Fan foundation steel bar quantity statistical method and system based on specification constraints

Abstract

This application discloses a method and system for statistically calculating the quantity of steel reinforcement in wind turbine foundations based on specification constraints, relating to the field of intelligent quantity accounting. The method includes: acquiring design and structural data and construction scan point clouds of the wind turbine foundation; registering and denoising the point clouds based on reference control points to obtain a net point cloud; extracting cross-sectional features by layering and slicing along the height and detecting candidate sections; constructing specification constraint factors based on theoretical specification parameters, scoring and filtering candidate sections to obtain effective sections; performing spatial consistency verification and anomaly correction by combining layout rules and design tolerances; statistically calculating the quantity of steel reinforcement of each specification and outputting a statistical report. Therefore, high-precision quantity calculation that can be verified according to specifications can be achieved under complex site conditions such as shading and overlap.

Description

Technical Field

[0001] This application relates to the field of intelligent engineering quantity calculation technology, and in particular to a method and system for statistically analyzing the quantity of steel reinforcement in wind turbine foundations based on specification constraints. Background Technology

[0002] In the field of wind power engineering construction, wind turbine foundations, as key structures bearing enormous loads, typically employ large-diameter concrete frustums or cylindrical designs. Their internal reinforcement arrangement is characterized by high density, multiple specifications intersecting, and layered distribution within a large volume of concrete. Accurate quantity surveying, especially the calculation of reinforcement quantities, directly impacts project cost control, structural safety assessment, and acceptance of construction compliance. However, due to the unique and complex structure of wind turbine foundations, traditional quantity surveying methods are insufficient to meet the demands of modern, precise construction.

[0003] Currently, methods for calculating rebar quantity are mainly divided into manual calculation, two-dimensional graphical methods, and automatic quantity calculation based on general BIM software. Manual and two-dimensional methods rely on mechanically reviewing design drawings and empirical estimations. When dealing with irregular, three-dimensional, irregularly shaped rebar structures like wind turbine foundations, these methods are prone to omissions or errors, and are inefficient. Although some projects have introduced BIM technology, existing BIM quantity calculations often rely on standardized component families, which are difficult to flexibly adapt to the complex rebar bends, intersections, and variable cross-section arrangements in wind turbine foundations. This results in the generated schedules often failing to accurately reflect the actual spatial parameters of each rebar.

[0004] Furthermore, with the development of machine vision technology, some image recognition-based rebar counting solutions have emerged in the industry. However, these solutions are mostly limited to rebar processing yards or material storage areas, and are designed for strip-shaped, discrete rebars with clearly visible cross-sections. At wind turbine foundation construction sites, rebars are usually tied into cages, partially obscured by formwork or covered by concrete, resulting in extremely limited visible cross-sections with severe occlusion and overlap. Current visual analysis systems struggle to achieve high-precision quantity counting and compliance verification in complex occlusion environments. Summary of the Invention

[0005] This application provides a method, system, storage medium, computer program product, and electronic device for statistically analyzing the quantity of steel reinforcement in wind turbine foundations based on specification constraints. This is intended to at least solve the problem in the prior art where it is difficult to achieve high-precision and verifiable automated statistical analysis of steel reinforcement quantities in wind turbine foundations under complex three-dimensional structures and on-site obstruction and interference conditions.

[0006] In a first aspect, embodiments of this application provide a method for statistically analyzing the quantity of reinforcing steel bars in a wind turbine foundation based on specification constraints. The method includes: acquiring design and structural data of the wind turbine foundation and scanned point cloud data from the construction site; the design and structural data includes theoretical specification parameters and arrangement rules for the reinforcing steel bars, and the scanned point cloud data reflects the three-dimensional spatial morphology of the reinforcing steel cage; based on reference control points in the design and structural data, performing coordinate system registration and noise reduction preprocessing on the scanned point cloud data to generate net point cloud data consistent with the coordinates of the design and structural data; and performing layered slicing processing on the net point cloud data along the height direction to generate cross-sections of the reinforcing steel bars at different height levels. The system collects distributed slice feature data and performs target detection on each slice feature data to obtain an initial cross-section candidate set. It then constructs a specification constraint factor using the theoretical specification parameters and evaluates and filters the initial cross-section candidate set based on the specification constraint factor to identify valid rebar cross-sections that meet design specifications. The specification constraint factor characterizes the degree of agreement between the candidate targets and the theoretical specification parameters. Based on the arrangement rules and design tolerance range, it performs spatial consistency verification and anomaly correction on the valid rebar cross-sections. Based on the corrected rebar cross-section detection results, it counts the number of rebars of each specification and outputs a rebar quantity statistics report.

[0007] Secondly, embodiments of this application provide a system for statistically analyzing the quantity of reinforcing steel bars in wind turbine foundations based on specification constraints. The system includes: a data acquisition unit for acquiring design and structural data of the wind turbine foundation and scanned point cloud data from the construction site; the design and structural data includes theoretical specification parameters and arrangement rules for the reinforcing steel bars, and the scanned point cloud data reflects the three-dimensional spatial morphology of the reinforcing steel cage; a data registration and preprocessing unit for performing coordinate system registration and noise reduction preprocessing on the scanned point cloud data based on reference control points in the design and structural data, generating net point cloud data consistent with the coordinates of the design and structural data; and a slicing and target detection unit for performing layered slicing processing on the net point cloud data along the height direction, generating data reflecting different heights. The system collects slice feature data of the cross-sectional distribution of the reinforcing bars, and performs target detection on each slice feature data to obtain an initial candidate set of cross-sections; a specification constraint and screening unit is used to construct specification constraint factors using the theoretical specification parameters, and to evaluate and screen the initial candidate set of cross-sections based on the specification constraint factors to identify effective reinforcing bar cross-sections that meet the design specifications; the specification constraint factors are used to characterize the degree of conformity between the candidate targets and the theoretical specification parameters; a verification and statistical reporting unit is used to perform spatial consistency verification and anomaly correction on the effective reinforcing bar cross-sections based on the arrangement rules and design tolerance range, count the number of reinforcing bars of each specification based on the corrected reinforcing bar cross-section detection results, and output a reinforcing bar quantity statistical report.

[0008] Thirdly, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the specification-constrained wind turbine foundation reinforcement quantity counting method of any embodiment of this application.

[0009] Fourthly, embodiments of this application provide a storage medium storing a computer program thereon, characterized in that, when the program is executed by a processor, it implements the steps of the method for calculating the quantity of steel reinforcement in wind turbine foundations based on specification constraints, according to any embodiment of this application.

[0010] Fifthly, embodiments of this application provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the method for calculating the quantity of steel reinforcement in a wind turbine foundation based on specification constraints, according to any embodiment of this application.

[0011] The method and system for calculating the quantity of steel reinforcement in wind turbine foundations based on specification constraints provided in this application can achieve at least the following technical effects:

[0012] (1) The construction site scan point cloud and the design structure data are established with the same coordinates and the same elevation. On this unified benchmark, the three-dimensional complex shape of the steel cage is transformed into a layered slice feature expression along the height direction. As a result, the steel reinforcement quantity statistics no longer rely on local perspectives or scattered sections selected manually. Instead, the cross-sectional distribution of each height layer is uniformly characterized based on a repeatable spatial benchmark, so that the subsequent identification and summarization have a stable statistical caliber. At the same time, the layered slice decouples the three-dimensional overlapping relationship into a multi-layer cross-sectional distribution, so that the input of target identification is transformed from a complex spatial occlusion scene into a relatively regular cross-sectional detection problem, thereby improving the reproducibility and traceability of the statistical process under different construction conditions and different scanning densities.

[0013] (2) Introducing specification constraint factors constructed from theoretical specification parameters to evaluate and screen cross-section candidates with confidence, and further combining layout rules and design tolerances to perform spatial consistency verification and anomaly correction, thereby realizing a structured closed loop of "identification-verification-correction". Specifically, specification constraint factors require that the test results meet the requirements of conformity with the design specifications, and can suppress pseudo-section interference caused by noise clusters, cross component edges, etc. before entering the counting stage, avoiding the accumulation and amplification of false detections in the quantity summarization stage; while layout rules and tolerance range provide secondary constraints oriented towards structural laws, making the consistency of the spatial distribution of steel bars of the same specification a verifiable condition, thereby performing rule-driven correction and calibration of phenomena such as local missed detection and fracture detection, and finally obtaining quantity results statistically based on specification dimensions with consistency endorsement, and forming a statistical report output that can be directly used for engineering accounting and review.

