Artificial Intelligence Power Grid Building Imaging Method and Engine Based on Large Language Model

Abstract

The present invention provides an artificial intelligence power grid building imaging method and engine based on a large language model, which relates to the technical field of power grid building imaging. The method includes: using the large language model to read project information and power grid building design standards to obtain building requirements; using the large language model to detect and correct outliers in the data; using a three-dimensional reconstruction model to gradually infer and restore the three-dimensional shape of the power grid building object according to the two-dimensional CAD image data after data preprocessing, wherein the building elements are identified based on the large language model to determine whether the building elements meet the current scene and construction conditions. The artificial intelligence power grid building imaging engine includes a three-dimensional image reconstruction module and a large language model processing module. The present invention can achieve accurate modeling of power grid buildings from two dimensions to three dimensions, improve data quality and modeling efficiency, and optimize building layouts.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid building imaging, and in particular to an artificial intelligence power grid building imaging method and engine based on a large language model. Background Art

[0002] In the field of traditional power construction, there are many drawbacks in the drawing and application of building structure drawings. For a long time, the two-dimensional drawing software AutoCAD has dominated the process of drawing building structure drawings, and its drawing results are mainly stored in formats such as DWG and DXF. These formats are significantly different from the formats used by three-dimensional modeling software such as Revit, SketchUp, 3dsMax, etc., resulting in obstacles to data sharing and interaction between the two, and a gap between two-dimensional design and three-dimensional space.

[0003] This gap has brought many problems in practical applications. On the one hand, during the design phase, designers design based on two-dimensional drawings, which may not fully consider the actual situation of the building in three-dimensional space, resulting in defects in the design plan in three-dimensional space. For example, some parts of the building may collide or conflict in three-dimensional space, affecting the structural safety and construction feasibility of the building. On the other hand, during the construction phase, construction workers mainly refer to two-dimensional drawings for construction. Due to the lack of intuitive guidance from three-dimensional models, construction workers may find it difficult to accurately understand the design intent, which may easily lead to construction errors or substandard construction quality.

[0004] In addition, the separation of two-dimensional design and three-dimensional space also brings many hidden dangers to the safety assessment of power buildings. For example, in terms of dual power supply settings, design and construction based only on two-dimensional drawings may not accurately determine whether the dual power supply settings are reasonable and whether they comply with relevant standards and specifications, thus affecting the reliability and safety of power supply. In terms of the construction space of emergency power supply equipment, due to the lack of intuitive display of three-dimensional models, it may not be possible to accurately reserve enough space, resulting in the inability to install or operate emergency power supply equipment normally, affecting the emergency power supply capacity of the building. In terms of building space layout, unreasonable layout may lead to the inability to ensure the safety of grounding equipment, such as the distance between the grounding equipment and other parts of the building does not meet safety requirements, increasing the risk of electric shock or short circuit. At the same time, problems such as repeated grounding and equipotential bonding may also occur, affecting the overall safety and stability of power buildings.

[0005] In summary, the separation of two-dimensional design and three-dimensional space in the traditional power construction field not only affects the efficiency and quality of design and construction, but also brings many hidden dangers to the safety assessment of power buildings. There is an urgent need for a method that can effectively integrate two-dimensional design and three-dimensional modeling to solve these problems. Summary of the invention

[0006] The present invention aims to solve at least one of the above technical problems existing in the prior art.

[0007] To this end, a first aspect of the present invention provides an artificial intelligence power grid building imaging method based on a large language model.

[0008] A second aspect of the present invention provides an artificial intelligence power grid building imaging engine based on a large language model.

[0009] The present invention provides an artificial intelligence power grid building imaging method based on a large language model, including:

[0010] Obtaining basic data, where the basic data at least includes two-dimensional CAD image data for describing a power grid building and project information; using a large language model to read the project information and power grid building design standards to obtain building requirements;

[0011] Performing data preprocessing on the two-dimensional CAD image data; the data preprocessing includes format conversion, data cleaning, and data augmentation; wherein, in data cleaning, a large language model is used to detect and correct outliers in the data;

[0012] Constructing a three-dimensional reconstruction model, where the three-dimensional reconstruction model gradually infers and restores the three-dimensional shape of the power grid building object according to the two-dimensional CAD image data after data preprocessing; wherein, the power grid building site information and two-dimensional feature information of various building units in the site are obtained through feature extraction, and the building unit is a power grid functional unit composed of at least one electrical device; classifying and labeling the extracted two-dimensional feature information through semantic segmentation, and, based on a large language model, identifying building elements to determine whether the building elements meet the current scenario and construction conditions;

[0013] Selecting a matching basic building unit model from a prefabricated model database according to the two-dimensional feature information; placing the basic building unit model into the three-dimensional coordinates of the power grid building site according to the coordinate information and in combination with the layout information of the plan view; and based on a six-view rendering engine, establishing a mapping relationship between the two-dimensional image and its corresponding basic building unit model, and performing a three-dimensional imaging and modeling task in space.

[0014] According to the above technical solution of the present invention, the artificial intelligence power grid building imaging method based on a large language model may further have the following additional technical features:

[0015] In the above technical solution, the two-dimensional CAD image data records the power grid building site information and image information of various building units in the site, and the image information includes the two-dimensional coordinates, geometric shapes, and surface features of the power grid building;

[0016] The project information includes at least one of the project name, project type, special situation description file, risk description file, CAD file description list, building review file list, and voxel points;

[0017] The special situation description file records the special information of the designated building, and the special information includes the location, size, and angle of the designated building;

[0018] The risk description file is used to restrict some buildings where construction is prohibited;

[0019] The CAD file description list is used to describe the function of each two-dimensional CAD image data;

[0020] The building review file list is used to supplement and verify building parameters.

[0021] In the above technical solution, the large language model adopts the large language model specialized for CPM-Bee architecture; the use of the large language model to detect and correct outliers in the data includes:

[0022] The large language model specialized for CPM-Bee architecture identifies the height information, distance information, and angle information in the two-dimensional CAD image data that deviate from the required range according to its own knowledge base and the building review file list, marks and corrects the deviated data, and restricts it within the required range;

[0023] For the data information of the designated building recorded in the special situation description file but not meeting the requirements of the knowledge base of the large language model specialized for CPM-Bee architecture and the building review file list, the data is detected and corrected based on the required range specified in the special situation description file.

