A Precise Positioning Construction Method for Wind Power Steel Pipe Piles Based on BIM Pattern Recognition Technology

Abstract

This invention pertains to the field of offshore wind power equipment installation and construction applications, and discloses a method for precise positioning of wind turbine steel pipe piles based on BIM pattern recognition technology. The method retrieves steel pipe pile parameter data from a constructed BIM model, acquires construction scene primitive features, and obtains real-world image data of the construction scene via high-precision satellite imagery. Image fusion technology is used to process these two types of image data to obtain precise positioning image features for the wind turbine steel pipe piles. This precise positioning image feature, steel pipe pile parameter data, and a pre-defined precise positioning method are used to construct a precise positioning construction model for the wind turbine steel pipe piles. This invention combines the BIM-constructed image and the high-precision satellite-acquired image, and uses a scene-based improved image recognition model to output the positioning method. This assists managers in making precise positioning decisions during construction. This not only helps reduce positioning construction risks but also improves the accuracy of construction positioning and the level of intelligence in wind turbine steel pipe pile construction.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing equipment application technology, and in particular relates to a method for precise positioning and construction of wind power steel pipe piles based on BIM pattern recognition technology. Background Technology

[0002] Traditional offshore wind turbine foundation construction suffers from problems such as low positioning accuracy, reliance on manual labor, and limited construction efficiency. In traditional construction, steel pipe pile positioning mainly relies on total stations, GPS-RTK surveying equipment, and manual visual inspection and experience. This method has significant limitations: in the dynamic marine environment, measurements are easily affected by interference, resulting in horizontal positioning accuracy typically only controlled within 50-100mm, with vertical deviations reaching 0.5%-1%; the positioning process requires multiple stops for adjustments, with single-pile positioning taking 30-60 minutes; and construction quality is highly dependent on the technical skill and experience of the surveyors, lacking full-process digital recording and real-time correction capabilities. These factors collectively constrain the efficiency, cost control, and project quality of large-scale wind farm construction.

[0003] With the deep application of BIM (Building Information Modeling) technology in civil engineering, and the rapid development of computer vision and pattern recognition algorithms, new technical solutions have been provided for the aforementioned problems. BIM technology can integrate multi-source information such as foundation design, geology, topography, and marine environment to construct a high-precision, parametric 3D construction model, providing a unique and reliable digital benchmark for on-site positioning. Meanwhile, pattern recognition technologies, such as image recognition, point cloud registration, and deep learning, can now capture pile posture in real time using devices like visual sensors and laser scanners, and intelligently match and analyze deviations with the BIM model, achieving sub-centimeter-level recognition accuracy.

[0004] Therefore, combining the precise design model of BIM with the real-time perception and intelligent judgment capabilities of pattern recognition to form a closed-loop control system of "digital design - on-site perception - intelligent decision-making - automatic correction" has become an inevitable technological trend for improving the positioning accuracy and construction automation level of offshore wind power steel pipe piles. This application achieves a technological leap from extensive experience-based construction to precise and intelligent construction through the deep integration of BIM and pattern recognition.

[0005] Therefore, several challenges remain. These include how to retrieve relevant data based on the constructed BIM model, how to combine high-precision satellite imagery for precise positioning and installation of wind turbine piles, how to build existing data sources based on the BIM model, and how to perform format conversion processing of multi-source data to adapt to model changes. Current technologies often use a single model for automated construction positioning. However, there are still issues to address. Specifically, how to combine multiple models to personalize data from existing wind turbine pile installation scenarios, adapting image recognition algorithms to output precise positioning methods for wind turbine steel pipe piles in different installation scenarios, thus helping construction managers make accurate positioning decisions, reducing positioning risks, improving the accuracy of construction positioning, and enhancing the intelligence level of wind turbine steel pipe pile construction. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a method for precise positioning and construction of wind power steel pipe piles based on BIM pattern recognition technology.

[0007] In a first aspect of the present invention, a method for precise positioning and construction of wind turbine steel pipe piles based on BIM pattern recognition technology is provided, the method comprising:

[0008] S1. Construct a BIM model of the wind power steel pipe pile construction scene, retrieve the steel pipe pile parameter data in the BIM model, obtain the primitive features of the construction scene, use high-precision satellite to obtain real image data of the construction scene, and obtain the accurate positioning method of wind power steel pipe piles calibrated by technical personnel for the construction scene.