[0014] This technical solution elevates the counting of rebar quantity from a simple visual / inspection count to a verifiable and correctable engineering quantity calculation process by strongly constraining the point cloud cross-section detection results through design specifications and layout rules. This ensures that the output quantity of rebar of each specification not only comes from the detection results themselves, but also meets the requirements of design parameter conformity and spatial distribution consistency, thereby significantly improving the credibility and verifiability of the quantity statistics results in engineering scenarios. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart is shown as an example of a method for counting the number of steel bars in a wind turbine foundation based on specification constraints, according to an embodiment of this application.

[0017] Figure 2 This document shows an example of an operation flowchart illustrating the generation of net point cloud data through preprocessing in a method according to an embodiment of this application.

[0018] Figure 3 This document illustrates an example of an operation flowchart for obtaining an initial cross-section candidate set through layered slicing and target detection in a method according to an embodiment of this application.

[0019] Figure 4 A flowchart illustrating an example of identifying effective steel reinforcement sections by constructing specification constraint factors according to an embodiment of this application is shown.

[0020] Figure 5This document shows an example of an operation flowchart for spatial consistency verification and anomaly correction of an effective steel reinforcement section according to an embodiment of this application;

[0021] Figure 6 A schematic diagram illustrating the principle mechanism of an example of a method for counting the number of steel bars in a wind turbine foundation based on specification constraints, according to an embodiment of this application;

[0022] Figure 7 A schematic diagram showing the comparative experimental results of counting accuracy and stability under various occlusion and noise environments based on different methods is presented.

[0023] Figure 8 This diagram illustrates the comparison of the mean absolute error variation of different methods under various occlusion rates.

[0024] Figure 9 A structural block diagram of an example of a specification-constrained wind turbine foundation reinforcement quantity counting system according to an embodiment of this application is shown. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] It should be noted that currently, in related technologies, the academic and engineering communities have attempted to solve the problem of steel bar quantity statistics from two dimensions: high-level BIM applications and intelligent vision algorithms, respectively. However, there are still significant limitations for the specific scenario of wind turbine foundations.

[0027] In terms of deepening the application of BIM technology, some scholars have proposed an automated structural design and cost optimization model based on a visual programming platform. This model designs components such as floor slabs and beams through custom nodes and converts the calculated cross-sectional area of ​​the reinforcing steel into weight to estimate material costs. Although this model achieves automatic calculation of the usage of some regular structures, its application is mainly aimed at conventional building structures such as slabs and beams, focusing on macro cost optimization rather than precise quantity statistics. It is difficult to directly promote this model when facing complex geometric configurations such as wind turbine truncated cone foundations. In addition, related research techniques have also proposed methods to achieve parametric modeling and quantity calculation through interaction between external data tables and 3D models. Although this solves the problem of modifying some component family parameters, the generated detailed tables often rely on manual parameter maintenance and are difficult to cover the variable reinforcement patterns in wind turbine foundations. Furthermore, it lacks an automatic verification mechanism for reinforcement specification constraints (such as diameter and spacing consistency).

[0028] In the fields of machine vision and deep learning, some studies utilize improved instance segmentation algorithms or region convolutional neural network models to solve the problem of counting dense rebars, improving counting accuracy in rebar yard scenarios through image segmentation and large-size image processing strategies. Other studies attempt to identify rebar ends using UAV imagery combined with target detection algorithms. However, these visual solutions primarily target ideal scenarios (such as processing yards or material storage areas) where the rebar cross-section is fully visible and lighting conditions are good. At wind turbine foundation construction sites, rebar cages are typically already tied and often obscured by concrete casings, formwork, or fixed supports, preventing cameras from capturing complete cross-sectional features. More critically, current visual analysis algorithms usually only focus on the recognition and counting of geometric features, lacking a mechanism for deeply integrating on-site perception data with design specifications (such as rebar diameter, theoretical spacing, and number of layers). This makes it impossible to determine whether the identified rebar conforms to design specifications and to automatically detect construction deviations.

[0029] It should be understood that the above description of the relevant technologies is intended only to help the public better understand the inventive spirit and motivation of this application, and is not intended to limit this application. Furthermore, the technical solutions described in the above-mentioned relevant technologies are not prior art, and may also be undisclosed technical solutions, such as those under research or in the laboratory stage.

[0030] The technical solutions in this application, including the collection, storage, use, processing, transmission, provision, and disclosure of users' personal information, comply with relevant laws and regulations and do not violate public order and good morals.

[0031] Figure 1 A flowchart illustrating an example of a method for counting the number of steel reinforcement bars in a wind turbine foundation based on specification constraints, according to an embodiment of this application, is shown.

[0032] Regarding the execution subject of the method in the embodiments of this application, it can be any controller or processor with computing or processing capabilities, such as the digital quantity calculation and quality verification system platform for wind turbine engineering. By using "design specifications and layout rules" as strong constraints, the uncertain identification process in the on-site point cloud is transformed into a measurable, filterable, and verifiable statistical closed loop. This enables the steel reinforcement quantity statistics to not only output quantity results, but also to complete consistency verification and anomaly correction under the same rule system. Therefore, it can obtain stable and engineering-acceptable statistical conclusions under complex spatial morphology and on-site interference conditions.

[0033] In some examples, it may be integrated into an electronic device or terminal through software, hardware, or a combination of both, and the type of terminal or electronic device may be diverse.

[0034] like Figure 1As shown, in step S110, the design and construction data of the wind turbine foundation and the scanning point cloud data of the construction site are obtained. The design and construction data includes the theoretical specifications and arrangement rules of the steel bars, and the scanning point cloud data reflects the three-dimensional spatial shape of the steel cage.

[0035] Specifically, the design and construction data can originate from BIM model files or digital construction drawings provided by the design unit. The system parses and extracts the theoretical specifications (such as design diameter and length) and arrangement rules (such as theoretical spacing in the circumferential or radial direction, layer distribution, etc.) of various types of steel bars in the wind turbine foundation, which serve as the "standard true values" for subsequent verification. Simultaneously, a 3D laser scanner or a vision sensor with depth perception capabilities is used to perform a comprehensive scan of the already tied steel cages at the construction site, acquiring massive amounts of point cloud data reflecting the actual three-dimensional spatial morphology of the steel cages. Thus, two sets of basic datasets—"theoretical standards" and "on-site conditions"—are constructed, providing a data foundation for subsequent high-precision statistics based on a combination of virtual and real data.

[0036] In step S120, based on the reference control points in the design and construction data, coordinate system registration and noise reduction preprocessing are performed on the scanned point cloud data to generate net point cloud data with coordinates consistent with the design and construction data.

[0037] It should be noted that since the point cloud data from on-site scanning is typically located in the scanner's local coordinate system, which does not coincide with the coordinate system of the design data and contains a large amount of environmental noise, the steps in this embodiment first identify the reference control points (such as the base center, edge corners, etc.) that coexist in the design model and the on-site data. The spatial transformation relationship between the two is then calculated to accurately map and unify the chaotic on-site point cloud to the design coordinate system, achieving "virtual-real alignment." Subsequently, a denoising preprocessing operation is performed on the outliers in the original point cloud (such as noise caused by dust, flying insects, etc.) to filter out background interference. This not only unifies the spatial reference for subsequent processing but also significantly reduces the computational load and false detection rate of subsequent algorithms by purifying the data.

[0038] In step S130, the net point cloud data is sliced ​​along the height direction to generate slice feature data that reflects the cross-sectional distribution of steel bars at different heights. Target detection is then performed on each slice feature data to obtain an initial cross-section candidate set.

[0039] It should be noted that, considering the extremely complex three-dimensional structure of the wind turbine foundation reinforcement cage, direct single-unit segmentation in three-dimensional space is extremely difficult. Therefore, this embodiment employs a dimensionality reduction processing logic. The slicing direction is determined based on the orientation of the reinforcement to be counted (e.g., horizontal slices along the height axis are used for longitudinal reinforcement, and vertical slices along the radial direction are used for horizontal circumferential reinforcement; this embodiment uses longitudinal reinforcement as an example). The three-dimensional net point cloud data is divided into a series of continuous slices. For each slice, it is converted into slice feature data (such as a projection density map or feature image) that reflects the cross-sectional shape distribution of the reinforcement at that height level, thus transforming the complex three-dimensional entanglement problem into a target detection problem across multiple two-dimensional planes. Next, image processing or geometric feature extraction algorithms are used to search for regions exhibiting reinforcement cross-sectional features (such as circular or near-circular) within these slice feature data, extracting an initial cross-sectional candidate set containing all potential reinforcement targets. This effectively simplifies the data processing dimension, enabling rapid location of all possible reinforcement positions and ensuring that no visible targets are missed.

