distribution characteristics of the annotation line. First, the parameter equation of the annotation line is obtained by using the Hough transform, i.e. ρ = xcosθ + ysinθ, and the direction angle θ of the annotation line is calculated. The starting point coordinates (x1, y1) and the end point coordinates (x2, y2) of the annotation line are determined by the end point detection algorithm, and the length L of the annotation line is calculated as L = √[(x2-x1)²+(y2-y1)²]. According to the position of the end point, an extension vector v = (cosθ, sinθ) is established, and the extension distance D is set as 1.5 times the length of the annotation line. For each annotation line, a feature parameter vector F = [D, θ, (x2, y2)] is generated, where D is the extension distance, θ is the direction angle, and (x2, y2) is the end point coordinates. Through this parameterized description, the spatial pointing features of the annotation line can be accurately expressed, providing basic data for subsequent symbol matching.

[0055] Step 402: Calculate the matching degree score of the annotation line and each steel bar symbol according to the steel bar symbol position feature and the feature vector, and the matching degree score is determined by the distance decay coefficient and the angle deviation coefficient.

[0056] Specifically, the matching degree score is calculated based on the steel bar symbol position feature and the annotation line feature parameter. First, a search area is established with the end point (x2, y2) of the annotation line as the center and the extension distance D as the radius. For each steel bar symbol in the search area, the center coordinates (cx, cy) and the symbol range [x, y, w, h] are extracted. The distance decay coefficient is calculated as α = exp(-d² / 2σ²), where d is the Euclidean distance from the end point to the symbol center, and σ is the distance standard deviation (set as D / 3). The angle deviation coefficient is calculated as β = cos(θ-φ), where θ is the direction angle of the annotation line and φ is the angle of the line connecting the end point to the symbol center. The final matching degree score S is w1 × α + w2 × β, where w1 = 0.6 and w2 = 0.4 are the weight coefficients. This dual constraint scoring mechanism based on distance and angle can effectively identify the most likely annotation target.

[0057] Step 403: Select the steel bar symbol with the highest matching degree score as the target steel bar symbol, and obtain the positioning coordinates of the target steel bar symbol.

[0058] Specifically, the matching scores of all the symbols in the search area are sorted, and the symbol with the highest score and exceeding the threshold T=0.75 is selected as the target steel bar symbol. When there are multiple candidate symbols and the score difference is less than ε=0.1, an additional constraint is introduced: the angle between the label line and the line connecting the symbol center is calculated, and the symbol with the smallest angle is selected as the target symbol. The positioning coordinates P=(cx, cy) of the target steel bar symbol are obtained, and the type information t (such as Φ20) and the coverage range R=[x, y, w, h] of the symbol are recorded. Through this multi-level screening mechanism, the labeling target can be accurately determined in a complex labeling environment, and the complete location information can be extracted.

[0059] Step 404: Determine the contour boundary point set associated with the target steel bar symbol in the component contour feature according to the positioning coordinates, and extract the component parameters based on the contour boundary point set.

[0060] Specifically, a search window W (30x30 pixels) is set with the positioning coordinates P as the center, and the boundary point set E={pi} within the window is extracted in the component contour feature. The DBSCAN clustering algorithm is used to group the boundary points and remove outliers to obtain the main contour point set E'. The least squares method is used to fit the geometric primitives (straight line segments, circular arcs, etc.) of the contour boundary to obtain the basic contour description of the component. Then the component parameters are calculated: the direction angle α of the component is obtained through the principal direction analysis, the length l and the width w of the contour are calculated by calculating the minimum circumscribed rectangle, and the diameter d of the component is determined based on the symbol type t. Finally, the component parameter vector C=[α, l, w, d] is generated, which completely describes the geometric features of the steel bar component. This parameter extraction method based on local features can accurately obtain the component information associated with the target symbol, providing accurate data for engineering applications.

[0061] Step 50: Obtain the associated steel bar information of the target steel bar symbol in the adjacent recognition area of the target recognition area, verify the component parameters based on the associated steel bar information, and generate the recognition result of the target recognition area.

[0062] In the embodiments of the present application, the associated steel bar information refers to other steel bar symbols and their attribute information (including symbol type, spatial position, and distribution characteristics) located in the same component as the target steel bar symbol, as well as the complete parameter information of the component (including the geometric dimensions, spatial position, and steel bar arrangement requirements) of these steel bar symbols. These information together constitute the component-level context features of the target steel bar symbol, which are used to verify and improve the accuracy of the recognition result.