[0024] In the above technical solution, the grid building site information and the two-dimensional feature information of various power equipment in the site include the planar building position feature information, planar building unit feature information, and building unit connection relationship feature information;

[0025] The obtaining of the grid building site information and the two-dimensional feature information of various building units in the site through feature extraction includes:

[0026] Performing convolution processing on the preprocessed two-dimensional CAD image data using multiple two-dimensional convolutional layers; among them, each two-dimensional convolutional layer is equipped with a convolutional kernel of a set size and stride, and these convolutional kernels slide in the two-dimensional space to extract different features in the two-dimensional CAD image data; a non-linear activation function is added after each two-dimensional convolutional layer to introduce non-linear transformation and output the feature map to be processed;

[0027] Perform a max pooling operation on the feature map to be processed, select the maximum eigenvalue within each pooling window as the output, and generate a local feature vector;

[0028] Pass the local feature vector through a fully connected layer for feature fusion and transformation, learn the connection relationships and layout rules between different building units, and combine the local feature vectors into a global feature vector.

[0029] In the above technical solution, the classification and annotation of the two-dimensional feature information through semantic segmentation includes:

[0030] In the decoder, use a transposed convolutional layer to perform an upsampling operation on the global feature vector, expand the low-resolution feature map into a higher-resolution feature map, and restore the original spatial resolution of the data with the decoded feature map;

[0031] For each pixel or voxel in the decoded feature map, calculate the probability distribution of its belonging to each building unit type through the Softmax function, and obtain the type label most likely to belong to each pixel or voxel, including:

[0032] For the global feature vector output by the fully connected layer without passing through the activation function , n is the total number of classes, is the score of the sample belonging to the class, and the calculation process of the Softmax function is:

[0033]

[0034] represents the conditional probability symbol, y is the true class label, refers to the probability that the output class is t under the condition of the given input feature x; r represents the class index (traversing all classes), ∈{1,2,…,n}; represents the unactivated feature vector the nth element of, and this score is calculated by the last layer of the neural network; the total number of classes n, that is, the total number of elements of the input vector, adopts the number of building unit types defined by the electric power industry standard, which determines the number of probabilities that the Softmax function needs to calculate;

[0035] According to the classification result, generate a corresponding segmentation mask for each class, and display the distribution of building units in the three-dimensional space through the segmentation mask;

[0036] In the three-dimensional reconstruction model, classify the planar building position feature information, planar building unit feature information, and building unit connection relationship feature information respectively.

[0037] In the above technical solution, the classification and annotation of the two-dimensional feature information through semantic segmentation further includes:

[0038] Introduce skip connections in the decoder to fuse the feature maps at the corresponding levels in the encoder with the upsampled feature maps in the decoder.

[0039] In the above technical solution, the identification of building elements based on the large language model and the judgment of whether the building elements meet the current scene and construction conditions include:

[0040] The large language model judges whether the building elements meet the current scene according to the building requirements read based on the project information for the planar building position feature information, that is, judges whether the positions of each building unit are reasonable; marks and records the building elements that do not meet the current scene, and can directly output the error information for designers' reference;

[0041] The large language model judges whether the current building unit meets the construction requirements of the corresponding building unit based on the building requirements read based on the project information for the planar building unit feature information;

[0042] After completing the determination of the current scene and construction conditions, the large language model supplements the detailed drawing information of the building unit according to the precautions entered in the knowledge base and project information.

[0043] In the above technical solution, select a matching basic building unit model in the prefabricated model database according to the two-dimensional feature information, place the basic building unit model into the three-dimensional coordinates of the power grid building site according to the coordinate information and in combination with the layout information of the plan view; and based on the six-view rendering engine, establish a mapping relationship between the two-dimensional image and its corresponding three-dimensional voxel point model, and perform the spatial three-dimensional imaging and modeling task, including:

[0044] Select a matching basic building unit model in the prefabricated model database according to the extracted planar building unit feature information; wherein, the prefabricated model database includes three-dimensional voxel point models of power grid functional units belonging to different types of power stations, and any three-dimensional voxel point model is defined as a basic building unit model;

[0045] Place the basic building unit model into the three-dimensional coordinates of the power grid building site according to the coordinate information and in combination with the layout information of the plan view;

[0046] After the six - view rendering engine receives the processed input image sequence, it first encodes these image sequences into a series of hidden states. These hidden states capture the temporal dependencies in the image sequence and contain spatial features; subsequently, these hidden states are transmitted to the decoder part. When processing these hidden states, the decoder first applies 3D convolution operations to extract and fuse spatio - temporal features. Immediately afterwards, non - linear characteristics are introduced through a non - linear activation function (sigmoid), enabling the model to learn more complex mapping relationships. The resolution of the hidden states is gradually increased until the target output resolution is reached.

[0047] In the above - mentioned technical solution, it further includes:

[0048] Perform edge smoothing on the generated segmentation mask to remove possible isolated noise points and small - area misclassified regions; for the possible holes in the segmentation mask, use a hole - filling algorithm based on neighborhood information for repair;

[0049] Present the final semantic segmentation result to power designers in an intuitive 3D visualization form; overlay the segmentation mask on the original 3D imaging data, and identify different power equipment categories with different colors or textures.

[0050] An artificial - intelligence power - grid building imaging engine based on a large - language model provided by the present invention is applied to the artificial - intelligence power - grid building imaging method based on a large - language model as described in any of the above - mentioned technical solutions. The artificial - intelligence power - grid building imaging engine includes: a three - dimensional image reconstruction module and a large - language model processing module; the three - dimensional image reconstruction module receives two - dimensional CAD image data for describing a power - grid building; the large - language model processing module is used to receive project information and use the large - language model to read the project information and power - grid building design standards to obtain building requirements;

[0051] The three - dimensional image reconstruction module includes:

[0052] A data pre - processing unit that performs data pre - processing on the two - dimensional CAD image data; the data pre - processing includes format conversion, data cleaning, and data augmentation; among them, when performing data cleaning, the large - language model is called to detect and correct outliers in the data;

[0053] A feature extraction module for obtaining two - dimensional feature information of the power - grid building site information and various power equipment in the site through feature extraction;

[0054] A semantic segmentation module for classifying and labeling the two - dimensional feature information through semantic segmentation; the large - language model processing module uses the large - language model to identify building elements based on the processing results of the feature extraction module and the semantic segmentation module, and determines whether the building elements meet the current scene and construction conditions;

[0055] A three-dimensional imaging module selects a matching basic building unit model in a prefabricated model database according to the two-dimensional feature information; places the basic building unit model into the three-dimensional coordinates of the power grid building site according to the coordinate information and in combination with the layout information of the plan view; and based on a six-view rendering engine, establishes a mapping relationship between the two-dimensional image and its corresponding basic building unit model, and executes the spatial three-dimensional imaging and modeling task.