[0009] S2. The construction scene primitive features and the real image data of the construction scene are processed using image positioning fusion technology to obtain the precise positioning image features of the wind power steel pipe pile;

[0010] S3. Construct a construction model for the precise positioning of wind power steel pipe piles using the image features of the precise positioning of the wind power steel pipe piles, the parameter data of the steel pipe piles, and the calibrated precise positioning method of the wind power steel pipe piles;

[0011] S4. Based on the aforementioned wind power steel pipe pile precise positioning construction model, output an adaptive wind power steel pipe pile precise positioning method for different construction scenarios to assist construction management personnel in making judgments on precise positioning construction.

[0012] Furthermore, the steel pipe pile parameter data includes pile length, wall thickness, and anti-corrosion coating thickness.

[0013] Furthermore, the construction scene primitive features are obtained by processing and integrating the seabed soil layer distribution data, rock layer thickness data, rock layer bearing capacity data, seawater depth, sea surface wind speed data, and sea wave data of each image unit in the BIM model.

[0014] Furthermore, the real image data of the construction scene obtained by high-precision satellite is RGB three-channel image data. The image positioning and fusion technology first comprehensively processes the pixel channel image data of the image unit corresponding to the primitive features of the construction scene.

[0015] Furthermore, the image positioning fusion technology vertically stitches together the processed construction scene primitive features and the real image data of the construction scene to obtain the precise positioning image features of the wind power steel pipe pile.

[0016] Furthermore, the wind power steel pipe pile precise positioning construction model adopts a multi-cascaded classification machine learning model.

[0017] Furthermore, the multi-cascaded classification machine learning model is obtained by cascading a Fisher criterion classifier into a neural network model. The numerical values ​​obtained by inputting the precise positioning image features of the wind turbine steel pipe pile into the Fisher criterion classifier are fused with the parameter data of the steel pipe pile and input into the activation function of the neural network model.

[0018] A precise positioning construction system for wind power steel pipe piles based on BIM pattern recognition technology is also provided. This system implements a precise positioning construction method for wind power steel pipe piles based on BIM pattern recognition technology. The system includes a construction scene BIM model construction module, a steel pipe pile parameter data acquisition module, a construction scene primitive feature processing acquisition module, a high-precision satellite scene image acquisition module, a wind power steel pipe pile precise positioning construction model construction module, and a wind power steel pipe pile construction positioning analysis module.

[0019] The construction scenario BIM model building module is used to build a BIM model of the wind power steel pipe pile construction scenario and retrieve the steel pipe pile parameter data from the BIM model.

[0020] The steel pipe pile parameter data acquisition module is used to acquire the steel pipe pile parameter data in the BIM model of the wind power steel pipe pile construction scenario.

[0021] The construction scene primitive feature processing and acquisition module is used to acquire construction scene primitive features.

[0022] The high-precision satellite scene image acquisition module is used to acquire real image data of the construction scene using high-precision satellites.

[0023] The wind power steel pipe pile precise positioning construction model construction module: processes the construction scene primitive features and the real image data of the construction scene using image positioning fusion technology to obtain the precise positioning image features of the wind power steel pipe pile; and constructs the precise positioning construction model of the wind power steel pipe pile using the precise positioning image features of the wind power steel pipe pile, the steel pipe pile parameter data, and the calibrated precise positioning method of the wind power steel pipe pile.

[0024] The wind power steel pipe pile construction positioning analysis module: Based on the wind power steel pipe pile precise positioning construction model, it outputs an adaptive wind power steel pipe pile precise positioning method for different construction scenarios, assisting construction managers in making judgments on precise positioning construction.

[0025] Furthermore, the wind power steel pipe pile precise positioning construction model adopts a multi-cascaded classification machine learning model;

[0026] The multi-cascaded classification machine learning model is obtained by cascading a Fisher criterion classifier into a neural network model. The numerical values ​​obtained by inputting the precise positioning image features of the wind power steel pipe pile into the Fisher criterion classifier are fused with the parameter data of the steel pipe pile and input into the activation function of the neural network model.