[0063] The recognition result refers to the labeling relationship and engineering data obtained through image analysis, including the mapping relationship between the label line and the target steel bar symbol, and the component parameters corresponding to the target steel bar symbol.

[0064] Specifically, a search range of 100 pixels is set in the adjacent recognition area with the positioning coordinates of the target reinforcement symbol as the center, and the associated reinforcement information of the same type of reinforcement symbol is extracted. First, the basic attributes (diameter, grade) and geometric parameters (length, spacing) of the reinforcement in the adjacent area are obtained, and the associated parameter vector R=[d, g, l, s] is established. By comparing the consistency of the component parameters C of the target reinforcement symbol and the associated parameters R, the verification score V=∑wi×|Ci-Ri| / Ri is calculated, where wi is the weight coefficient of each parameter. When the verification score V is less than the threshold value 0.15, it is confirmed that the component parameters are valid; otherwise, the abnormal parameters are corrected based on the associated reinforcement information. Finally, the recognition result is generated, including but not limited to the labeled relationship T, the symbol attribute S, the component parameter C and the verification result V.

[0065] On the basis of the above-mentioned embodiments, as another optional embodiment, the step of verifying the component parameters based on the associated reinforcement information and generating the recognition result of the target recognition area can further include steps 501-504: Step 501: Establish reinforcement connection constraints based on associated reinforcement information, including continuity requirements and component parameter consistency requirements.

[0066] Specifically, the constraint condition model is established by analyzing the associated reinforcement information. First, the reinforcement symbol set S={si} and its attribute vector A={ai} in the same component are extracted, where ai contains information such as reinforcement type and position coordinates. Based on the continuity requirement of the reinforcement, the spatial constraint function f1(si, sj)=|dij-d0| / d0 is established, where dij is the spacing between adjacent reinforcements, and d0 is the standard spacing. For component parameter consistency, the parameter constraint function f2(pi, p0)=|pi-p0| / p0 is established, where pi is the current component parameter, and p0 is the reference parameter value. The constraint condition vector C=[f1, f2] is generated, and the threshold values T1=0.15 and T2=0.2 are set for subsequent parameter verification.

[0067] Step 502: Verify the component parameters in the target recognition area according to the constraint conditions, and mark the parameter items that do not meet the constraint conditions.

[0068] Specifically, the multi-dimensional parameter checking method is used to evaluate the component parameters. For each set of component parameters P = [p1, p2,..., pn] of the target recognition area, the compliance with the constraint conditions is calculated. First, the reinforcement continuity is checked. When the abnormal value of the spacing between adjacent reinforcements exceeds the set threshold T1, the corresponding parameter is marked as an abnormal item E1. Then the component parameter consistency is verified. The parameter deviation rate δ = |pi - μi| / σi is calculated, where μi is the parameter mean and σi is the standard deviation. When δ > 3, the parameter is marked as an abnormal item E2. An abnormal marking matrix M is established, where Mij = 1 indicates that the jth parameter of the ith component needs to be corrected. A multi-level parameter checking mechanism is used to ensure the reliability of the recognition result.

[0069] Step 503: Correct the marked parameter items to generate corrected component parameters.

[0070] Specifically, the marked parameter items are corrected based on the optimal estimation theory. First, a parameter correction model is constructed: for the continuity abnormal item E1, the average spacing dave of adjacent reinforcements is used to calculate the correction value: p' = p x (1 ± |dave - d| / dave). For the consistency abnormal item E2, the weighted average method is used for correction: p'' = Σwipi / Σwi, where wi = exp(-δi² / 2) is the credibility weight. When multiple abnormal items exist simultaneously, an iterative optimization strategy is used: one parameter is corrected at a time, the constraint conditions are updated, and this process is repeated until all parameters meet the constraint requirements or the maximum number of iterations is reached. Finally, the corrected component parameter set P' is generated to ensure the engineering reasonableness of the parameters.

[0071] Step 504: Generate the recognition result of the target recognition area, which includes the correspondence between the annotation lines and the target reinforcement symbol, and the corrected component parameters corresponding to the target reinforcement symbol.

[0072] Specifically, first, the annotation relationship data structure R = {L, S, P'} is established, where L is the annotation line set, including the direction vector v and the endpoint coordinates (x, y); S is the target reinforcement symbol set, including the symbol type t and the position coordinates (cx, cy); P' is the corrected component parameter set. For each annotation relationship, a mapping function map(Li) → {Si, P'i} is generated, which represents the reinforcement symbol Si pointed to by the annotation line Li and its corresponding component parameter P'i. At the same time, the credibility score Q = exp(-Σ|δi| / n) of the recognition result is calculated, where δi is the correction amount and n is the number of parameters. Finally, the complete recognition result document is output, including the annotation correspondence, component parameters and credibility evaluation, providing accurate data support for engineering applications.