[0056] In summary, due to the adoption of the above technical features, the beneficial effects of the present invention are as follows:

[0057] The present invention integrates two-dimensional CAD data and a large language model to achieve accurate three-dimensional modeling of power grid buildings, improve data quality and modeling efficiency, optimize building layouts, enhance safety and compliance, break down data barriers between software, reduce manual intervention, and promote the intelligent development of power construction.

[0058] The additional aspects and advantages of the present invention will become apparent in the following description section or be learned through the practice of the present invention. Description of the Drawings

[0059] The above and / or additional aspects and advantages of the present invention will become apparent and be readily understood from the description of the embodiments in conjunction with the following drawings, wherein:

[0060] Figure 1 is a flowchart of an artificial intelligence power grid building imaging method based on a large language model according to an embodiment of the present invention;

[0061] Figure 2 is a schematic diagram of the process of a 2D convolutional neural network layer converting an input image into a feature map to be processed according to an embodiment of the present invention;

[0062] Figure 3 is a schematic diagram of a 3D-LSTM network structure according to an embodiment of the present invention;

[0063] Figure 4 is a schematic diagram of the preliminary 3D-LSTM drawing effect according to an embodiment of the present invention;

[0064] Figure 5 is a schematic diagram of the GRU unit-assisted drawing effect according to an embodiment of the present invention;

[0065] Figure 6 is a schematic diagram of the effect after mean square error correction according to an embodiment of the present invention;

[0066] Figure 7 is a schematic diagram of the effect after smoothing processing and hole filling according to an embodiment of the present invention. Detailed Embodiments

[0067] To better understand the above objects, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments may be combined with each other.

[0068] In the following description, numerous specific details are set forth to facilitate a thorough understanding of the present invention. However, the present invention may be practiced in other ways than those described herein. Therefore, the scope of the present invention is not limited by the specific embodiments disclosed below.

[0069] The following refers to Figures 1 to 7 to describe an artificial intelligence power grid building imaging method and engine based on a large language model according to some embodiments of the present invention.

[0070] Some embodiments of the present application provide an artificial intelligence power grid building imaging method based on a large language model.

[0071] The first embodiment of the present invention proposes an artificial intelligence power grid building imaging method based on a large language model, including the following steps S1 - S4.

[0072] S1. Obtain basic data, where the basic data at least includes two - dimensional CAD image data for describing a power grid building and project information; use the large language model to read the project information and power grid building design standards to obtain building requirements.

[0073] Specifically, the two - dimensional CAD image data records the power grid building site information and the image information of various building units in the site. The image information includes the two - dimensional coordinates, geometric shapes, and surface features of the power grid building. In some embodiments, the two - dimensional CAD image data can be two - dimensional imaging data from different channels, such as CAD files formed by lidar (LiDAR) scan data, aerial photogrammetry data, etc. These data contain rich spatial information of power facilities, such as the two - dimensional coordinates, geometric shapes, and surface features of transmission lines, substations, main workshops, dust removal devices, induced draft fan sites, and chimney flues.

[0074] The project information includes at least one of the project name, project type, special situation description file, risk description file, CAD file description list, building review file list, and voxel points; the richer the content of the project information, the more beneficial it is for the large language model to obtain clear building requirements, and then analyze and judge the rationality, safety, etc. of the project. In the present disclosure, it is described by taking the project information including all the above - mentioned information as an example. Among them, the special situation description file, risk description file, CAD file description list, and building review file list can all be input into the large language model in the form of.csv format files.

[0075] The project name is used to generally describe the current engineering project and supports multiple languages, such as Chinese, English, etc.; the project type is used to distinguish the category of the current engineering project to facilitate subsequent work such as model selection. In a specific embodiment, the project types include: thermal power project, new energy project, nuclear power project, power grid project, substation project, power transmission project, system planning, design and dispatching project, communication project, thermal engineering project, desulfurization project, construction project, environmental project, survey project, and other projects. The above project types cover the common types of power grid projects.

[0076] The special situation description file records the special information of the specified building, and the special information includes the location, size, and angle of the specified building; the risk description file is used to restrict some buildings that are prohibited from construction; the CAD file description list is used to describe the function of each two-dimensional CAD image data to facilitate the engine to model in an orderly manner; the building review file list is used to supplement and verify the building parameters. The voxel point is a cube unit in three-dimensional space and is used to represent discrete points in three-dimensional images or volume data.

[0077] The power grid building design standard refers to the current design standards and design specifications of each building unit in the power grid. By reading the power grid building design standard, the large language model can have corresponding data processing capabilities.

[0078] In a specific embodiment, the large language model performs composition analysis on the two-dimensional CAD image data according to the CAD file description list, clarifies various elements and attributes included in the drawing, and uses the CAD file description list to perform name correspondence on the currently input CAD image data to ensure the accuracy of the drawing information. Subsequently, it is sorted according to the type of CAD drawing (such as architecture, machinery, electrical, etc.) for subsequent processing.

[0079] In a specific embodiment, the large language model adopts a large language model specialized for CPM-Bee architecture. It should be noted that CPM-Bee is a fully open-source and commercially available large language model with tens of billions of parameters in Chinese and English. It adopts the Transformer autoregressive architecture and can be used for the semantic library of lightweight power grid construction documents and the regular formula judgment of building problems.

[0080] S2. Perform data preprocessing on the two-dimensional CAD image data; the data preprocessing includes format conversion, data cleaning, and data augmentation; among them, in data cleaning, the large language model is used to detect and correct outliers in the data.

[0081] In some embodiments, in step S2, two-dimensional CAD image data in various different formats are uniformly converted into a processable standard format, including unifying requirements such as drawing format and size, scale, drawing lines, fonts, symbols, etc., ensuring data consistency and compatibility for subsequent processing and analysis. At the same time, some unprocessable data information is deleted and recorded in the error log. In a specific embodiment, the standard data format standard adopts the content provided in the document "Rules for Electrical Engineering CAD Drafting of the People's Republic of China National Standard" GB / T 18135—2000. Additionally, in this process, unified conversion of length units and unified adjustment of building orientation can be performed on all two-dimensional CAD image data.