[0027] Therefore, the beneficial effects of this invention are as follows: It retrieves steel pipe pile parameter data from the constructed BIM model, obtains construction scene primitive features and real-world image data of the construction scene acquired through high-precision satellite imagery, and combines these features with image positioning fusion technology to obtain precise positioning image features for wind power steel pipe piles. Using these precise positioning image features, steel pipe pile parameter data, and a pre-defined precise positioning method, it constructs a precise positioning construction model for wind power steel pipe piles. Finally, it uses an image recognition algorithm to process the data and obtain an instantaneous precise positioning method for wind power steel pipe piles. This invention combines BIM-constructed images and high-precision satellite images, and utilizes a cascaded image recognition algorithm model adapted to wind power steel pipe pile construction scenarios to output an instantaneous precise positioning method for wind power steel pipe piles. This helps construction managers make precise positioning decisions during construction. This not only helps reduce positioning risks but also improves the accuracy of construction positioning and enhances the intelligence level of wind power steel pipe pile construction.

[0028] Further embodiments and improvements of the present invention will be described in conjunction with the accompanying drawings and specific examples. Attached Figure Description

[0029] Figure 1 This is a flowchart of the wind power steel pipe pile precise positioning construction method based on BIM pattern recognition technology of the present invention;

[0030] Figure 2 This is a schematic diagram of the wind power steel pipe pile precise positioning construction system based on BIM pattern recognition technology of the present invention;

[0031] Figure 3 This is a simulation example diagram in the BIM model of wind power steel pipe piles in this invention;

[0032] Figure 4 This is a schematic diagram illustrating the principle of the multi-cascade classification machine learning model in an embodiment of the present invention;

[0033] Figure 5 This is a schematic diagram of an electronic device structure for implementing the method of the present invention in an embodiment of the present invention. Detailed Implementation

[0034] The invention will now be further described in conjunction with the accompanying drawings and specific embodiments. The improved neural network model used in this invention is a multi-machine learning model that combines a Fisher criterion classifier with a traditional neural network model in pattern recognition. This model is a specific modification of the model in the wind power steel pipe pile construction positioning scenario of this application. Specifically, it is a cascaded neural network model that processes and fuses image features and then inputs them into the improved neural network model.

[0035] The wind turbine steel pipe pile precise positioning construction model of the present invention assists in decision-making on the installation and construction positioning of offshore wind power equipment. It belongs to the offshore wind power equipment installation and wind farm construction, and is applied to the wind energy equipment installation industry, and therefore belongs to the wind energy industry.

[0036] In a first aspect of the present invention, a method for precise positioning and construction of wind turbine steel pipe piles based on BIM pattern recognition technology is provided, the method comprising:

[0037] S1. Construct a BIM model of the wind power steel pipe pile construction scene, retrieve the steel pipe pile parameter data in the BIM model, obtain the primitive features of the construction scene, use high-precision satellite to obtain real image data of the construction scene, and obtain the accurate positioning method of wind power steel pipe piles calibrated by technical personnel for the construction scene.

[0038] S2. The construction scene primitive features and the real image data of the construction scene are processed using image positioning fusion technology to obtain the precise positioning image features of the wind power steel pipe pile;

[0039] S3. Construct a construction model for the precise positioning of wind power steel pipe piles using the image features of the precise positioning of the wind power steel pipe piles, the parameter data of the steel pipe piles, and the calibrated precise positioning method of the wind power steel pipe piles;

[0040] S4. Based on the aforementioned wind power steel pipe pile precise positioning construction model, output an adaptive wind power steel pipe pile precise positioning method for different construction scenarios to assist construction management personnel in making judgments on precise positioning construction.

[0041] In this application, to improve the scientific rigor of wind turbine pile installation, the technical personnel applied BIM technology, already mature in the civil engineering field, to the wind turbine pile installation scenario. BIM typically refers to Building Information Modeling. The construction of a BIM model for a wind turbine pile installation scenario is a systematic digital modeling process, encompassing the entire lifecycle of data integration from geological survey to pile foundation completion. The construction of the BIM model follows the general operational form of building an information model after collecting relevant data in the field, such as soil layer distribution, rock layer depth, and bearing capacity parameters. For offshore facility installation, it also includes water depth, wave conditions, and wind speed, simulating wind turbine pile construction.

[0042] Furthermore, the steel pipe pile parameter data includes pile length, wall thickness, and anti-corrosion coating thickness.