[0073] On the basis of the above embodiment, as another optional embodiment, a steel reinforcement engineering drawing intelligent recognition method further includes the following processes: Specifically, ambiguous scene data is first collected and organized. For each ambiguous scene in the target recognition area, its feature vector F = [f1, f2, f3] is extracted, where f1 represents the spatial distribution characteristics of the marking line (direction, length, density), f2 represents the position characteristics of the reinforcement symbol (coordinates, type, distribution), and f3 represents the contour characteristics of the component (boundary, shape, size). The corresponding recognition result R is recorded, including the correct marking correspondence and component parameters. In this way, a sample set D = { (Fi, Ri)} is accumulated and constructed.

[0074] The sample set D is divided into a training sample set and a validation sample set in a ratio of 8:2 using stratified sampling. To ensure consistency in data distribution, the distribution of ambiguous types is considered during division, so that ambiguous scenes of different types maintain similar proportions in the two data sets. The training sample set is used for model training, and the validation sample set is used to evaluate model performance. A deep learning model is constructed based on the training sample set, using a multi-branch neural network structure, including a feature extraction layer, a relationship reasoning layer, and an output layer. The feature extraction layer uses a convolutional neural network to extract visual features of the image, the relationship reasoning layer uses a graph attention network to process the association between the marking line and the symbol, and the output layer generates the marking correspondence probability and the component parameters. The model is trained in an end-to-end manner, and the loss function includes the correspondence cross-entropy loss Lc and the parameter regression loss Lr: L = aLc + bLr, where a = 0.6 and b = 0.4 are balance coefficients.

[0075] The initial recognition model obtained by training is evaluated and optimized using the validation sample set. The performance indicators of the model on the validation set are calculated: the correspondence accuracy Pacc, the parameter relative error Perr, and the comprehensive score S = yPacc + (1-y)exp(-Perr), where y = 0.7 is the weight coefficient. Based on the evaluation results, the model is iteratively optimized by adjusting the network structure (such as adding attention mechanisms), optimizing the training strategy (such as learning rate scheduling), and introducing regularization constraints, until the performance indicators meet the requirements, and the target recognition model M is obtained.

[0076] Finally, the electronic image of the ambiguous scene to be recognized is input into the optimized target recognition model M. The model first extracts the multi-scale features of the image, then analyzes the potential correspondence between the marking line and the reinforcement symbol through the relationship reasoning network, and finally outputs the recognition result R', including the marking correspondence map (Li) → Si and the confidence score. This deep learning-based method can automatically learn complex marking rules, effectively handle various ambiguous scenes, and significantly improve the accuracy and robustness of recognition. By continuously collecting new sample data for model updating, the adaptability of the system is further improved.

[0077] See Figure 2A module schematic diagram of an intelligent recognition system for steel bar engineering drawings provided by an embodiment of the application, wherein the system comprises: A recognition area division module, configured to acquire an electronic image corresponding to a steel bar engineering drawing, and divide the electronic image into a plurality of recognition areas; An ambiguous area recognition module, configured to detect a marking relationship between a marking line and a steel bar component in each of the recognition areas, and determine a target recognition area with an ambiguous scene based on the marking relationship, the ambiguous scene including at least one of marking pointing confusion, low steel bar symbol recognition degree, and component overlap; A feature parameter extraction module, configured to construct a spatial feature grid corresponding to the target recognition area, and acquire spatial distribution features of the marking line, steel bar symbol position features, and component contour features in the target recognition area based on the spatial feature grid; A corresponding relationship determination module, configured to determine a target steel bar symbol pointed to by the marking line according to the spatial distribution features of the marking line and the steel bar symbol position features, and extract component parameters corresponding to the target steel bar symbol based on the component contour features; An identification result generation module, configured to acquire associated steel bar information of the target steel bar symbol in adjacent recognition areas of the target recognition area, verify the component parameters based on the associated steel bar information, and generate an identification result of the target recognition area.

[0078] Optionally, the recognition area division module is further configured to acquire a distribution density of the marking line in the electronic image, and generate a density heat map based on the distribution density; Determine a gathering area of the marking line according to the density heat map; Take the gathering area as a reference point for area division, and divide the plurality of recognition areas according to a preset size range with the reference point as the center.