[0082] In some embodiments, due to possible influence by environmental factors during the imaging process, there may be noise points in the data, so data cleaning is required.

[0083] Specifically, the data is denoised through a filtering algorithm (Gaussian filtering) to improve the data quality and reduce the interference of noise on subsequent analysis. For each pixel point ( ) in the two-dimensional CAD image data, its value after Gaussian filtering is , and the calculation formula is:

[0084]

[0085] where, represents the Gaussian filtering at the position ( ); represents the coordinates of the original pixel value; represents the Gaussian standard deviation, the smaller it is, the weaker the smoothing effect and the more image details are retained. In this embodiment, takes the value of 0.13, and the filtering intensity is moderate; e represents the base of the natural logarithm; represents a double integral over the entire two-dimensional plane, that is, summing over every possible and values.

[0086] During the data cleaning process, the CPM-Bee building-specialized large language model identifies height information, distance information, and angle information in the two-dimensional CAD image data that deviate from the required range according to its own knowledge base and the list of building review documents, marks and corrects the deviated data, and restricts it within the required range; for the data information of the specified building that does not meet the requirements of the CPM-Bee building-specialized large language model knowledge base and the building review document list but is recorded in the special situation description file, the data is detected and corrected based on the required range specified in the special situation description file.

[0087] In some embodiments, data augmentation is performed by means of rotation, translation, scaling, cropping, etc.

[0088] Specifically, random rotation and translation operations are performed on the two-dimensional CAD image data to simulate the observation perspectives at different angles and positions, increase the diversity of the data, improve the generalization ability of the model, and enable it to better cope with various changes in the actual scenario. The data is scaled and cropped according to the preset scale range, the key areas related to power design are retained, the irrelevant background information is removed, and at the same time, the consistency of the data at different scales is ensured, which is convenient for the model to learn and extract features.

[0089] To implement the above data augmentation function, the graphic edge detection calculation formula adopted in this embodiment is gradient calculation. First, define the input image as I, where x and y are the abscissa and ordinate of the pixels in the image respectively, and the value of I(x, y) represents the gray value of the pixel. The Sobel operator consists of two 3×3 convolution kernels and is used to calculate the gradient component in the horizontal direction and the gradient component in the vertical direction , and the specific calculation formula is as follows:

[0090]

[0091] The calculation formula for the convolution operation is:

[0092]

[0093]

[0094] Among them, represents the position index coordinates of the output feature map; represents the gradient component in the horizontal direction at the pixel ; represents the gradient component in the vertical direction at the pixel (x, y); I(x + i, y + j) represents the gray value at the corresponding position (x + i, y + j) in the input image.

[0095] After calculating the gradient components, continue to calculate the gradient magnitude (or called the gradient amplitude). In this embodiment, the maximum value between the two is taken as the gradient magnitude, which is represented by the following formula:

[0096]

[0097] After calculating the gradient magnitude, judge the image edge, and the judgment method is as follows:

[0098]

[0099] Among them, represents the set threshold of the gradient magnitude. In this disclosure The value is set to 127, and pixels with a gradient magnitude greater than the threshold are marked as edge pixels: Edge(x, y) is the output binary image, where pixels with a value of 1 represent edges and pixels with a value of 0 represent non-edges.

[0100] S3. Construct a 3D reconstruction model (3D-EPowerNet model), which gradually infers and restores the 3D shape of the grid building objects based on the 2D CAD image data after data preprocessing; among them, the grid building site information and the 2D feature information of various building units in the site are obtained through feature extraction, and the building unit is a grid functional unit composed of at least one power equipment; the extracted 2D feature information is classified and labeled through semantic segmentation, and based on the large language model, the building elements are identified to determine whether the building elements meet the current scene and construction conditions.

[0101] Specifically, the grid building site information and the 2D feature information of various power equipment in the site include planar building position feature information, planar building unit feature information, and building unit connection relationship feature information; the planar building position feature information records at least the coordinate positions of all building units; the planar building unit feature information records at least the appearance shape of a single building unit; the building unit connection relationship feature information records at least the connection relationship between the connected building units.

[0102] In some embodiments, the obtaining of the grid building site information and the 2D feature information of various building units in the site through feature extraction includes the following steps S311-S313.

[0103] S311. In the 2D convolutional neural network layer (2D-CNN), multiple two-dimensional convolutional layers are used to perform convolutional processing on the preprocessed 2D CAD image data; among them, each two-dimensional convolutional layer is equipped with a convolutional kernel of a set size and stride, and these convolutional kernels slide in the two-dimensional space to extract different features in the 2D CAD image data; a non-linear activation function is added after each two-dimensional convolutional layer to introduce non-linear transformation, and the feature map to be processed is output. For example, the shallow convolutional layer can extract smaller local features, such as the detailed structure of the power supply room; while the deep convolutional layer can capture a larger range of context information, such as the overall trend and layout of the transmission line.

[0104] In a specific embodiment, the expression of the 2D convolutional neural network layer is as follows:

[0105]

[0106] Among them, represents the output feature map to be processed; represents the convolutional kernel weight matrix; Represents the input feature map, i.e., the preprocessed 2D CAD image data; The output feature map to be processed can be updated, optimized through training, and used for high-level feature extraction The number of rows of the convolutional kernel; Represents the number of columns of the convolutional kernel; Represents the row stride of the convolutional kernel; Represents the column stride of the convolutional kernel. In this design, the stride length is selected as 1; Since usually CAD drawings are monochromatic drawings, the number of channels can be set to 3. Then the process of the 2D convolutional neural network layer converting the input image into the feature map to be processed is as Figure 2 shown.

[0107] After each convolutional layer, add a non-linear activation function (ReLU) to introduce non-linear transformation, enhance the expression ability of the model, enable the network to learn more complex feature mapping relationships, and thus better fit the diverse forms and features of power facilities. The formula of the ReLU activation function is:

[0108]

[0109] where, Represents the output value after being processed by the activation function, Represents the input value. This means that when the input value is greater than 0, the output value is the input value itself; when the input value is less than or equal to 0, the output value is zero.

[0110] S312. Perform a max pooling operation on the feature map to be processed, select the maximum feature value within each pooling window as the output, and generate a local feature vector.