[0043] The length, wall thickness, and anti-corrosion coating thickness of wind turbine steel pipe piles are strongly correlated with the selection of the offshore installation location, and these three factors jointly determine the bearing capacity, durability, and economy of the pile foundation. Pile length is mainly constrained by water depth and geological conditions—the greater the water depth and the deeper the bearing stratum, the longer the required pile length. Wall thickness is positively correlated with environmental loads such as wave force and ocean current impact; the farther from shore, the deeper the water, and the harsher the sea conditions, the thicker the wall needs to be to resist greater horizontal loads and fatigue stress. Anti-corrosion coating thickness is closely related to the corrosive environment of seawater, such as salinity, temperature, and current velocity; the greater the distance from shore, the greater the water depth, and the stronger the seawater corrosivity, the thicker the anti-corrosion coating needs to be to extend its service life. Therefore, the structural dimensions and anti-corrosion requirements of the pile foundation determine the selection of the offshore location from multiple perspectives. To ensure the safety and economy of wind turbine pile installation and construction, and to ensure the simplicity and accuracy of subsequent model data processing, the above three characteristic parameters are selected as one of the model's input parameters to control the accuracy of the model's positioning method output.

[0044] In this embodiment, spatial feature vectors are constructed using the three types of parameter feature values ​​described above to achieve the model's applicability to data processing. One of the constructed spatial feature vectors is... This indicates that the length of the offshore wind power steel pipe pile is 110 meters, the wall thickness is 80 millimeters, and the average thickness of the anti-corrosion coating is 600 μm. This is one of the spatial feature vectors used for model training in this invention. The dimension is removed to facilitate the calculation of the model later. At the same time, the spatial feature vector used to train the model in this invention is not limited to this one spatial feature vector.

[0045] Furthermore, the construction scene primitive features are obtained by comprehensively processing the seabed soil layer distribution data, rock layer thickness data, rock layer bearing capacity data, seawater depth, sea surface wind speed data, and sea wave data of each image unit in the BIM model. The calculation formula for the feature processing of its image unit (x, y) is as follows:

[0046] In the formula, Let be the grayscale value feature of image unit (x, y), and n be the number of pixels contained in image unit (x, y), which is generally taken as 9. This represents the seabed soil layer distribution data for the i-th pixel within an image unit (x, y). The seabed soil layer distribution data is the average value of the depths of each soil layer at a fixed depth when a typical subsea wind turbine pile is driven. The image unit (x, y) contains the rock layer thickness data corresponding to the i-th pixel. This represents the rock bearing capacity data corresponding to the i-th pixel within the image unit (x, y). Since the bearing capacity of seabed rocks is generally measured in MPa, and 1 MPa is... N / square meter, to prevent the model from generalizing the input data and achieving a large bias, The parameters are reduced using a logarithmic function. Let (x, y) be the seawater depth containing the i-th pixel in the image unit. The image unit (x, y) contains the sea surface wind speed data for the i-th pixel. This application provides sea wave data containing the i-th pixel within an image unit (x, y). The coefficients of each parameter in this application are obtained by experts using the Analytic Hierarchy Process (AHP) to score the weights of each parameter influencing the installation location of offshore wind turbine piles. Subsequently, the calculated comprehensive parameter data for the image unit undergoes standardized image feature processing. Data standardization is a fundamental step in machine learning, data analysis, and engineering calculations. Its core purpose is to eliminate dimensional differences and transform features of different scales and units into a unified, comparable range. The standardization method used in this application is Logistic function standardization. The Logistic function, when used for standardization, is a non-linear transformation that maps any real number to the (0, 1) interval, ensuring that the standardized image feature values ​​are within a balanced range. This creates a more efficient and accurate foundation for subsequent fusion of high-precision satellite-acquired real-world image data of the construction scene to achieve feature changes, enabling the comprehensive parameter data to express grayscale features in the image unit (x, y). The obtained construction scene primitive features can both represent the adaptability of wind power steel pipe pile installation within the image unit and facilitate subsequent model processing of image features. In this embodiment, all parameters have undergone dimensionless processing to facilitate subsequent calculations and model processing.

[0047] In this application, since the BIM model only contains identifiers for various data, in order to enable subsequent processing of the scene-based neural network model, construction scene primitive features are obtained from the data in the BIM model that are highly important to the installation location of the wind turbine steel pipe piles. The construction scene in this application is composed of multiple image units, which are combined to form the aforementioned construction scene primitive features. A simple 4x4 construction scene primitive feature is expressed as a spatial feature vector as follows:

[0048] Among them The same applies to other cases, and will not be elaborated upon here.