[0079] Optionally, the ambiguous area recognition module is further configured to acquire a graphic feature parameter of each recognition area, wherein the graphic feature parameter includes a marking line feature, a steel bar symbol feature, and a component feature; Calculate a feature score of each evaluation index according to a preset feature evaluation index system and an image feature parameter, wherein the evaluation index includes a marking pointing confusion index, a symbol recognition degree index, and a component overlap index; Compare each of the feature scores with a preset evaluation criterion, and take a recognition area with any feature score not satisfying the corresponding evaluation criterion as a target recognition area.

[0080] Optionally, the feature parameter extraction module is further configured to calculate a feature density value of each grid unit in the target recognition area, the feature density value including a marking line density, a steel bar symbol density, and a component density; determine a key feature region of the target recognition region based on the feature density value, the key feature region including a marking line aggregation region, a steel bar symbol distribution region, and a component dense region; extract a marking line direction and length parameter of the marking line aggregation region, and generate a marking line spatial distribution feature of the target recognition region; determine a symbol coordinate and region range of the steel bar symbol distribution region, and generate a steel bar symbol position feature of the target recognition region; extract an edge contour point set of the component dense region, and generate a component contour feature of the target recognition region.

[0081] Optionally, the correspondence determining module is further configured to extract an extension direction and end position of the marking line according to the spatial distribution feature of the marking line, and establish a feature parameter pointed by each marking line, the feature parameter including an extension distance, a direction angle, and an end coordinate; calculate a matching degree score of the marking line and each steel bar symbol according to the steel bar symbol position feature and the feature vector, the matching degree score being determined by a distance decay coefficient and an angle deviation coefficient; select a steel bar symbol with the highest matching degree score as a target steel bar symbol, and obtain a positioning coordinate of the target steel bar symbol; determine a contour boundary point set associated with the target steel bar symbol in the component contour feature according to the positioning coordinate, and extract a component parameter based on the contour boundary point set.

[0082] Optionally, the recognition result generating module is further configured to establish a steel bar connection constraint condition based on the associated steel bar information, the constraint condition including a continuity requirement of the steel bar and a component parameter consistency requirement; verify the component parameter in the target recognition region according to the constraint condition, and mark a parameter item that does not meet the constraint condition; correct the marked parameter item to generate a corrected component parameter; generate a recognition result of the target recognition region, the recognition result including a correspondence between the marking line and the target steel bar symbol pointed by the marking line, and the corrected component parameter corresponding to the target steel bar symbol.

[0083] Optionally, the recognition result generating module is further configured to generate a sample set of an ambiguity scene and a corresponding recognition result of a plurality of target recognition regions, divide the sample set into a training sample set and a verification sample set; train an initial recognition model based on the training sample set, and verify and optimize the initial recognition model based on the verification sample set to obtain a target recognition model; input an electronic image with an ambiguity scene to be recognized into the target recognition model, and output a corresponding correspondence between the marking line and the steel bar symbol.

[0084] It should be noted that the system provided by the above embodiment is only exemplified by the above division of functional modules when realizing its functions. In actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above described functions. In addition, the system and method embodiments provided by the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be described here.

[0085] The embodiment of the present application further provides a computer storage medium, which can store a plurality of instructions. The instructions are suitable for being loaded by a processor and executing the intelligent identification method of a steel bar engineering drawing provided by the above embodiment. The specific execution process can be referred to the specific description of the above embodiment, which will not be described here.

[0086] Please refer to Figure 3 The present application also discloses an electronic device. Figure 3 is a structural schematic diagram of an electronic device disclosed by the embodiment of the present application. The electronic device 300 can include at least one processor 301, at least one network interface 304, a user interface 303, a memory 305, and at least one communication bus 302.

[0087] The communication bus 302 is used to realize the connection and communication between the components.

[0088] The user interface 303 can include a display screen (Display) and a camera (Camera). Optionally, the user interface 303 can further include a standard wired interface and a wireless interface.

[0089] The network interface 304 can optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0090] The processor 301 can include one or more processing cores. The processor 301 connects various parts within the server through various interfaces and lines, performs various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 305, and calling data stored in the memory 305. Alternatively, the processor 301 can be implemented in at least one of a hardware form of a digital signal processing (DSP), a field-programmable gate array (FPGA), and a programmable logic array (PLA). The processor 301 can integrate a combination of one or more of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. Among them, the CPU mainly processes operating systems, user interfaces, and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; and the modem is used for processing wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor 301, but can be realized by a separate chip.