[0040] In step S140, a specification constraint factor is constructed using theoretical specification parameters, and the confidence level of the initial cross-section candidate set is evaluated and screened based on the specification constraint factor to identify the effective steel reinforcement cross-section that meets the design specifications. The specification constraint factor is used to characterize the degree of agreement between the candidate target and the theoretical specification parameters.

[0041] It should be noted that the initial candidate set often contains many artifacts formed by construction debris, concrete spots, etc., which are difficult to distinguish based on geometric shape alone.

[0042] In the embodiments of this application, prior design knowledge is introduced. A quantified specification constraint factor is constructed using theoretical specification parameters (such as design radius and design spacing) extracted from previous steps. This factor serves as an evaluation index to measure the degree to which each candidate target conforms to the design specifications in terms of size and neighborhood relationship. The system scores the confidence level of each target in the initial candidate set based on this factor, retaining those targets that highly conform to the design specifications and eliminating false targets with large size deviations or unreasonable spatial distribution. By adopting this mechanism, not only is the recognition accuracy significantly improved, but the pain point of traditional visual methods being unable to distinguish steel reinforcement specifications is also effectively solved, achieving intelligent recognition based on the design.

[0043] In step S150, based on the layout rules and design tolerance range, the spatial consistency of the effective steel bar cross section is checked and anomalies are corrected. The quantity of steel bars of each specification is counted based on the corrected steel bar cross section test results, and a steel bar quantity statistics report is output.

[0044] Here, after obtaining the effective rebar cross-section, the system further performs a global perspective, combining the rebar layout rules (such as circumferential distribution patterns and inter-layer continuity) and the design-allowed construction tolerance range to perform an overall spatial consistency check on the detection results. If gaps violating the layout logic are found in local areas (such as missed detections due to occlusion) or abnormal abrupt changes, the system will perform logical inference and correction based on contextual information to fill in missing targets or correct deviations. Finally, the system summarizes the final detection results after multiple checks and corrections, classifies and statistically analyzes the actual quantity, location, and deviation of rebar of different specifications, and automatically generates a visualized rebar quantity statistics report. This ensures the integrity and compliance of the final delivered data, providing an objective and reliable basis for project acceptance and settlement.

[0045] Figure 2 A flowchart illustrating an example of generating net point cloud data through preprocessing in a method according to an embodiment of this application is shown.

[0046] like Figure 2 As shown, in step S210, a spatial transformation model between the design coordinate system and the field scanning coordinate system is constructed, and a coarse-to-fine hierarchical registration strategy is executed.

[0047] Specifically, a rigid transformation model is constructed between the design coordinate system and the on-site scanning coordinate system. This is used to describe the positional relationship between the field scan data points and the design model, and to perform a coarse-to-fine two-level coordinate registration, where... This represents any point in the on-site scan point cloud. This represents the rotation matrix, used to describe the angular difference between two coordinate systems. This represents the displacement vector, used to describe the spatial offset between the coordinate origin and the origin. This represents the coordinates of the transformed point.

[0048] In some implementations, the basic geometric center and edge contour features in the scanned point cloud data are first extracted, and then aligned with the reference control points in the design and construction data to complete coarse registration.

[0049] Here, to avoid getting stuck in local optima during fine registration, this step first performs coarse registration. For example, the base plane is extracted using RANSAC plane fitting, and the basic geometric center (e.g., the center of the frustum base) and edge contour features are extracted from the scanned point cloud data. These features are then aligned with the reference control points in the design and construction data, and the initial rotation matrix is ​​solved. and initial displacement vector This completes the initial coarse registration, enabling the approximate orientation of the field data to be quickly aligned with the design model, providing a good initial orientation for subsequent high-precision calculations.

[0050] Subsequently, an iterative optimization algorithm is used to minimize the spatial distance error between the point cloud and the design model, and the optimal spatial transformation parameters are solved to accurately unify the scanned point cloud data to the design coordinate system.

[0051] Specifically, using the initial coarse registration result as the initial value, the Iterative ClosestPoint (ICP) algorithm is used to construct an objective function that minimizes the Euclidean distance. This is then used to perform refined iterations on the scanned point cloud data to solve for the optimal rotation matrix. and optimal displacement vector This unifies the scanned point cloud data to the design coordinate system.

[0052] For example, at this stage, the results obtained from coarse registration and As initial values, the Iterative Closest Point (ICP) algorithm is used to construct an objective function that minimizes the Euclidean distance. Through iterative calculation, the rotation matrix is ​​continuously fine-tuned. and displacement vector This continues until the spatial distance error between the scanned point cloud and the surface of the design model converges to a preset minimum value. The optimal rotation matrix is ​​then obtained through this process. and optimal displacement vector The system transforms massive amounts of on-site scan point cloud data into the design coordinate system, thereby eliminating coordinate deviations caused by the arbitrary placement of scanners on the construction site and ensuring the consistency of the spatial reference for subsequent steel bar comparison based on the design location.

[0053] In step S220, hybrid denoising processing is performed on the registered point cloud data to obtain net point cloud data.

[0054] It should be noted that due to the complex environment of construction sites, the original point cloud often contains dust noise from the air and background objects that are not steel reinforcement. In this embodiment, a density-based statistical outlier removal algorithm is used to remove discrete noise from the scanned point cloud data. Specifically, the average distance from each point in the point cloud to its neighboring points is calculated, and points whose average distance is significantly greater than the global statistical standard deviation are considered discrete noise and removed, thereby purifying the background noise.

[0055] Furthermore, based on the analysis of local geometric features of the point cloud, the normal vector and local curvature of each point in the remaining point cloud are calculated. Planar point cloud clusters with consistent normal vector directions and local curvatures below the smoothness threshold are identified and determined to be interference from the surface of the concrete casing or construction template, and are removed. Linear point cloud clusters with high curvature features are retained as net point cloud data.

[0056] Here, to separate the reinforcing bars from the concrete casing and formwork, the system further analyzes the local geometric features of the point cloud to calculate the normal vector and local curvature of each point in the remaining point cloud. Specifically, the system uses geometric differences for logical judgment: it identifies point cloud clusters with highly consistent normal vector directions and local curvatures below a preset smoothness threshold, determines them as flat concrete casing surfaces or construction formwork interference, and removes them; it retains linear point cloud clusters with high curvature characteristics (i.e., drastic changes in normal vectors). Thus, it effectively removes the interfering data surrounding the reinforcing bars, generating net point cloud data with a high signal-to-noise ratio.

[0057] In some implementations, in order to accurately separate the "linear feature" of reinforcing bars from "planar interference" such as concrete casings and formwork, the system can use principal component analysis (PCA) based on local neighborhoods to construct geometric morphology discrimination logic.

[0058] Specifically, for any point in the remaining point cloud Search for it Construct a local covariance matrix from the set of points in the neighborhood. :

[0059] Equation (1)

[0060] In the formula, As the neighborhood center, This is a vector transpose operation. This indicates the number of point clouds in a local neighborhood (or neighborhood scale). For points within the neighborhood. By analyzing the covariance matrix... Eigenvalue decomposition is performed to obtain the three non-negative eigenvalues ​​of the local covariance matrix. And their corresponding feature vectors. The system constructs a linearity index based on these three feature values. and flatness index :

[0061] Equation (2)

[0062] In the formula, The largest eigenvalue represents the principal direction where the point cloud distribution is most discrete (with the largest variance); These are intermediate eigenvalues; The minimum eigenvalue represents the direction in which the point cloud distribution is most compact (with the smallest variance) (usually corresponding to the normal direction of the plane). Therefore, the relative magnitudes of these three values ​​determine the geometric shape of the point cloud (e.g., ...). Represents a linear structure; (Represents planar structures).

[0063] Based on the above indicators, a classification and elimination strategy will be implemented:

[0064] Identifying planar interference: When the flatness of a point... Higher than the preset planar threshold (e.g.) and the smallest eigenvalue When the value is extremely small (representing a very weak change in the normal vector), the point is determined to belong to the surface of the concrete casing or the construction formwork and is therefore discarded.

[0065] Preserving the linear skeleton: when the linearity at a certain point Significantly higher than flatness (e.g.) And the local surface variation (which is equal to...) When the value is high, the point is determined to be a reinforcing bar with significant one-dimensional linear characteristics and is retained.