[0049] Furthermore, the high-precision satellite image data of the construction scene is RGB three-channel image data. The image positioning and fusion technology first comprehensively processes the pixel channel image data of the image unit corresponding to the primitive features of the construction scene. The calculation formula is as follows:

[0050] In the formula, For each image unit (x, y) corresponding to the primitive features of the construction scene, n represents the real image data of the construction scene, where n is the number of pixels contained in the image unit (x, y). Let be the grayscale feature value of the R channel of the i-th pixel within the image unit (x, y). Let G be the grayscale feature value of the i-th pixel in the image unit (x, y) in channel G. Let be the grayscale feature value of the B channel of the i-th pixel within the image unit (x, y). Since wind turbine steel pipe piles are often installed in environments far from the coastline, which are clean and open sea areas, blue light penetrates the deepest and contains the most information. Therefore, the weight value of the blue light channel in the three channels is set to a maximum of 0.6, and the corresponding image channels for the region are 0.3 and 0.1, respectively.

[0051] In this embodiment, the selection of image feature data corresponds to the pixel coordinates of the construction scene primitive features. Simultaneously, the features of a region image unit are used to provide a panoramic display of the wind turbine steel pipe pile installation construction scene. This reduces the pressure on subsequent model data processing and, more importantly, performs image localization fusion on the image features, making the acquired image features more representative of the subsequent location selection of the wind turbine steel pipe piles. In this embodiment, a simple spatial feature vector representing a 4x4 image unit in the real construction scene image data used for image localization fusion is:

[0052] The representation of each image unit, such as in the spatial feature vector , This represents the spatial feature vector value after processing the three-channel grayscale values.

[0053] Furthermore, the image positioning fusion technology vertically stitches together the processed construction scene primitive features and the real image data of the construction scene to obtain the precise positioning image features of the wind power steel pipe pile.

[0054] The spatial feature vector obtained after image localization fusion technology in this embodiment is:

[0055] Furthermore, the wind power steel pipe pile precise positioning construction model adopts a multi-cascaded classification machine learning model.

[0056] Multiple classifier cascades are an efficient hierarchical decision architecture that achieves progressive classification from coarse to fine and from simple to complex by sequentially combining multiple classifiers. The core idea is to use simple and fast classifiers to filter out a large number of negative samples, allowing complex and sophisticated classifiers to focus on processing a small number of difficult samples.

[0057] Furthermore, the multi-cascaded classification machine learning model is obtained by cascading a Fisher criterion classifier into a neural network model. The numerical values ​​obtained by inputting the precise positioning image features of the wind turbine steel pipe pile into the Fisher criterion classifier are fused with the steel pipe pile parameter data and input into the activation function of the neural network model. The calculation formula for the Fisher criterion classifier is as follows:

[0058] In the formula, The image features are used for precise positioning of the wind turbine steel pipe piles. The output is the feature value fused with the parameter data of the steel pipe pile. The normal vector perpendicular to the hyperplane is obtained by training the feature values ​​fused from the precise positioning image features of the wind turbine steel pipe pile and the output parameter data of the steel pipe pile.

[0059] Single strong classifiers, such as Fisher's criterion, SVM, and deep neural networks, treat all samples equally, resulting in a large number of simple samples consuming the same computational resources as difficult samples. Cascaded classification schemes break down the classification task into multiple stages, allowing 95% of simple negative samples to be quickly rejected in the first few layers. Since wind turbine steel pipe pile installation takes place against an ocean background, a simple Fisher's criterion classifier can quickly filter out features with a strong ocean background, thus adapting to the limited computing power of the equipment. This enables efficient and accurate positioning applications in scenarios with simple backgrounds but diverse targets and high real-time requirements, such as precise positioning and identification of wind turbine steel pipe pile installation. See the diagram for the principle of the multi-cascaded classification machine learning model. Figure 4 As shown.

[0060] Furthermore, the activation function of the neural network model is:

[0061] In the formula, L is the activation function value, and L is the feature value obtained after the input of the precise positioning image features of the wind power steel pipe pile into the Fisher criterion classifier and the feature value obtained by fusing and splicing the steel pipe pile parameter data, which is then linearly transformed by weights and biases.