[0111] Specifically, perform a max pooling operation on the feature map, select the maximum feature value within each pooling window as the output, effectively reduce the spatial dimension of the feature map, reduce the amount of calculation and the number of parameters, and at the same time retain the most important feature information. This helps to prevent the occurrence of overfitting, improve the generalization performance of the model, and enable the model to focus on the learning and recognition of key features.

[0112] In a specific embodiment, the expression of the max pooling operation is:

[0113]

[0114] where, Represents the value of the local feature vector at position (i,j), i.e., the maximum value after pooling; Represents the pooling stride, i.e., the number of pixels the window slides each time, and the value range is from 1 to 3, specifically determined according to the required feature detail level; represents the number of rows of the pooling window; represents the number of columns of the pooling window; , represents selecting the maximum value of the pooling window within the given index range m ∈ [1, P], n ∈ [1, Q].

[0115] In this application, the max pooling formula is used to slide a pooling window of a fixed size on the feature map to be processed and select the maximum feature value within each window as the output, thereby reducing the spatial dimension of the data while retaining the most important feature information.

[0116] In a specific embodiment, the pooling layer downsamples the feature map output by the convolutional layer through a sliding window (such as 2×2 2×2), significantly reducing the amount of data. For example, a 1000×1000 1000×1000 transmission line floor plan may be compressed to 500×500 500×500 after pooling, significantly reducing the computational complexity of subsequent modules (such as fully connected layers, 3D-LSTM).

[0117] S313. Feature fusion and transformation are performed on the local feature vectors through a fully connected layer to learn the connection relationships and layout rules between different building units, and the local feature vectors are combined into global feature vectors.

[0118] Specifically, after multiple convolutional and pooling operations, the two-dimensional feature tensor is flattened into a one-dimensional vector, and further feature fusion and transformation are performed through a fully connected layer. The fully connected layer combines local features into a global feature representation, providing a more abstract and representative feature vector for subsequent classification or regression tasks. In this process, the model can learn the complex relationships and patterns between different power building units, such as the connection methods between different types of boilers, condensers, high and low pressure heaters, condensate pumps, main transformers, distribution devices, batteries, control panels, etc.

[0119] In a specific embodiment, the expression of the fully connected layer is as follows:

[0120]

[0121] where, represents the output vector of the fully connected layer, that is, the global feature vector, and the value range is ; represents the input vector of the fully connected layer, that is, the local feature vector, and the value range is ; represents the weight matrix, ; represents the bias vector, and the value range is ; N represents the number of elements in the output vector;

[0122] For example, assume there is a fully connected layer with an input vector = [1, 2] (dimension = 2), and the output vector needs to have 3 elements (i.e., N = 3).

[0123]

[0124] Then the output vector can be obtained through the following calculation:

[0125]

[0126] This output vector can be used as the input for the next layer.

[0127] In some embodiments, classifying and labeling the two-dimensional feature information through semantic segmentation includes the following steps S321 - S323.

[0128] S321. In the decoder, use a transposed convolutional layer to perform upsampling on the global feature vector, expanding the low-resolution feature map into a higher-resolution feature map to restore the original spatial resolution of the data with the decoded feature map.

[0129] Specifically, to restore the original spatial resolution of the data, a transposed convolutional layer is used to perform upsampling on the output of the fully connected layer. The transposed convolutional layer expands the low-resolution feature map into a high-resolution feature map by means of a transposed convolution kernel, gradually restoring it to the same or similar size as the input data for pixel-level classification and labeling. It can be understood that the encoder part corresponds to the feature extraction part, and the decoder corresponds to the content of the above transposed convolutional layer.

[0130] In some embodiments, skip connections are also introduced in the decoder to fuse the feature maps at the corresponding levels in the encoder with the upsampled feature maps in the decoder. This connection method can transmit the coarse-grained feature information extracted in the encoder to the decoder, assisting the decoder to better restore the detailed information. Especially when dealing with areas where the boundaries of power facilities are blurred or there are occlusions, it can improve the accuracy and integrity of semantic segmentation.

[0131] S322. For each pixel or voxel in the decoded feature map, calculate the probability distribution of its belonging to each building unit type through the Softmax function to obtain the type label that each pixel or voxel is most likely to belong to.

[0132] Specifically, the Softmax function converts the original scores output by the model into probability values, making the sum of the probabilities of all classes equal to 1, thereby obtaining the class label that each pixel or voxel is most likely to belong to.

[0133] In a specific embodiment, for the global feature vector output by the fully connected layer without passing through the activation function , where n is the total number of categories, is the score that the sample belongs to the th category. The calculation process of the Softmax function is as follows:

[0134]

[0135] represents the conditional probability symbol, y is the true class label, refers to the probability that the output category is t under the condition of the given input feature x; r represents the category index (traversing all categories), ∈{1,2,…,n}; represents the nth element of the unactivated feature vector , and this score is calculated by the last layer of the neural network; the total number of categories n, that is, the total number of elements of the input vector, adopts the number of building unit types defined by the power industry standard, which determines the number of probabilities that the Softmax function needs to calculate.

[0136] The following is an example to illustrate the process of step S322:

[0137] Input data: the feature of a certain pixel point in the CAD drawing;

[0138] Model unactivated feature value: =[3.5, 1.2, -0.5], corresponding to the scores that this pixel point belongs to the boiler, distribution device, and load-bearing wall in sequence;

[0139] The Softmax function calculation is as follows:

[0140]

[0141] It means that this pixel has an 88% probability of belonging to the boiler and can be used to guide the subsequent 3D model generation.

[0142] Through the above processing of the 2D CAD drawing, the information of the power grid building site and various power building units in the site can be obtained, providing a basic condition for the next three-dimensional structure processing.

[0143] S323. According to the classification results, generate corresponding segmentation masks for each category, and display the distribution of building units in the three-dimensional space through the segmentation masks. Specifically, the segmentation mask is a binary image or volume data with the same dimension as the input data, where the positions corresponding to the target category are marked as 1, and the background and other categories are marked as 0. These segmentation masks can intuitively display the distribution of power facilities in the 3D space, providing accurate geographical information and facility layout references for power designers.

[0144] After the above steps S2 to S3, first, the unprocessed two-dimensional CAD image data (usually, in the same project, the two-dimensional CAD image data includes at least the views of six faces of the imaging target, including the top, bottom, and four side faces), the corrected two-dimensional CAD image data obtained through steps such as position rotation, size ratio adjustment, and perspective switching, and then the preprocessed two-dimensional CAD image data obtained through sharpness improvement and noise reduction processing. After feature extraction and classification annotation described in step S3 based on the preprocessed two-dimensional CAD image data, the plane building position feature information, plane building unit feature information, and building unit connection relationship feature information are classified respectively.