[0066] Through the geometric dimensionality reduction analysis based on eigenvalue tensors, the system can mathematically distinguish between "slender cylinders (reinforcing bars)" and "large-area curved / planar surfaces (casings / formwork)," thereby effectively removing the interference data surrounding the reinforcing bars and generating net point cloud data with high signal-to-noise ratio, clearing the way for subsequent slicing and feature extraction.

[0067] This application's embodiments employ a combination of "coarse-to-fine hierarchical registration" and "geometric feature-based hybrid denoising" to solve the dual challenges of inconsistent coordinates and severe background interference in wind turbine foundation scenarios. The hierarchical registration strategy utilizes a rigid transformation model combined with feature alignment and ICP iteration to ensure sub-millimeter-level alignment between on-site data and design benchmarks, preventing the algorithm from getting trapped in local optima. Meanwhile, the curvature and normal vector-based denoising logic leverages the physical characteristics of "reinforcing steel with linear high curvature and formwork with planar low curvature" to accurately remove complex construction backgrounds from the source. This improves the quality of the data source, allowing subsequent algorithms to focus on clean reinforcing steel skeleton data, thereby significantly improving the robustness and statistical accuracy of reinforcing steel identification.

[0068] Figure 3 The diagram illustrates an example of obtaining an initial cross-section candidate set through layered slicing and target detection in a method according to an embodiment of this application.

[0069] like Figure 3 As shown, in step S310, the slice thickness is set according to the minimum rebar diameter in the theoretical specification parameters to ensure that the slice data can contain complete rebar cross-sectional information.

[0070] Specifically, considering the three-dimensional complexity of the steel reinforcement arrangement in the wind turbine foundation, direct three-dimensional segmentation is extremely difficult. Therefore, a dimensionality reduction strategy of breaking down the entire structure into smaller parts is adopted. The system analyzes the design and construction data and extracts the diameter parameters of the smallest steel reinforcement. The thickness of the layered slices is set accordingly. To satisfy the sampling theorem constraints The net point cloud data is divided into multiple consecutive height layer slices. This ensures that each rebar is captured by at least two slices in the vertical direction, thus avoiding the adhesion of multiple rebars due to excessively thick slices or the loss of cross-sectional information due to excessively thin slices. This achieves efficient division of 3D point cloud data into multiple consecutive height layer slices without losing structural features.

[0071] In step S320, for each slice, a two-dimensional projection grid is constructed, the number of point clouds falling into each grid cell is counted and normalized, and a grayscale image reflecting the point cloud distribution density of that slice is generated.

[0072] Specifically, for each height layer slice, the system projects its contained 3D discrete point cloud onto a horizontal plane and divides it into a fine 2D ocupancy grid. Then, the number of point clouds falling into each grid cell is counted and mapped to pixel grayscale values ​​between 0 and 255, thereby generating the corresponding slice grayscale density map for that layer. This enables the conversion from unstructured point cloud data to structured image data, allowing subsequent processing to fully utilize mature image processing algorithms. Furthermore, the introduction of grayscale values ​​preserves the physical density characteristics of the reinforcing bars, which are "dense at the center and sparse at the edges."

[0073] In step S330, morphological structural elements are adaptively constructed based on the minimum rebar diameter. Morphological edge enhancement operations are performed on the grayscale image using the structural elements. The edge contour of the rebar section is highlighted by calculating the difference between the morphological dilation result and the morphological erosion result, and a morphological gradient feature map is generated as slice feature data.

[0074] Here, to effectively address the edge blurring problem caused by uneven illumination or sparse point clouds during on-site scanning, a morphological gradient algorithm based on specification constraints is introduced. Specifically, it utilizes structuring elements... grayscale density map of slices Perform morphological gradient operations to generate morphological gradient feature maps of the reinforcing bar edge profile. Furthermore, the morphological gradient feature map was used as the slice feature data.

[0075] For example, based on the minimum steel bar diameter Adaptively construct a disk-shaped morphological structural element Its size is set to This is done to precisely match the geometric dimensions of the steel reinforcement cross-section. Subsequently, this structural element is used... grayscale density map of slices Perform morphological gradient operations to generate morphological gradient feature maps of the reinforcing bar edge profile. Its mathematical expression is:

[0076] Equation (3)

[0077] In the formula, This represents a morphological expansion operation used to expand highlighted areas to fill voids inside the reinforcing bars. The grayscale image represents a slice of the layer. This represents a morphological erosion operation, used to shrink highlighted areas to remove background noise. pixel coordinates; structuring element The size is set to To match the geometric dimensions of the steel reinforcement cross-section.

[0078] In equation (3), by calculating the difference between the expansion map and the corrosion map, large areas of uniform regions in the image (such as the background of the casing) are eliminated, which greatly highlights the gray-scale transition edge of the steel bar section and provides a high-contrast feature map for subsequent detection.

[0079] In step S340, the morphological gradient feature map is used as input to perform Hough circle transformation, and the high response region is located by using the gradient magnitude accumulation voting mechanism. The center and radius of several circular targets are extracted to form an initial cross-section candidate set.

[0080] Here, based on the generated morphological gradient feature map The system utilizes the Hough Gradient Method for geometric shape detection. The algorithm uses edge pixels with high response in the gradient feature map as voting points, accumulating votes on possible center positions and radii in the parameter space. By setting an accumulator threshold, the system can robustly extract target regions exhibiting significant circular features from complex background noise and resolve their center coordinates and radii, thus constructing an initial cross-section candidate set containing all potential rebar sections.

[0081] This application's embodiments employ an adaptive morphological gradient enhancement strategy based on specification constraints, solving the problem of edge detection failure in traditional visual methods on low-quality point cloud slices. The system utilizes physical world design parameters (minimum rebar diameter) To guide the algorithm parameters (slice thickness) in the digital world With structural elements This constructs a mapping logic between physical and digital spaces. Furthermore, by applying morphological gradient operations, not only are the internal voids in the rebar cross-section effectively filled, but also large-area background interference is automatically filtered out using difference operations, significantly improving the signal-to-noise ratio at the rebar edges. This enables the subsequent Hough transform to accurately locate the geometric center of the rebar even under extreme conditions where the point cloud is sparse or adhered.

[0082] Figure 4 A flowchart illustrating an example of identifying valid steel reinforcement sections by constructing specification constraint factors according to an embodiment of this application is shown.

[0083] like Figure 4 As shown, in step S410, the design and construction data corresponding to the current slice level are parsed, and the theoretical design radius and theoretical design adjacent spacing of the steel bars in that level are extracted as prior constraint parameters.

[0084] It should be noted that in actual construction, the specifications of the reinforcing bars at different heights may vary due to stress requirements. Therefore, the system first indexes the BIM design model or parametric database based on the height coordinates of the current slice within the wind turbine foundation to accurately extract the design true value corresponding to that level. Specifically, it extracts the theoretical design radius of the reinforcing bars. (For example, a radius of 16mm corresponding to a single 32mm diameter rebar, or a radius of 20mm corresponding to a single 40mm diameter rebar) and the theoretical design adjacent spacing between adjacent rebars in the same layer. (e.g., 150mm) These two parameters constitute the benchmark for subsequent algorithms to determine "what kind of target is a real steel bar", providing a rigid prior basis for screening based on specification constraints.

[0085] In step S420, for any candidate target in the initial cross-section candidate set, a cross-section is constructed that includes geometric consistency evaluation and spatial topology consistency evaluation.

[0086] Here, the specification constraint factor is configured such that when the detection radius of a candidate target deviates from the theoretical design radius, or when the distance between it and its neighboring targets deviates from the theoretical design proximity distance, the weight of the candidate target is reduced using an exponential decay function logic.

[0087] To effectively eliminate false targets caused by construction debris, concrete spots, etc., this embodiment introduces a physical constraint mechanism. The system calculates each candidate target. Specification constraint factor This factor quantitatively characterizes the degree of conformity between the target and the design specifications using the following formula:

[0088] Equation (4)

[0089] In the formula, Represents the first in the initial cross-section candidate set One candidate target section; Represents candidate target cross section The detection radius; This represents the theoretical design radius of a single steel bar. Indicates the theoretical design adjacent spacing; Represents candidate target cross section The set of spatial neighborhood target sections (e.g., neighbors within a certain range of Euclidean distance); Represents the first in the neighborhood set One neighboring target section; Represents candidate target cross section Target cross section with neighboring The Euclidean distance between the centers of the circles; and These are the radius deviation weighting coefficient and the spacing deviation weighting coefficient, used to adjust the penalty intensity; specification constraint factor. By utilizing the exponential decay property, the weight of candidate target sections that deviate from the prior constraint parameters is reduced.