[0062] In this embodiment, the final output value of the neural network is used to determine the precise positioning method of the wind power steel pipe pile. If the final output value of the neural network is 0.23, the precise positioning method of the wind power steel pipe pile is determined to be (44, 235), which is the coordinate of the precise positioning of the wind power steel pipe pile. The above situation is the corresponding result obtained after training the model in this application. This application is not limited to this result in determining the precise positioning method of the wind power steel pipe pile.

[0063] Since activation functions play a crucial role in neural networks, they not only provide the network with nonlinear mapping capabilities but also affect the network's performance, convergence speed, and generalization ability. The personal information recovery model in this application is constructed using the most basic activation functions.

[0064] A neural network is a computational model inspired by biological nervous systems. It consists of multiple layers, including input layers, hidden layers, and output layers, with each layer containing several neurons. Each neuron receives a weighted sum of the outputs from the previous layer, adds a bias, and then performs a non-linear transformation through an activation function to produce the output. This structure gives neural networks strong adaptability, enabling them to automatically adjust the connection weights between neurons through training, gradually approximating the complex functional relationship between input and output.

[0065] Activation functions are core components of neural networks. By introducing nonlinear transformations, they enable networks to overcome the limitations of linear models and learn and fit complex data distributions and mapping relationships. In constructing the personal information recovery model in this application, the selection of the activation function directly affects the model's expressive power, training convergence speed, and final generalization performance.

[0066] In model design, it is necessary to select an appropriate activation function based on factors such as task characteristics, data properties, and computational resources. For example, the Sigmoid function can compress the output to the (0,1) interval, making it suitable for probability-based output scenarios; ReLU and its variants can alleviate the vanishing gradient problem and improve training efficiency; and some activation functions exhibit a better balance between gradient and accuracy in certain deep networks. Activation functions not only endow networks with nonlinear modeling capabilities but also influence the stability of gradient propagation and the convergence behavior of model training.

[0067] Therefore, in the construction of this model, the activation function will be systematically selected based on the hierarchical characteristics of the network architecture, the inherent pattern of data distribution, and the dynamic requirements of the training process, so as to ensure that the model has strong representation ability, efficient training characteristics, and good generalization performance in the output of the wind power steel pipe pile precision positioning method.

[0068] A precise positioning construction system for wind power steel pipe piles based on BIM pattern recognition technology is also provided. This system implements a precise positioning construction method for wind power steel pipe piles based on BIM pattern recognition technology. The system includes a construction scene BIM model construction module, a steel pipe pile parameter data acquisition module, a construction scene primitive feature processing acquisition module, a high-precision satellite scene image acquisition module, a wind power steel pipe pile precise positioning construction model construction module, and a wind power steel pipe pile construction positioning analysis module.

[0069] The construction scenario BIM model building module is used to build a BIM model of the wind power steel pipe pile construction scenario and retrieve the steel pipe pile parameter data from the BIM model.

[0070] The steel pipe pile parameter data acquisition module is used to acquire the steel pipe pile parameter data in the BIM model of the wind power steel pipe pile construction scenario.

[0071] The construction scene primitive feature processing and acquisition module is used to acquire construction scene primitive features.

[0072] The high-precision satellite scene image acquisition module is used to acquire real image data of the construction scene using high-precision satellites.

[0073] The wind power steel pipe pile precise positioning construction model construction module: processes the construction scene primitive features and the real image data of the construction scene using image positioning fusion technology to obtain the precise positioning image features of the wind power steel pipe pile; and constructs the precise positioning construction model of the wind power steel pipe pile using the precise positioning image features of the wind power steel pipe pile, the steel pipe pile parameter data, and the calibrated precise positioning method of the wind power steel pipe pile.

[0074] The wind power steel pipe pile construction positioning analysis module: Based on the wind power steel pipe pile precise positioning construction model, it outputs an adaptive wind power steel pipe pile precise positioning method for different construction scenarios, assisting construction managers in making judgments on precise positioning construction.

[0075] Furthermore, the wind power steel pipe pile precise positioning construction model adopts a multi-cascaded classification machine learning model.