[0145] In some embodiments, after completing feature extraction and annotation classification, in step S3, the large language model is used to identify building elements and determine whether the building elements meet the current scene and construction conditions, including:

[0146] The large language model judges whether the building elements meet the current scene according to the building requirements read based on the project information for the plane building position feature information, that is, judges whether the positions of each building unit are reasonable; for the building elements that do not meet the current scene, they are marked and recorded, and the error information can be directly output for designers to refer to.

[0147] The large language model judges whether the current building unit meets the construction requirements of the corresponding building unit based on the building requirements read based on the project information for the plane building unit feature information; similarly, the building units that do not meet the construction conditions are marked and recorded, and the error information can be directly output for designers to refer to.

[0148] After completing the determination of the current scene and construction conditions, the large language model supplements the detailed drawing information of the building units according to the precautions entered in the knowledge base and project information.

[0149] In addition, in this step, the error information marked by the large language model or the building units that do not meet the construction conditions can be marked in red for differential display in the finally generated model.

[0150] S4. Select a matching basic building unit model from the prefabricated model database according to the two-dimensional feature information; place the basic building unit model into the three-dimensional coordinates of the power grid construction site according to the coordinate information and in combination with the layout information of the plan view; and based on the six-view rendering engine, establish a mapping relationship between the two-dimensional image and its corresponding basic building unit model, and perform the tasks of three-dimensional imaging and modeling in space.

[0151] Specifically, step S4 includes the following steps S41 - S43.

[0152] S41. Select a matching basic building unit model from the prefabricated model database according to the extracted planar building unit feature information; wherein, the prefabricated model database includes three-dimensional voxel point models of power grid functional units belonging to different types of power stations, and any three-dimensional voxel point model is defined as a basic building unit model.

[0153] Specifically, the basic building unit models contained in the prefabricated model database are shown in Table 1:

[0154] Table 1 Example Table of Basic Building Unit Models

[0155]

[0156] The basic building unit model records the size description information, 3D modeling information, physical constraint condition information, etc. of the basic building unit.

[0157] S42. Place the basic building unit model into the three-dimensional coordinates of the power grid construction site according to the coordinate information and in combination with the layout information of the plan view; specifically, when we use 2D - CNN to extract the planar building position feature information, planar building unit feature information, and building unit connection relationship feature information, we can match the corresponding basic building unit model from the prefabricated model library according to the planar building unit feature information, and place the basic building unit model into the three-dimensional coordinates of the building site according to the coordinate information provided by the large language model and in combination with the layout information of the plan view.

[0158] After receiving the processed input image sequence, the six-view rendering engine first encodes these image sequences into a series of hidden states; these hidden states capture the temporal dependencies in the image sequence and contain spatial features; subsequently, these hidden states are sent to the decoder part. When processing these hidden states, the decoder first applies 3D convolutional operations to extract and fuse spatio-temporal features. Immediately afterwards, a non-linear activation function (sigmoid) is introduced to introduce non-linearity, enabling the model to learn more complex mapping relationships. Gradually increase the resolution of the hidden states until the target output resolution is reached.

[0159] Specifically, the underlying principle of the six-view rendering engine is based on the 3D-LSTM network and GRU (Gated Recurrent Unit).

[0160] As Figure 3 shown, the 3D-LSTM network consists of a set of structured LSTM units with restricted connections to form a three-dimensional grid structure. Six CAD drawings are preprocessed into three-dimensional data and divided into a grid of N×N×N, where N is the spatial resolution.

[0161] Each 3D-LSTM unit (or 3D-lsmunit) has an independent hidden state and input / output gates for processing information from adjacent units and the previous time step. The formula for updating the 3D-LSTM unit state is as follows:

[0162]

[0163]

[0164]

[0165]

[0166]

[0167]

[0168] Where represents a specific position in 3D space, respectively represent the indices in the three dimensions; in 3D-LSTM, such indices are used to locate specific inputs and states; t represents the time step, and t - 1 is used to indicate the previous moment relative to the current moment; represents the input data of the 3D-LSTM network; , , respectively represent the activation values of the forget gate, input gate, and output gate; and respectively represent the cell state and candidate cell state; represents the hidden state; σ is the sigmoid function; represents element-wise multiplication; tanh represents the hyperbolic tangent function, which is used as the activation function in LSTM to compress the values between -1 and 1; represents the weight, represents the bias parameter, and The subscripts are used to mark the corresponding gate or unit status. Among them, subscript f corresponds to the forget gate, subscript i corresponds to the input gate, subscript o corresponds to the output gate, and subscript c corresponds to the unit status. Based on the conventional LSTM, the weight is set to whether there is such a pixel point on the directly read CAD drawing. If it exists, it is 1; if it does not exist, it is 0. b is set to the weight provided by the large language model. For example, for the pixel point at the window position of the power supply room, theoretically there is no window, so it is 0. At this time, the drawing effect is as Figure 4 shown.

[0169] The GRU layer is placed after the 3D-LSTM layer to assist the 3D-LSTM in processing time series information. Especially when long-term dependencies need to be captured, it is used to process time series data and global context information. The update gate and reset gate of the GRU unit can be respectively expressed as:

[0170]

[0171]

[0172]

[0173]

[0174] Among them, represents the update gate; represents the reset gate; represents the hidden state in the GRU unit; represents the candidate hidden state; represents the input data of the GRU layer; represents the weight matrix corresponding to the update gate; represents the weight matrix corresponding to the reset gate; represents the weight matrix of the input vector; represents the weight matrix corresponding to the hidden state; ⊙ represents element-wise multiplication.

[0175] During the drawing process, the weights are dynamically adjusted according to the characteristic information of the power grid prefabrication model library to make each building more in line with the power grid building. At this time, the drawing effect is as Figure 5 shown.

[0176] Based on the assistance of the GRU unit in drawing, the mean squared error (MSE) calculation method is used as the loss function:

[0177]

[0178] Among them, represents the loss function of the GRU unit; represents the The true value of a sample is the known observed data; denotes the predicted value of the sample, which is the data calculated by the model; denotes the number of samples, i.e., the total number of samples contained in the dataset. The effect diagram obtained after removing the loss error is as Figure 6 shown.