[0090] In equation (4), the first term is the geometric consistency term, when the detection radius... The closer to the design radius The closer this term is to 1, the greater the deviation; this term decays exponentially towards 0. This adjusts the sensitivity to radius error. The second term is the spatial topology consistency term; the closer the distance between the candidate target and its nearest neighbor is to the design distance... The closer this value is to 1, the more it utilizes the regularity of the steel reinforcement arrangement to penalize targets that are "isolated" or have "abnormal spacing." This adjusts the sensitivity to spacing deviations. Furthermore, the two factors are multiplied to form the final specification constraint factor, ensuring that only targets with "correct dimensions and correct positions" receive high scores.

[0091] In step S430, visual features and specification constraints are fused to calculate the overall confidence score for each candidate target.

[0092] Here, the overall confidence score is a weighted sum of the visual gradient response intensity of the candidate target in the slice feature data and the specification constraint factor. Specifically, to balance "clear visibility" (visual features) and "matching" (specification constraints), the system calculates the overall confidence score for each candidate target. The calculation formula is as follows:

[0093] Equation (5)

[0094] In the formula, Represents candidate target cross section The normalized gradient response intensity in the slice feature data (e.g., the voting value normalization result of the Hough transform) is used to characterize the visual saliency of the target in the image. and These are the visual feature weights and the specification constraint weights (e.g., they can be set to...). =0.4, =0.6), used to balance the importance of both; in practical applications, it is recommended (For example This expresses the design philosophy of strong specification constraints, that is, when visual features are ambiguous, priority is given to adopting the target that conforms to the design rules.

[0095] Through this weighted fusion, even if a real steel bar has a weak visual response due to obstruction, it can still get a high comprehensive score as long as its size and position height meet the design specifications, thus avoiding missed detection.

[0096] In step S440, an iterative screening is performed on the initial cross-section candidate set based on the comprehensive confidence score.

[0097] Specifically, the target with the highest score is retained first, and redundant targets with too high overlap with it are suppressed until all candidate targets are traversed, thereby outputting an effective steel reinforcement section that meets the design specifications.

[0098] In some implementations, after obtaining the comprehensive score of all candidate targets, the system can employ an iterative non-maximum suppression (NMS) strategy for final cleaning. For example, firstly, all candidate targets are sorted... The targets are sorted from highest to lowest score, and the highest-scoring target is selected as the "confirmed rebar". Then, targets whose spatial overlap (IoU) exceeds a preset threshold (e.g., 0.3) are considered duplicate detections of the same rebar and are removed. This process is repeated until all candidate targets have been processed, ensuring that in the final output set of valid rebar cross-sections, each rebar retains only one optimal detection box, and that the detection result is the optimal solution verified by both visual evidence and design logic.

[0099] It should be noted that traditional visual methods rely solely on image features, making them highly susceptible to being misled by distracting objects of similar shapes (such as round pebbles or concrete spots). This application's embodiments construct a confidence evaluation system that deeply integrates visual perception and design priors, introducing specification constraint factors. The "hard constraints" of the physical world (design radius) Design spacing The rebar is transformed into "soft weights" in the algorithm world, and an exponential decay function is used to accurately reduce the weight of pseudo-targets that deviate from design specifications. In addition, by applying comprehensive confidence analysis, strong prior knowledge is used to complete and correct the logic when visual information is incomplete (such as partial occlusion), which improves the accuracy and robustness of rebar identification under complex working conditions.

[0100] Figure 5 A flowchart illustrating an example of spatial consistency verification and anomaly correction for an effective steel reinforcement section according to an embodiment of this application is shown.

[0101] like Figure 5 As shown, in step S510, based on the arrangement rules, the effective steel reinforcement sections are classified and assigned to the corresponding design specification set using the nearest neighbor matching principle.

[0102] Specifically, after completing the initial confidence screening, the system obtained a set of valid rebar sections. To evaluate the reasonableness of this set of detection results from a global perspective, it is first necessary to map these disorganized detection points back to the logical framework of the design. For example, the system reads the layout rules in the design construction data (e.g., the inner circle consists of 32mm diameter rebars, and the outer circle consists of 25mm diameter rebars) and calculates each valid section using the nearest neighbor matching principle. The distance from the theoretical reinforcement positions on the design drawings. Based on the distance minimization criterion, each effective section is assigned to its corresponding set of design specifications. Among them The index representing the rebar specification category enables logical alignment between test data and design data, laying the foundation for subsequent calculation of overall error by specification.

[0103] In step S520, a global specification error evaluation model is constructed based on the classification results. This model quantifies the sum of geometric shape deviation and topological distribution deviation of all effective sections in the current level relative to the ideal design model through weighted aggregation.

[0104] Here, geometric shape deviation represents the difference between the detection radius and the theoretical radius, topological distribution deviation represents the difference between the actual spacing and the theoretical spacing, and the model includes weight terms to balance the numerical magnitude difference between the radius and the spacing.

[0105] In some implementations, to accurately determine whether the detection quality of the current slice layer meets the standard, this embodiment constructs an energy function that can characterize the "overall disorder," namely the global specification error function. This function no longer focuses on the gains or losses of individual targets, but rather aggregates the total deviations of all effective sections within the current level relative to the ideal design model. More specifically, the global specification error function is expressed by the following formula:

[0106] Equation (6)

[0107] In the formula, This represents the global specification error function value, used to quantitatively evaluate the overall quality of the current level. This indicates the total number of different types of steel bars included in the design and construction data; Indicates being classified as the first A subset of effective cross-sections of various specifications; Indicates belonging to a subset The One effective cross-section instance; This indicates the effective cross section. The detection radius; Indicates the relationship with the first The theoretical design radius corresponding to each specification; Indicates the effective cross section The actual physical distance between its nearest spatial neighbor section; Indicates the relationship with the first The theoretical design adjacent spacing corresponding to each specification; and These are the radius error weight and the spacing error weight, used to balance the differences in numerical magnitudes.

[0108] It is particularly important to note that, and These are the radius error weight and the spacing error weight, respectively. The logic behind introducing these two weights is to balance the numerical magnitude difference between the radius error (usually small, such as a few millimeters) and the spacing error (usually large, such as tens of millimeters), preventing the large spacing error from masking the small radius error, and ensuring that the two types of features have a balanced influence in the error model. Furthermore, since the spacing error (usually tens of millimeters) is significantly larger than the radius error (usually a few millimeters), it is recommended to set [weight missing]. Alternatively, the error term can be normalized to ensure that both terms contribute to the total energy. Their contributions are on the same order of magnitude.

[0109] In step S530, the global specification error evaluation value of the global specification error evaluation model is calculated, and the global specification error evaluation value is compared with the tolerance threshold set based on the design tolerance range. If the global specification error evaluation value exceeds the tolerance threshold, it is determined that the detection result of the current level has a systematic deviation or local occlusion anomaly, and an anomaly correction operation is triggered.

[0110] For example, if If the current level of detection results has a systematic deviation or local occlusion anomaly, then an anomaly correction operation will be triggered.

[0111] Specifically, the global error value at the current level It intuitively reflects the "distance" between the on-site construction (or test results) and the perfect design model. The system then compares this value with a preset tolerance threshold. The threshold was compared. This is the energy threshold calculated based on the allowable construction deviation range in engineering acceptance specifications (such as the "Acceptance Regulations for Wind Power Project Construction"). If This means that the detection results at the current level have a significant systematic bias (such as an overall alignment error) or local occlusion anomalies (such as a large gap caused by the casing occlusion). If the number of cases increases abnormally, the system determines that the detection results cannot be directly used for counting and immediately triggers the subsequent abnormal correction process (i.e., logical completion operation for missing parts) to ensure the integrity of the final statistical data.

[0112] This application establishes a global quality verification mechanism based on design isomorphism. Unlike traditional methods that focus only on the recognition rate of individual targets, this scheme constructs a global specification error function to restore discrete detection points into a holistic structure with topological relationships. By weighted aggregation of geometric shape deviations and topological distribution deviations, this mechanism can sensitively capture structural defects caused by occlusion (manifested as a significant increase in the spacing error term) or specification deviations caused by construction errors (manifested as a significant increase in the radius error term). Thus, the verification logic from local to global is equivalent to adding an engineering supervision and review mechanism to the vision algorithm, effectively preventing statistical distortion caused by local false detections or missed detections.

[0113] In some examples of embodiments of this application, the anomaly correction operation includes a missing completion operation for occluded areas.