[0076] Multiple classifier cascades are an efficient hierarchical decision architecture that achieves progressive classification from coarse to fine and from simple to complex by sequentially combining multiple classifiers. Its core idea is to use simple and fast classifiers to filter out a large number of negative samples, allowing complex and sophisticated classifiers to focus on processing a small number of difficult samples.

[0077] The multi-cascaded classification machine learning model is obtained by cascading a Fisher criterion classifier into a neural network model. The numerical values ​​obtained by inputting the precise positioning image features of the wind turbine steel pipe pile into the Fisher criterion classifier are fused with the steel pipe pile parameter data and input into the activation function of the neural network model. The calculation formula for the Fisher criterion classifier is as follows:

[0078] In the formula, The image features are used for precise positioning of the wind turbine steel pipe piles. The output is the feature value fused with the parameter data of the steel pipe pile. The normal vector perpendicular to the hyperplane is obtained by training the feature values ​​fused from the precise positioning image features of the wind turbine steel pipe pile and the output parameter data of the steel pipe pile.

[0079] Single strong classifiers, such as Fisher's criterion, SVM, and deep neural networks, treat all samples "equally," causing many simple samples to consume the same computational resources as difficult samples. Cascaded solutions break down the classification task into multiple stages, allowing 95% of simple negative samples to be quickly rejected in the first few layers. Since wind turbine steel pipe pile installations occur against an ocean background, a simple Fisher's criterion classifier can quickly filter out features with a strong ocean background, thus adapting to the limited computing power of the equipment. This enables efficient and accurate positioning applications in scenarios with simple backgrounds but diverse targets and high real-time requirements, such as precise positioning and identification of wind turbine steel pipe pile installations.

[0080] The activation function of the neural network model is:

[0081] In the formula, L is the activation function value, and L is the feature value obtained after the input of the precise positioning image features of the wind power steel pipe pile into the Fisher criterion classifier and the feature value obtained by fusing and splicing the steel pipe pile parameter data, which is then linearly transformed by weights and biases.

[0082] Therefore, the beneficial effects of this invention are as follows: It retrieves steel pipe pile parameter data from the constructed BIM model, obtains construction scene primitive features and real-world image data of the construction scene acquired through high-precision satellite imagery, and combines these features with image positioning fusion technology to obtain precise positioning image features for wind power steel pipe piles. Using these precise positioning image features, steel pipe pile parameter data, and a pre-defined precise positioning method, a precise positioning construction model for wind power steel pipe piles is constructed. An image recognition algorithm model is then used to process the data to obtain an instantaneous precise positioning method for wind power steel pipe piles. This invention combines BIM-constructed images and high-precision satellite-acquired images, and utilizes an image recognition cascade algorithm model adapted to the wind power steel pipe pile construction scenario to output an instantaneous precise positioning method for wind power steel pipe piles. This assists construction managers in making precise positioning decisions during construction. This not only helps reduce positioning construction risks but also improves the accuracy of construction positioning and enhances the intelligence level of wind power steel pipe pile construction.

[0083] Of course, it is understood that each embodiment of the present invention can achieve one of the effects individually, and the combination of multiple embodiments of the present invention can achieve all the above effects. However, it is not required that each embodiment of the present invention achieve all the above advantages and effects, because each embodiment of the present invention can constitute a separate technical solution and make one or more contributions to the prior art. Each embodiment of the present invention is not affected, and the solution can still be implemented even if one embodiment is deleted.

[0084] For any module structures not specifically defined in this invention, the existing technical descriptions shall prevail. The prior art mentioned in the foregoing background and specific embodiments sections can be considered part of this invention and used to understand the meaning of certain technical features or parameters.

Claims

1. A method for precise positioning and construction of wind power steel pipe piles based on BIM pattern recognition technology, characterized in that, The method includes: S1. Construct a BIM model of the wind power steel pipe pile construction scene, retrieve the steel pipe pile parameter data in the BIM model, obtain the primitive features of the construction scene, use high-precision satellite to obtain real image data of the construction scene, and obtain the accurate positioning method of wind power steel pipe piles calibrated by technical personnel for the construction scene. S2. The construction scene primitive features and the real image data of the construction scene are processed using image positioning fusion technology to obtain the precise positioning image features of the wind power steel pipe pile; S3. Construct a construction model for the precise positioning of wind power steel pipe piles using the image features of the precise positioning of the wind power steel pipe piles, the parameter data of the steel pipe piles, and the calibrated precise positioning method of the wind power steel pipe piles; S4. Based on the aforementioned wind power steel pipe pile precise positioning construction model, output an adaptive wind power steel pipe pile precise positioning method for different construction scenarios to assist construction management personnel in making judgments on precise positioning construction.