[0091] The memory 305 can include a random access memory (RAM) and a read-only memory (ROM). Alternatively, the memory 305 includes a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 305 can include a program storage area and a data storage area, wherein the program storage area can store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area can store data involved in the above-mentioned various method embodiments, etc. The memory 305 can alternatively be at least one storage device located away from the aforementioned processor 301. Referring to Figure 3 The memory 305 as a kind of computer storage medium can include an operating system, a network communication module, a user interface module and an application program of a kind of steel bar engineering drawing intelligent identification method.

[0092] In Figure 3In the electronic device 300 shown, the user interface 303 is mainly used to provide an interface for the user to input, and obtain data input by the user; and the processor 301 can be used to invoke an application program stored in the memory 305 and storing a steel bar engineering drawing intelligent recognition method, which, when executed by one or more processors 301, causes the electronic device 300 to perform the method of one or more of the above-described embodiments. It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all described as a combination of a series of actions, but those skilled in the art should know that the present application is not limited to the order of the actions described, because according to the present application, certain steps can be performed in other order or at the same time. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0093] In the above embodiments, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0094] In several embodiments provided by the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only schematic. The division of the units is only a logical function division. There can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between the units can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical or other forms.

[0095] The units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0096] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present alone, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0097] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable memory. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a memory and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned memory includes: a U disk, a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0098] The above are only exemplary embodiments of the present disclosure, and cannot limit the scope of the present disclosure. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure are still within the scope of the present disclosure. Other embodiments of the present disclosure will be readily apparent to those skilled in the art upon considering the specification and practicing the true principles of the present disclosure.

[0099] The present application is intended to cover any variations, uses or adaptive changes of the present disclosure that follow the general principles of the present disclosure and include common knowledge or conventional technical means in the technical field not described in the present disclosure. The specification and examples are only considered as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for intelligent recognition of a reinforcement engineering drawing, characterized in that, The method comprises: obtaining an electronic image corresponding to a steel bar engineering drawing, and dividing the electronic image into a plurality of recognition regions; detecting a marking relationship between a marking line and a steel bar component in each of the recognition regions, and determining a target recognition region with an ambiguous scene based on the marking relationship, the ambiguous scene including at least one of marking direction confusion, low steel bar symbol recognition, and component overlap; constructing a spatial feature grid corresponding to the target recognition region, and obtaining spatial distribution features of the marking line, steel bar symbol position features, and component contour features in the target recognition region based on the spatial feature grid; determining a target steel bar symbol pointed to by the marking line according to the spatial distribution features of the marking line and the steel bar symbol position features, and extracting component parameters corresponding to the target steel bar symbol based on the component contour features; obtaining associated steel bar information of the target steel bar symbol in adjacent recognition regions of the target recognition region, verifying the component parameters based on the associated steel bar information, and generating a recognition result of the target recognition region.

2. The method of claim 1, wherein, The dividing of the electronic image into a plurality of recognition regions comprises: obtaining a distribution density of the marking line in the electronic image, and generating a density heat map based on the distribution density; determining a gathering region of the marking line according to the density heat map; taking the gathering region as a reference point for region division, and dividing the plurality of recognition regions according to a preset size range with the reference point as the center.

3. The intelligent recognition method of a reinforcement engineering drawing according to claim 1, characterized in that, The determining of the target recognition region with the ambiguous scene based on the marking relationship comprises: obtaining a graphic feature parameter of each recognition region, wherein the graphic feature parameter includes a marking line feature, a steel bar symbol feature, and a component feature; calculating a feature score of each evaluation index according to a preset feature evaluation index system and an image feature parameter, wherein the evaluation index includes a marking direction confusion index, a symbol recognition index, and a component overlap index; comparing each feature score with a preset evaluation criterion, and taking a recognition region with any feature score not satisfying the corresponding evaluation criterion as the target recognition region.

4. The intelligent recognition method of a reinforcement engineering drawing according to claim 1, characterized in that, The spatial feature grid comprises a plurality of grid units, and the obtaining of the spatial distribution features of the marking line, the steel bar symbol position features, and the component contour features in the target recognition region based on the spatial feature grid comprises: calculating a feature density value of each grid unit in the target recognition region, the feature density value including a marking line density, a steel bar symbol density, and a component density; determining a key feature region of the target recognition region based on the feature density value, the key feature region including a marking line gathering region, a steel bar symbol distribution region, and a component dense region; extracting a marking line direction and length parameter of the marking line gathering region, and generating a marking line spatial distribution feature of the target recognition region; determining a symbol coordinate and a region range of the steel bar symbol distribution region, and generating a steel bar symbol position feature of the target recognition region; extracting an edge contour point set of the component dense region, and generating a component contour feature of the target recognition region.