[0114] More specifically, firstly, the system traverses the set of valid rebar sections within the current level, analyzes the spatial topological relationships between adjacent sections, and calculates their actual physical spacing. Here, after completing global error verification, if an anomaly is determined in the current level, the system will initiate refined completion logic. The system first establishes topological adjacency relationships on the two-dimensional plane of the current height level based on the spatial location of the valid rebar sections (e.g., through Delaunay triangulation or simple chain sorting), determining each pair of spatially adjacent sections. Then, it calculates the Euclidean distance between the centers of adjacent sections, i.e., the actual physical spacing. .

[0115] If the actual physical distance between adjacent sections is detected Larger than the theoretical design spacing ,For example Greater than If the area is 1.5 times larger than the design specifications and is a reinforced area in the design data (excluding pre-reserved holes), then it is determined that there is a missing rebar due to obstruction. Here, the system uses geometric logic to determine that if there is a visual blind spot between two adjacent sections caused by the concrete casing or insufficient lighting, it can be determined that there is a missing rebar.

[0116] Then, based on the spatial coordinates of the known effective reinforcement sections at both ends of the adjacent sections. and and theoretical design spacing Inferring the theoretical number of missing reinforcing bars .

[0117] Specifically, to fill this visual blind spot, the system first needs to determine "how many bars are missing." This is done by utilizing the "anchor points" at both ends—which have already been identified with high confidence—that are, the known effective rebar cross-sections. and Combined with the standard spacing in the design specifications The system uses an inference calculation model to quantitatively calculate the theoretical number of missing steel bars. .

[0118] For example, the inference calculation model is as follows:

[0119] Equation (7)

[0120] Regarding the explanation of equation (7), This represents the theoretical number of missing steel bars (integer value). This represents the rounding function, used to eliminate non-integer interference caused by measurement errors; This represents the Euclidean distance between the center coordinates of the known effective steel reinforcement sections at both ends. In equation (7), by calculating how many standard spacing units are included in the total span between the two points, subtracting 1 gives the number of steel reinforcements that should exist in the middle but have not been detected.

[0121] Then, using linear spatial interpolation logic, a corresponding number of virtual steel reinforcement cross-section center coordinates are generated between the known coordinates at both ends of adjacent cross-sections according to the theoretical design spacing. The generated virtual steel reinforcement cross-sections are then merged into the set of valid steel reinforcement cross-sections, thereby obtaining the corrected steel reinforcement cross-section detection results.

[0122] Here, after determining the number of missing reinforcement bars, the system uses linear vector interpolation to reconstruct the spatial location of the missing bars. The system assumes that the reinforcement bars are uniformly distributed within a short distance and uses unit direction vectors to guide the interpolation generation.

[0123] Specifically, no. Theoretical center coordinates of the missing rebar It can be calculated using the following formula (8):

[0124] Equation (8)

[0125] In equation (8), Indicates the generated first The coordinates of the center of a virtual steel reinforcement section; Represents the coordinates of the known effective reinforcement section at the starting end; This is the index variable for missing rebar, with values ​​ranging from 1 to... ; This represents the unit direction vector from the starting point to the ending point, used to ensure that the interpolation point lies strictly on the line connecting the two endpoints.

[0126] Finally, based on the calculated coordinates Generate virtual steel reinforcement sections (radius can be the theoretical radius). The process involves merging all virtual rebar sections into the set of valid rebar sections, resulting in a corrected rebar section detection result. This reconnects the previously broken rebar chains caused by obstruction, leading to a corrected rebar section detection result that conforms to design logic.

[0127] Through the embodiments of this application, a logical completion mechanism based on topological adjacency is constructed. Traditional visual algorithms often fail to address the common occlusion problems (such as mud, formwork, and equipment occlusion) at wind turbine foundation construction sites, resulting in undercounting. By using detected high-confidence targets as spatial anchors and combining them with the strong prior constraint of design spacing, the visually "invisible" is transformed into logically "inferable." Through quantitative calculations using formulas (7) and (8), the system can accurately "virtually reconstruct" missing steel bars in blind areas without relying on pixel features, greatly improving the completeness and reliability of statistical results under complex working conditions.

[0128] Regarding the implementation details of outputting the rebar quantity statistics report in step S150, in some examples of embodiments of this application, firstly, the corrected rebar cross-section detection results are sorted according to rebar specification category. Aggregate and count the total number of detections for each algorithm specification. And retrieve the corresponding total number of theoretical designs. .

[0129] Specifically, after completing anomaly correction and missing data completion, the system obtains a complete and logically consistent list of valid rebar sections. At this point, the system categorizes each section into its corresponding specification category based on its attributes (such as diameter and grade). Below (e.g., "32mm diameter main reinforcement"). Subsequently, the system automatically counts the total number of reinforcement bars actually detected by the algorithm for each specification. Simultaneously, the system indexes the design and construction data again, retrieving the total theoretical design quantity of this specification as specified in the design drawings. This achieved synchronous alignment between measured and theoretical values, providing a data foundation for subsequent deviation analysis.

[0130] Then, define the permissible deviation thresholds for various specifications. And construct segmented arbitration decision logic to determine the statistical quantity of the final output. .

[0131] Considering the complexity of engineering sites, measured data often exhibits fluctuations. To ensure the scientific validity and security of the output results, the system does not directly output the measured values ​​but instead introduces an intelligent arbitration mechanism. The system first sets an allowable deviation threshold for each specification based on engineering acceptance standards or historical experience. (e.g., allowable deviation) (Root). Next, the following piecewise arbitration function is constructed to calculate the statistical count of the final output. :

[0132] Equation (9)

[0133] In the formula, This indicates the final output number. Statistical count of class specifications; This represents the total number of detections after the algorithm was modified. This indicates the total number of theoretical designs in the design and construction data; Indicates the first The permissible deviation threshold corresponding to each specification.

[0134] This reflects two different decision-making logics: on the one hand, when the algorithm detects the difference between the total number of tests and the theoretically designed total number... When the detection result is within the allowable deviation threshold, the total number of detections by the algorithm is considered valid. As an output that truly reflects the situation on site. On the other hand, when the difference... When the deviation exceeds the allowable threshold, the system determines that the detection result has a serious deviation and uses the total theoretical design quantity as the output result. Here, when the difference exceeds the allowable range, the system determines that there may be serious occlusion interference, algorithm failure, or a major construction accident. To prevent erroneous measured data from misleading subsequent settlements, the system adopts a conservative strategy and uses the total theoretical design quantity. As an output, it is used to ensure the security and compliance of accounting data.

[0135] Furthermore, based on the statistical quantities of the final output Generate a steel reinforcement quantity statistics report; for cases where the difference exceeds the limit, such as when... When an anomaly is detected in the steel reinforcement quantity statistics report, an anomaly warning is triggered, the location of the anomaly area is marked, and a point cloud visualization snapshot of that area is associated with it for manual review.

[0136] Here, the system is based on the calculations obtained. The system automatically generates standardized steel reinforcement quantity statistics reports, listing the specifications, locations, and quantities of various types of steel reinforcement. In particular, for specifications that trigger the "fallback correction" logic (i.e., exceeding the difference limit), the system not only highlights the anomaly in the report but also utilizes the visualization capabilities of 3D point clouds to automatically capture a point cloud snapshot of the abnormal area (such as missing or severely overlapping areas) and embed it into the report. This provides engineering managers with a visually appealing review portal, enabling them to quickly pinpoint the source of the problem and ensuring traceable evidence.

[0137] Traditional automated quantity surveying software often reports the exact amount calculated, and errors in the output data can easily lead to engineering disputes if the algorithm malfunctions or there is significant obstruction at the site. This application's embodiment establishes an intelligent decision-making and feedback mechanism based on tolerance arbitration, constructing segmented arbitration logic to integrate the authenticity of measured data with the authority of design data. Therefore, it can accurately reflect subtle deviations at the site under normal circumstances, and automatically activate safety circuit breakers in abnormal situations, forcing a regression to design values ​​to avoid major misjudgments. Simultaneously, the linked output of anomaly alerts and visual snapshots greatly enhances the practical value of the report, assisting manual inspection of potential hazards and truly achieving intelligent and closed-loop quantity surveying.

[0138] Figure 6 A schematic diagram illustrating the principle mechanism of an example of a specification-constrained method for counting the number of steel bars in a wind turbine foundation according to an embodiment of this application is shown. This mechanism employs a dual-stream fusion processing architecture that integrates on-site measured data and design construction data.