2. The method for precise positioning and construction of wind power steel pipe piles based on BIM pattern recognition technology as described in claim 1, characterized in that: The parameters of the steel pipe pile include pile length, wall thickness, and anti-corrosion coating thickness.

3. The method for precise positioning and construction of wind power steel pipe piles based on BIM pattern recognition technology as described in claim 1, characterized in that: The construction scene graphic features are obtained by processing and integrating the seabed soil distribution data, rock layer thickness data, rock layer bearing capacity data, seawater depth, sea surface wind speed data, and sea wave data of each image unit in the BIM model.

4. A method for precise positioning and construction of wind power steel pipe piles based on BIM pattern recognition technology as described in claim 1 or 3, characterized in that: The high-precision satellite image data of the construction scene is RGB three-channel image data. The image positioning and fusion technology first performs comprehensive processing on the pixel channel image data of the image unit corresponding to the primitive features of the construction scene.

5. The method for precise positioning and construction of wind power steel pipe piles based on BIM pattern recognition technology as described in claim 4, characterized in that: The image positioning fusion technology vertically stitches together the processed construction scene primitive features and the real image data of the construction scene to obtain the precise positioning image features of the wind power steel pipe pile.

6. A method for precise positioning and construction of wind power steel pipe piles based on BIM pattern recognition technology as described in claim 1, 2, or 3, characterized in that: The precise positioning construction model for wind power steel pipe piles adopts a multi-cascaded classification machine learning model.

7. The method for precise positioning and construction of wind power steel pipe piles based on BIM pattern recognition technology as described in claim 6, characterized in that: The multi-cascaded classification machine learning model is obtained by cascading a Fisher criterion classifier into a neural network model. The numerical values ​​obtained by inputting the precise positioning image features of the wind power steel pipe pile into the Fisher criterion classifier are fused with the parameter data of the steel pipe pile and input into the activation function of the neural network model.

8. A wind power steel pipe pile precise positioning construction system based on BIM pattern recognition technology. This system implements a method for precise positioning construction of wind power steel pipe piles based on BIM pattern recognition technology. The system includes a construction scene BIM model construction module, a steel pipe pile parameter data acquisition module, a construction scene primitive feature processing acquisition module, a high-precision satellite scene image acquisition module, a wind power steel pipe pile precise positioning construction model construction module, and a wind power steel pipe pile construction positioning analysis module. Its features are: The construction scenario BIM model building module is used to build a BIM model of the wind power steel pipe pile construction scenario and retrieve the steel pipe pile parameter data from the BIM model. The steel pipe pile parameter data acquisition module is used to acquire the steel pipe pile parameter data in the BIM model of the wind power steel pipe pile construction scenario. The construction scene primitive feature processing and acquisition module is used to acquire construction scene primitive features. The high-precision satellite scene image acquisition module is used to acquire real image data of the construction scene using high-precision satellites. The wind power steel pipe pile precise positioning construction model construction module: processes the construction scene primitive features and the real image data of the construction scene using image positioning fusion technology to obtain the precise positioning image features of the wind power steel pipe pile; and constructs the precise positioning construction model of the wind power steel pipe pile using the precise positioning image features of the wind power steel pipe pile, the steel pipe pile parameter data, and the calibrated precise positioning method of the wind power steel pipe pile. The wind power steel pipe pile construction positioning analysis module: Based on the wind power steel pipe pile precise positioning construction model, it outputs an adaptive wind power steel pipe pile precise positioning method for different construction scenarios, assisting construction managers in making judgments on precise positioning construction.

9. The wind power steel pipe pile precise positioning construction system based on BIM pattern recognition technology as described in claim 8, characterized in that: The precise positioning construction model for wind power steel pipe piles adopts a multi-cascaded classification machine learning model. The multi-cascaded classification machine learning model is obtained by cascading a Fisher criterion classifier into a neural network model. The numerical values ​​obtained by inputting the precise positioning image features of the wind power steel pipe pile into the Fisher criterion classifier are fused with the parameter data of the steel pipe pile and input into the activation function of the neural network model.

Images