[0179] In some embodiments, the artificial intelligence power grid building imaging method further includes steps S5 and S6.

[0180] Step S5 is a post-processing operation, specifically for edge smoothing of the generated segmentation mask, removing possible isolated noise points and small-area misclassified regions, making the edges of the segmentation result smoother and more natural, and conforming to the physical form and spatial distribution characteristics of actual power facilities. This helps to improve the quality and usability of the segmentation result and reduce errors in subsequent analysis and design. Also, for possible holes in the segmentation mask (regions that are not correctly classified due to occlusion or data loss), a hole filling algorithm based on neighborhood information is used for repair; for example, the reasonable class attribution of the hole region can be inferred according to the characteristics and spatial relationships of the surrounding classified pixels, further improving the integrity and accuracy of the segmentation result. The effect diagram after being processed by step S5 is as Figure 7 shown.

[0181] Step S6 is a visual output. Specifically, the 3D visualization interface output mainly depends on the 3D images generated by the present disclosure and outputs by connecting to other 3D software through the 3D imaging API. There are specifically the following two types:

[0182] Display by the drawing engine: Through the above operation steps, a 3D model image in black, white, and red colors is generated for designers to refer to and display.

[0183] Display by connecting to 3D drawing software: Through the API interface operation of 3Dmax2022, the final semantic segmentation result is presented to power designers in an intuitive 3D visualization form. The segmentation mask is superimposed on the original 3D imaging data, and different power facility categories are identified with different colors or textures, enabling designers to clearly observe and analyze the spatial distribution, mutual relationships of power facilities, and their interactions with the surrounding environment.

[0184] After processing, a list of problems generated during the operation of converting the 2D CAD drawing to a 3D view this time is output. At the same time, the modified information after the CPM-Bee building-specific large language model evaluates the characteristics of building units and construction locations during the execution process is also provided.

[0185] Some other embodiments of the present invention provide an artificial intelligence power grid building imaging engine based on a large language model for implementing the artificial intelligence power grid building imaging method based on a large language model described in any of the above embodiments, which includes: a three-dimensional image reconstruction module and a large language model processing module; the three-dimensional image reconstruction module receives two-dimensional CAD image data for describing a power grid building; the large language model processing module is used to receive project information and use the large language model to read the project information and power grid building design standards to obtain building requirements.

[0186] Specifically, the three-dimensional image reconstruction module includes:

[0187] A data preprocessing unit that preprocesses the two-dimensional CAD image data; the data preprocessing includes format conversion, data cleaning, and data augmentation; wherein, in data cleaning, a large language model is called to detect and correct outliers in the data;

[0188] A feature extraction module for obtaining two-dimensional feature information of the power grid building site information and various power equipment in the site through feature extraction;

[0189] A semantic segmentation module for classifying and annotating the two-dimensional feature information through semantic segmentation; the large language model processing module uses the large language model to identify building elements based on the processing results of the feature extraction module and the semantic segmentation module, and determines whether the building elements meet the current scene and construction conditions;

[0190] A three-dimensional imaging module that selects a matching basic building unit model in the prefabricated model database according to the two-dimensional feature information; places the basic building unit model into the three-dimensional coordinates of the power grid building site according to the coordinate information and in combination with the layout information of the floor plan; and based on the six-view rendering engine, establishes a mapping relationship between the two-dimensional image and its corresponding basic building unit model, and performs spatial three-dimensional imaging and modeling tasks.

[0191] In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiments or instances. Moreover, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples.

[0192] Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An artificial intelligence power grid building imaging method based on a large language model, characterized in that, Including: Obtain basic data, where the basic data at least includes two-dimensional CAD image data for describing power grid buildings and project information; Use a large language model to read project information and power grid building design standards to obtain building requirements; Perform data preprocessing on the two-dimensional CAD image data; the data preprocessing includes format conversion, data cleaning, and data augmentation; among them, in data cleaning, a large language model is used to detect and correct outliers in the data; Construct a three-dimensional reconstruction model, and the three-dimensional reconstruction model gradually infers and restores the three-dimensional shape of power grid building objects according to the two-dimensional CAD image data after data preprocessing; among them, the power grid building site information and the two-dimensional feature information of various building units in the site are obtained through feature extraction, and the building unit is a power grid functional unit composed of at least one power equipment; the extracted two-dimensional feature information is classified and labeled through semantic segmentation, and, based on a large language model, building elements are identified to determine whether the building elements meet the current scenario and construction conditions; Select a matching basic building unit model from the prefabricated model database according to the two-dimensional feature information; place the basic building unit model into the three-dimensional coordinates of the power grid building site according to the coordinate information and in combination with the layout information of the floor plan; and based on the six-view rendering engine, establish a mapping relationship between the two-dimensional image and its corresponding basic building unit model, and perform spatial three-dimensional imaging and modeling tasks.

2. The method for artificial intelligence power grid building imaging based on a large language model according to claim 1, wherein, The two-dimensional CAD image data records the power grid building site information and the image information of various building units in the site, and the image information includes the two-dimensional coordinates, geometric shapes, and surface features of the power grid building; The project information includes at least one of the project name, project type, special situation description file, risk description file, CAD file description list, building review file list, and voxel points; The special situation description file records the special information of the specified building, and the special information includes the location, size, and angle of the specified building; The risk description file is used to restrict the construction of some prohibited buildings; The CAD file description list is used to describe the role of each two-dimensional CAD image data; The building review file list is used to supplement and verify building parameters.

3. The method for imaging an artificial intelligence power grid building based on a large language model according to claim 2, wherein, The large language model uses a large language model specialized for CPM-Bee architecture; the use of the large language model to detect and correct outliers in the data includes: The large language model specialized for CPM-Bee architecture identifies the height information, distance information, and angle information in the two-dimensional CAD image data that deviate from the required range according to its own knowledge base and the building review file list, marks and corrects the deviated data, and limits it within the required range; For the data information of the specified building recorded in the special situation description file but not meeting the requirements of the knowledge base of the large language model specialized for CPM-Bee architecture and the building review file list, the data is detected and corrected based on the required range specified in the special situation description file.