[0139] like Figure 6 As shown, the upper-layer data stream is mainly responsible for the feature deconstruction of the field data: first, it performs registration on the collected field point cloud, and then converts the 3D data into sliced ​​grayscale images through layered slicing and density projection. Then, morphological gradient enhancement and Hough transform techniques are used to extract geometric features and generate an initial cross-section candidate set containing potential reinforcement targets; the lower-level data flow is mainly responsible for the logical construction of design priors: the system parses the theoretical specifications in the BIM design and construction data. and adjacent spacing Constructing quantitative specification constraint factors These two data streams are deeply integrated in the confidence filtering and global verification modules. Design constraints are used to perform weighted correction and overall consistency verification on the visual inspection results. Through completion operations for occluded areas and segmented arbitration decision logic, a visual statistical report containing point cloud snapshots of abnormal areas is finally generated.

[0140] Figure 7 The diagram shows the comparative experimental results of counting accuracy and stability under various occlusion and noise environments using different methods.

[0141] like Figure 7 As shown, in order to verify the effectiveness of the proposed method for counting the number of steel bars in wind turbine foundations based on specification constraints, a simulation dataset containing various degrees of shading (0% to 30% random distribution) and environmental noise interference was constructed. The method of this application, the MaskID deep learning algorithm in related technologies, and traditional manual statistical methods were compared and tested. Figure 7 The distribution of F1-scores for the three methods on 50 random samples is illustrated using a violin plot. Figure 7 As can be seen, the manual statistical method (right side) is affected by subjective factors and fatigue, resulting in a wide range of accuracy fluctuations and a large variance, indicating poor stability. While the MaskID algorithm (middle side) utilizes visual recognition technology, it still suffers from missed detections due to feature loss when handling complex occlusion, with F1-scores mainly concentrated between 0.97 and 0.99. In contrast, the proposed method (left side) benefits from incorporating theoretical specifications from the design and construction data as prior constraints. It effectively completes and corrects missing targets caused by occlusion through a global specification error function and segmented arbitration decision logic. Therefore, it significantly outperforms the comparative schemes in accuracy, with an average F1-score exceeding 0.99 and the most compact distribution. Thus, this result strongly demonstrates that the proposed method possesses extremely high counting accuracy and robustness under high occlusion and high interference conditions at wind turbine foundation construction sites.

[0142] Figure 8 A schematic diagram showing the relationship between the mean absolute error of different methods under various occlusion rates is presented.

[0143] like Figure 8As shown, to further quantify the robustness of the proposed method in dynamic occlusion environments, simulation scenarios with six different occlusion rate gradients from 0 to 0.5 (i.e., 50%) were set up, and the mean absolute error (MAE) of the three methods at each gradient was calculated. The trends in the curves clearly show that manual statistics (green legend) and the MaskID algorithm (orange legend), limited by their strong dependence on visual features, exhibit a significant linear upward trend in their error curves as the occlusion rate increases, indicating that the lack of visual information directly leads to a rapid decline in recognition accuracy. In contrast, the proposed method (blue legend), benefiting from the logical compensation effect of the specification constraint factor and the "missing information completion operation" performed on the occluded region, shows extremely smooth growth in its error curve under extreme conditions of continuously worsening occlusion rates, demonstrating strong stability. Specifically, even under extreme conditions where the occlusion rate reaches 0.5, the error of the method in this application is still controlled below 0.02, while the errors of MaskID and manual statistics have climbed to nearly 0.04 and 0.06 respectively. The significantly widened error gap between the three strongly proves that the method in this application effectively solves the problem of steel bar counting under complex interference environments by integrating prior design knowledge.

[0144] This application proposes an automatic statistical method based on specification constraints to address the unique difficulties in counting the number of reinforcing bars in wind turbine foundations. By comprehensively utilizing BIM design data and on-site point cloud scanning, and employing adaptive morphological gradient enhancement and global specification error verification strategies, it effectively overcomes the limitations of traditional vision technologies in handling complex 3D reinforcement layouts and severe occlusion. By introducing specification constraint factors, anomaly correction and completion, and segmented arbitration decision logic, this method deeply integrates prior design information throughout the entire detection process. This not only achieves accurate identification of the location, diameter, and spacing of reinforcing bar sections but also ensures the logical self-consistency of the statistical results under occlusion conditions. Experimental verification shows that this method significantly outperforms manual statistics and the MaskID algorithm in both accuracy and stability, providing a high-precision intelligent quantity calculation tool for the digital construction and compliance acceptance of large and complex reinforced concrete structures such as wind turbine foundations.

[0145] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of combined actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Secondly, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application. In the above embodiments, the descriptions of each embodiment have their own emphasis; for parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0146] Figure 9 A structural block diagram of an example of a specification-constrained wind turbine foundation reinforcement quantity counting system according to an embodiment of this application is shown.

[0147] like Figure 9 As shown, the wind turbine foundation rebar quantity statistics system 900 based on specification constraints includes a data acquisition unit 910, a data registration and preprocessing unit 920, a slicing and target detection unit 930, a specification constraint and screening unit 940, and a verification and statistical reporting unit 950.

[0148] The data acquisition unit 910 is used to acquire the design and construction data of the wind turbine foundation and the scanning point cloud data of the construction site; the design and construction data includes the theoretical specifications and arrangement rules of the reinforcing bars, and the scanning point cloud data reflects the three-dimensional spatial morphology of the reinforcing cage.

[0149] The data registration and preprocessing unit 920 is used to perform coordinate system registration and noise reduction preprocessing on the scanned point cloud data based on the reference control points in the design and construction data, so as to generate net point cloud data with coordinates consistent with the design and construction data.

[0150] The slicing and target detection unit 930 is used to perform layered slicing processing on the net point cloud data along the height direction, generate slice feature data reflecting the cross-sectional distribution of steel bars at different height layers, and perform target detection for each slice feature data to obtain an initial cross-section candidate set.

[0151] The specification constraint and screening unit 940 is used to construct specification constraint factors using the theoretical specification parameters, and to evaluate and screen the initial cross-section candidate set based on the specification constraint factors, so as to identify the effective steel reinforcement cross-sections that meet the design specifications; the specification constraint factors are used to characterize the degree of agreement between the candidate targets and the theoretical specification parameters.

[0152] The verification and statistical reporting unit 950 is used to perform spatial consistency verification and anomaly correction on the effective steel bar cross section based on the arrangement rules and design tolerance range, count the number of steel bars of each specification based on the corrected steel bar cross section detection results, and output a steel bar quantity statistical report.

[0153] In some embodiments, this application provides a non-volatile computer-readable storage medium storing one or more programs including execution instructions. The execution instructions can be read and executed by electronic devices (including but not limited to computers, servers, or network devices) to perform the steps of any of the above-described methods for calculating the quantity of steel reinforcement in wind turbine foundations based on specification constraints.

[0154] In some embodiments, this application also provides a computer program product, the computer program product including a computer program stored on a non-volatile computer-readable storage medium, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the steps of any of the above-described methods for calculating the quantity of steel reinforcement in wind turbine foundations based on specification constraints.

[0155] In some embodiments, this application also provides an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of a method for calculating the quantity of steel reinforcement in a wind turbine foundation based on specification constraints.

[0156] The above-described product can perform the methods provided in the embodiments of this application, and has the corresponding functional modules and beneficial effects for performing the methods. Technical details not described in detail in this embodiment can be found in the methods provided in the embodiments of this application.

[0157] The electronic devices in this application can exist in various forms, including but not limited to: mobile communication devices, ultra-mobile personal computer devices, portable entertainment devices, or other airborne electronic devices with data interaction functions.