5. The intelligent recognition method of a reinforcement engineering drawing according to claim 1, characterized in that, The method comprises the following steps: According to the spatial distribution characteristics of the annotation line and the reinforcement symbol position characteristics, the target reinforcement symbol pointed by the annotation line is determined, and the component parameters corresponding to the target reinforcement symbol are extracted based on the component contour characteristics, which comprises: According to the spatial distribution characteristics of the annotation line, the extension direction and the end position of the annotation line are extracted, and the characteristic parameters pointed by each annotation line are established, which comprises the extension distance, the direction angle and the end coordinates; According to the reinforcement symbol position characteristics and the characteristic vector, the matching degree score of the annotation line and each reinforcement symbol is calculated, which is determined by the distance attenuation coefficient and the angle deviation coefficient; The reinforcement symbol with the highest matching degree score is selected as the target reinforcement symbol, and the positioning coordinates of the target reinforcement symbol are obtained; 6. The intelligent recognition method of a reinforcement engineering drawing according to claim 1, characterized in that, According to the positioning coordinates, the contour boundary point set associated with the target reinforcement symbol is determined in the component contour characteristics, and the component parameters are extracted based on the contour boundary point set. The associated reinforcement information comprises the reinforcement symbols belonging to the same component as the target reinforcement symbol and the corresponding component parameters, the component parameters are verified based on the associated reinforcement information, and the identification result of the target identification area is generated, which comprises: Based on the associated reinforcement information, the reinforcement connection constraint condition is established, which comprises the continuity requirement of the reinforcement and the consistency requirement of the component parameters; According to the constraint condition, the component parameters in the target identification area are verified, and the parameter items that do not meet the constraint condition are marked; The marked parameter items are modified to generate the modified component parameters; 7. The intelligent recognition method of a reinforcement engineering drawing according to claim 1, characterized in that, The identification result of the target identification area is generated, which comprises the corresponding relationship between the annotation line and the target reinforcement symbol pointed by the annotation line, and the modified component parameters corresponding to the target reinforcement symbol. The method further comprises: The ambiguity scenarios and the corresponding identification results of a plurality of target identification areas are generated into a sample set, and the sample set is divided into a training sample set and a verification sample set; Based on the training sample set, an initial identification model is trained, and based on the verification sample set, the initial identification model is verified and optimized to obtain a target identification model; 8. A system for intelligent recognition of a reinforcement engineering drawing, characterized by, The electronic image with an ambiguity scenario to be identified is input into the target identification model, and the corresponding corresponding relationship between the annotation line and the reinforcement symbol is output. The system comprises: An identification area division module is used for obtaining an electronic image corresponding to a reinforcement engineering drawing, and dividing the electronic image into a plurality of identification areas; An ambiguity area identification module is used for detecting the annotation relationship between the annotation line and the reinforcement component in each identification area, and determining the target identification area with an ambiguity scenario based on the annotation relationship, wherein the ambiguity scenario comprises at least one of annotation pointing confusion, low reinforcement symbol recognition degree and component overlap; A feature parameter extraction module is used for constructing a spatial feature grid corresponding to the target identification area, and obtaining the spatial distribution characteristics of the annotation line, the reinforcement symbol position characteristics and the component contour characteristics in the target identification area based on the spatial feature grid; A correspondence determining module is configured to determine a target reinforcing bar symbol pointed by the marking line according to the spatial distribution feature of the marking line and the reinforcing bar symbol position feature, and extract a component parameter corresponding to the target reinforcing bar symbol based on the component contour feature; An identification result generating module is configured to acquire associated reinforcing bar information of the target reinforcing bar symbol in an adjacent identification area of the target identification area, verify the component parameter based on the associated reinforcing bar information, and generate an identification result of the target identification area.

9. A computer-readable storage medium, characterized in that, A computer readable storage medium stores a plurality of instructions, the instructions being adapted to be loaded and executed by a processor to implement the method of any one of claims 1-7.

10. An electronic device, comprising: An electronic device includes a processor, a memory, a user interface, and a network interface. The memory is configured to store instructions. The user interface and the network interface are configured to communicate with other devices. The processor is configured to execute the instructions stored in the memory to cause the electronic device to implement the method of any one of claims 1-7.

Images