4. The method for imaging an artificial intelligence power grid building based on a large language model according to claim 2, characterized in that, The grid building site information and the two-dimensional feature information of various power equipment in the site include planar building position feature information, planar building unit feature information, and building unit connection relationship feature information; The grid building site information and the two-dimensional feature information of various building units in the site obtained through feature extraction include: Performing convolution processing on the preprocessed two-dimensional CAD image data using multiple two-dimensional convolutional layers; among them, each two-dimensional convolutional layer is equipped with a convolutional kernel of a set size and stride, and these convolutional kernels slide in the two-dimensional space to extract different features in the two-dimensional CAD image data; adding a non-linear activation function after each two-dimensional convolutional layer to introduce non-linear transformation and output the feature map to be processed; Performing a max-pooling operation on the feature map to be processed, selecting the maximum eigenvalue within each pooling window as the output to generate a local feature vector; Passing the local feature vector through a fully connected layer for feature fusion and transformation, learning the connection relationship and layout rules between different building units, and combining the local feature vectors into a global feature vector.

5. The artificial intelligence power grid building imaging method based on a large language model according to claim 4, characterized in that, The classification and annotation of the two-dimensional feature information through semantic segmentation include: Performing an upsampling operation on the global feature vector using a transposed convolutional layer in the decoder, expanding the low-resolution feature map into a higher-resolution feature map to restore the original spatial resolution of the data with the decoded feature map; For each pixel or voxel in the decoded feature map, calculating the probability distribution of its belonging to each building unit type through the Softmax function to obtain the type label most likely to which each pixel or voxel belongs, including: For the global feature vector output by the fully connected layer without passing through the activation function , where n is the total number of categories, is the score that the sample belongs to the th category. The calculation process of the Softmax function is as follows: Denotes the conditional probability symbol, where y is the true class label, refers to the probability that the output class is t given the input feature x; r represents the class index, traversing all classes, ∈{1,2,…,n}; Denotes the nth element of the unactivated feature vector This score is calculated by the last layer of the neural network; the total number of classes n, which is the total number of elements of the input vector and is the number of building unit types defined by the power industry standard, determines the number of probabilities that the Softmax function needs to calculate; According to the classification result, generating a corresponding segmentation mask for each category, and displaying the distribution of building units in the three-dimensional space through the segmentation mask; Classifying the planar building position feature information, planar building unit feature information, and building unit connection relationship feature information respectively in the three-dimensional reconstruction model.

6. The method for imaging an artificial intelligence power grid building based on a large language model according to claim 5, wherein The classification and annotation of the two-dimensional feature information through semantic segmentation also include: Introducing skip connections in the decoder to fuse the feature map at the corresponding level in the encoder with the upsampled feature map in the decoder.

7. The method for artificial intelligence power grid building imaging based on a large language model according to claim 5, characterized in that The recognition of building elements based on a large language model to determine whether the building elements meet the current scene and construction conditions includes: The large language model judges whether the building elements meet the current scene according to the building requirements read based on the project information for the planar building position feature information, that is, judges whether the positions of each building unit are reasonable; marks the building elements that do not meet the current scene and records them; The large language model judges whether the current building unit meets the construction requirements of the corresponding building unit based on the building requirements read based on the project information for the planar building unit feature information; After completing the determination of the current scene and construction conditions, the large language model supplements the detailed drawing information of the building unit according to the precautions entered in the knowledge base and project information.

8. The method for imaging an artificial intelligence power grid building based on a large language model according to claim 7, wherein, Selecting a matching basic building unit model from the prefabricated model database according to the two-dimensional feature information, and placing the basic building unit model into the three-dimensional coordinates of the grid building site according to the coordinate information and in combination with the layout information of the plan view; Based on the six-view rendering engine, a mapping relationship is established between the two-dimensional image and its corresponding three-dimensional voxel point model to perform spatial three-dimensional imaging and modeling tasks, including: Selecting a matching basic building unit model in a prefabricated model database according to the extracted planar building unit feature information; wherein the prefabricated model database includes three-dimensional voxel point models of power grid functional units of different types of power stations, and defining any three-dimensional voxel point model as a basic building unit model; The basic building unit model is placed in the three-dimensional coordinates of the power grid construction site according to the coordinate information and the layout information of the plan; After receiving the processed input image sequence, the six-view rendering engine first encodes these image sequences into a series of hidden states; these hidden states capture the temporal dependencies in the image sequence and contain spatial features; then, these hidden states are transmitted to the decoder part; when processing these hidden states, the decoder first applies 3D convolution operations to extract and fuse spatial-temporal features; then, nonlinear characteristics are introduced through the nonlinear activation function sigmoid, so that the model can learn more complex mapping relationships; the resolution of the hidden state is gradually improved until the target output resolution is reached.

9. The artificial intelligence power grid building imaging method based on a large language model according to claim 1, characterized in that Also includes: Perform edge smoothing on the generated segmentation mask to remove possible isolated noise points and small misclassified areas; For possible holes in the segmentation mask, a hole filling algorithm based on neighborhood information is used to repair them; The final semantic segmentation results are presented to power designers in an intuitive 3D visualization form; the segmentation mask is superimposed on the original 3D imaging data to identify different power equipment categories with different colors or textures.

10. An artificial intelligence power grid building imaging engine based on a large language model, characterized in that, The artificial intelligence power grid building imaging method based on a large language model as described in any one of claims 1 to 9, wherein the artificial intelligence power grid building imaging engine comprises: a three-dimensional image reconstruction module and a large language model processing module; the three-dimensional image reconstruction module receives two-dimensional CAD image data for describing the power grid building; the large language model processing module is used to receive project information, and use the large language model to read the project information and power grid building design standards to obtain building requirements; The three-dimensional image reconstruction module comprises: A data preprocessing unit performs data preprocessing on the two-dimensional CAD image data; the data preprocessing includes format conversion, data cleaning and data enhancement; wherein, in data cleaning, a large language model is called to detect and correct abnormal values in the data; A feature extraction module is used to obtain the two-dimensional feature information of the power grid building site and various types of power equipment in the site through feature extraction; A semantic segmentation module is used to classify and label the two-dimensional feature information through semantic segmentation; the large language model processing module uses a large language model to identify building elements according to the processing results of the feature extraction module and the semantic segmentation module, and determines whether the building elements meet the current scene and construction conditions; A three-dimensional imaging module selects a matching basic building unit model from a prefabricated model database according to the two-dimensional feature information; places the basic building unit model into the three-dimensional coordinates of the power grid construction site according to the coordinate information and in combination with the layout information of the plan view; and based on a six-sided view rendering engine, establishes a mapping relationship between the two-dimensional image and its corresponding basic building unit model, and executes the tasks of three-dimensional imaging and modeling of space.

Images