[0158] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0159] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0160] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for statistically analyzing the quantity of reinforcing steel bars in wind turbine foundations based on specification constraints, characterized in that, The method includes: The design and construction data of the wind turbine foundation and the scanning point cloud data of the construction site are obtained; the design and construction data includes the theoretical specifications and arrangement rules of the reinforcing bars, and the scanning point cloud data reflects the three-dimensional spatial morphology of the reinforcing cage. Based on the reference control points in the design and construction data, coordinate system registration and noise reduction preprocessing are performed on the scanned point cloud data to generate net point cloud data with coordinates consistent with the design and construction data. The net point cloud data is sliced ​​along the height direction to generate slice feature data that reflects the cross-sectional distribution of steel bars at different height levels, and target detection is performed on each slice feature data to obtain an initial cross-section candidate set. Specification constraint factors are constructed using the theoretical specification parameters, and the confidence level of the initial cross-section candidate set is evaluated and screened based on the specification constraint factors to identify effective steel reinforcement cross-sections that meet the design specifications; the specification constraint factors are used to characterize the degree of agreement between the candidate targets and the theoretical specification parameters. Based on the arrangement rules and design tolerance range, the spatial consistency of the effective steel bar cross section is checked and anomalies are corrected. The quantity of steel bars of each specification is counted based on the corrected steel bar cross section test results, and a steel bar quantity statistics report is output. The step of constructing specification constraint factors using the theoretical specification parameters, and evaluating and screening the initial cross-section candidate set based on the specification constraint factors to identify effective reinforcement cross-sections that meet the design specifications, includes: Analyze the design and construction data corresponding to the current slice level, and extract the theoretical design radius and theoretical design adjacent spacing of the reinforcement in that level as a priori constraint parameters; For any candidate target in the initial cross-section candidate set, a specification constraint factor is constructed that includes geometric consistency evaluation and spatial topology consistency evaluation; the specification constraint factor is configured such that when the detection radius of the candidate target deviates from the theoretical design radius, or the distance between it and its neighboring targets deviates from the theoretical design adjacent distance, the weight of the candidate target is reduced using an exponential decay function logic. By fusing visual features and specification constraints, a comprehensive confidence score is calculated for each candidate target; the comprehensive confidence score is composed of the weighted sum of the visual gradient response intensity of the candidate target in the slice feature data and the specification constraint factor. Based on the comprehensive confidence score, an iterative screening is performed on the initial cross-section candidate set: the target with the highest score is retained first, and redundant targets with too high overlap are suppressed, until all candidate targets are traversed, thereby outputting an effective steel reinforcement cross-section that meets the design specifications.

2. The method according to claim 1, characterized in that, The step of performing coordinate system registration and noise reduction preprocessing on the scanned point cloud data based on the reference control points in the design and construction data to generate net point cloud data with coordinates consistent with the design and construction data includes: A spatial transformation model between the design coordinate system and the on-site scanning coordinate system is constructed, and a coarse-to-fine hierarchical registration strategy is implemented: First, the basic geometric center and edge contour features in the scanned point cloud data are extracted and aligned with the reference control points in the design construction data to complete the coarse registration; then, an iterative optimization algorithm is used to minimize the spatial distance error between the point cloud and the design model, and the optimal spatial transformation parameters are solved to accurately unify the scanned point cloud data to the design coordinate system. Hybrid denoising processing is performed on the registered point cloud data to obtain the net point cloud data: a density-based statistical outlier removal algorithm is used to remove discrete noise points in the scanned point cloud data; and based on the local geometric feature analysis of the point cloud, the normal vector and local curvature of each point in the remaining point cloud are calculated, and planar point cloud clusters with consistent normal vector directions and local curvatures below the smoothness threshold are identified, determined to be interference from the concrete casing surface or construction formwork and removed, while linear point cloud clusters with high curvature characteristics are retained as net point cloud data representing the steel reinforcement skeleton.

3. The method according to claim 1, characterized in that, The net point cloud data is sliced ​​along the height direction to generate slice feature data reflecting the cross-sectional distribution of steel bars at different height levels. Target detection is then performed on each slice feature data to obtain an initial cross-section candidate set, including: The slice thickness is set according to the minimum rebar diameter in the theoretical specifications to ensure that the slice data can contain complete rebar cross-sectional information; For each slice, a two-dimensional projection grid is constructed, the number of point clouds falling into each grid cell is counted and normalized, and a grayscale image reflecting the point cloud distribution density of the slice is generated. Based on the minimum rebar diameter, an adaptive morphological structural element is constructed. The structural element is used to perform morphological edge enhancement operations on the grayscale image. The edge contour of the rebar section is highlighted by calculating the difference between the morphological dilation result and the morphological erosion result. A morphological gradient feature map is generated as the slice feature data. The morphological gradient feature map is used as input to perform Hough circle transform. The high-response region is located by using the gradient magnitude accumulation voting mechanism. The center and radius of several circular targets are extracted to form the initial cross-section candidate set.

4. The method according to claim 1, characterized in that, The spatial consistency verification and anomaly correction of the effective steel reinforcement cross-section based on the arrangement rules and design tolerance range includes: Based on the arrangement rules, the effective steel reinforcement sections are classified and assigned to the corresponding design specification sets using the nearest neighbor matching principle; Based on the classification results, a global specification error evaluation model is constructed. The model quantifies the sum of geometric shape deviation and topological distribution deviation of all effective sections in the current level relative to the ideal design model through weighted aggregation. The geometric shape deviation represents the difference between the detection radius and the theoretical radius, and the topological distribution deviation represents the difference between the actual spacing and the theoretical spacing. The model also includes a weight term to balance the numerical magnitude difference between the radius and the spacing. Calculate the global specification error evaluation value of the global specification error evaluation model, and compare the global specification error evaluation value with the tolerance threshold set based on the design tolerance range; If the global specification error evaluation value exceeds the tolerance threshold, it is determined that the detection result of the current level has a systematic deviation or local occlusion anomaly, and an anomaly correction operation is triggered.

5. The method according to claim 4, characterized in that, The anomaly correction operation includes a missing completion operation for occluded areas, specifically including: Traverse the set of valid steel reinforcement sections within the current level, analyze the spatial topological relationship between adjacent sections, and calculate their actual physical spacing; If the actual physical spacing between adjacent sections is found to be greater than the theoretical design spacing, and the area is a reinforcement area in the design construction data, then it is determined that there is a lack of reinforcement due to obstruction. Based on the spatial coordinates of the known effective steel reinforcement sections at both ends of adjacent sections and the theoretical design spacing, the theoretical number of missing steel reinforcements is inferred. Using linear spatial interpolation logic, a corresponding number of virtual steel reinforcement cross-section center coordinates are generated between the known coordinates at both ends of the adjacent cross-sections according to the theoretical design spacing. The generated virtual steel reinforcement cross-sections are then merged into the set of effective steel reinforcement cross-sections, thereby obtaining the corrected steel reinforcement cross-section detection result.

6. The method according to claim 1, characterized in that, The process of calculating the quantity of steel bars of each specification based on the corrected steel bar cross-section test results and outputting a steel bar quantity statistics report includes: The corrected steel bar section detection results are aggregated according to steel bar specification category, the total number of algorithm detections for each specification is counted, and the corresponding total number of theoretical design detections is retrieved. Define permissible deviation thresholds for various specifications and construct segmented arbitration decision logic to determine the statistical quantities of the final output: When the difference between the total number of detections by the algorithm and the total number of theoretically designed detections is within the allowable deviation threshold, the detection result is deemed valid, and the total number of detections by the algorithm is adopted as the output result. When the difference exceeds the allowable deviation threshold, it is determined that the detection result has a serious deviation, and the theoretically designed total number is used as the output result; The steel reinforcement quantity statistics report is generated based on the final output statistics. If the difference exceeds the limit, an anomaly warning is triggered in the steel reinforcement quantity statistics report, the location of the abnormal area is marked and associated with the point cloud visualization snapshot of the area for manual review.

7. A system for counting the quantity of reinforcing steel bars in wind turbine foundations based on specification constraints, used to implement the method as described in any one of claims 1-6; characterized in that, The system includes: The data acquisition unit is used to acquire the design and construction data of the wind turbine foundation and the scanning point cloud data of the construction site; the design and construction data includes the theoretical specifications and arrangement rules of the reinforcing bars, and the scanning point cloud data reflects the three-dimensional spatial morphology of the reinforcing cage. The data registration and preprocessing unit is used to perform coordinate system registration and noise reduction preprocessing on the scanned point cloud data based on the reference control points in the design and construction data, so as to generate net point cloud data with coordinates consistent with the design and construction data. The slicing and target detection unit is used to perform layered slicing processing on the net point cloud data along the height direction, generate slice feature data reflecting the cross-sectional distribution of steel bars at different height layers, and perform target detection for each slice feature data to obtain an initial cross-section candidate set. The specification constraint and screening unit is used to construct specification constraint factors using the theoretical specification parameters, and to evaluate and screen the initial cross-section candidate set based on the specification constraint factors, so as to identify the effective steel reinforcement cross-sections that meet the design specifications; the specification constraint factors are used to characterize the degree of agreement between the candidate targets and the theoretical specification parameters. The verification and statistical reporting unit is used to perform spatial consistency verification and anomaly correction on the effective steel bar cross section based on the arrangement rules and design tolerance range, count the number of steel bars of each specification based on the corrected steel bar cross section detection results, and output a steel bar quantity statistical report.

Images