A Geological BIM Modeling and Quantity Calculation Method Based on Point Cloud Sparse Optimization

Abstract

This application proposes a geological BIM modeling and quantity calculation method based on point cloud sparse optimization, belonging to the field of geological engineering technology. The method includes: acquiring high-resolution point cloud data using UAV oblique photogrammetry, combining it with borehole exploration data, and performing denoising, classification, segmentation, lithological differentiation, and registration to construct a pre-excavation geological BIM model; acquiring regional point cloud data again after excavation, optimizing the point cloud accuracy of the slope toe area using an inverse distance weighted interpolation algorithm, and generating a post-excavation geological BIM model; combining a classification-optimized digital elevation model, employing a robust elevation difference calculation method based on neighborhood windows to detect the elevation difference of the model before and after excavation pixel-by-pixel, accurately calculating the volume change of each lithological region, and obtaining the quantity of work. This application improves the accuracy of the geological BIM model and the reliability of volume calculation through point cloud sparse optimization and elevation change detection technology, and is widely applicable to geological engineering and earthwork volume calculation scenarios.

Description

Technical Field

[0001] This application relates to the field of geological engineering technology, and in particular to a geological BIM modeling and engineering quantity calculation method based on point cloud sparse optimization. Background Technology

[0002] In geological engineering and excavation construction, accurately acquiring the three-dimensional morphology, internal structure, and excavation volume of geological bodies is the core foundation for construction planning and management. However, traditional geological information acquisition and modeling methods have significant shortcomings in terms of accuracy, comprehensiveness, and data integration capabilities. Furthermore, the detection accuracy of localized elevation changes is low, further limiting the accuracy of geological BIM model construction and quantity calculation, thus posing numerous challenges to construction management and decision-making.

[0003] Existing geological BIM modeling and dynamic calculation methods for excavation quantities primarily rely on point cloud data acquisition and traditional geometric modeling techniques. These methods, centered on point cloud data processing, digital elevation model (DEM) generation, and volume calculation, aim to model the three-dimensional morphology of geological bodies and perform statistical analysis of excavation quantities. Typically, multi-view point cloud data is acquired using UAV oblique photogrammetry or terrestrial laser scanning techniques, combined with borehole exploration data to establish a basic geological three-dimensional model. Mathematical calculations are then used to assess the excavation quantities.

[0004] While traditional methods have achieved geological BIM modeling and quantity calculation to some extent, their limitations stem from the poor adaptability of fixed algorithms to complex terrains and their inability to handle issues such as sparse point cloud data and insufficient representation of key features like slope toes. Furthermore, traditional fixed-threshold elevation difference detection methods perform poorly in nonlinear and non-stationary geological change scenarios, often neglecting the influence of internal geological structure and external environmental factors, thus limiting prediction accuracy and quantity calculation precision. This is particularly true in slope toe areas and boundary feature areas during excavation, where insufficient model accuracy can lead to significant statistical errors in quantity calculations, further restricting the application value of geological BIM models in actual construction. Summary of the Invention

[0005] This application aims to at least partially address one of the technical problems in the related art.

[0006] Therefore, the first objective of this application is to propose a geological BIM modeling and engineering quantity calculation method based on point cloud sparse optimization.

[0007] The second objective of this application is to propose a geological BIM modeling and engineering quantity calculation device based on point cloud sparse optimization.

[0008] To achieve the above objectives, the first aspect of this application proposes a geological BIM modeling and engineering quantity calculation method based on point cloud sparse optimization, including:

[0009] Oblique photography technology is used to acquire geological body image data and three-dimensional point cloud data, and the three-dimensional point cloud data is correlated and combined with borehole exploration data;

[0010] The three-dimensional point cloud data is preprocessed. Using the preprocessed three-dimensional point cloud data, image data and borehole exploration data, a geological BIM model is constructed before excavation. The data preprocessing includes denoising, classification, segmentation, lithology differentiation and registration.

[0011] After the excavation work was completed, oblique photography technology was used to obtain point cloud data of the same area, and the geological model was reconstructed. In order to address the problem of unclear and sparse point cloud identification in the slope foot area, the inverse distance weighted interpolation algorithm was used to optimize the data accuracy and construct the geological BIM model after excavation.

[0012] Based on lithology, the geological BIM model before excavation is divided into regions. For the digital elevation models before and after excavation, a robust elevation difference calculation method based on neighborhood windows is used to calculate the elevation difference of the geological BIM models before and after excavation pixel by pixel.

[0013] Based on the elevation difference calculation results, the volume change of each lithological zone is calculated to obtain the total engineering volume of the entire excavation area.

[0014] Optionally, the step of acquiring geological body image data and 3D point cloud data using oblique photography technology, and correlating and combining the 3D point cloud data with borehole exploration data, includes:

[0015] Multi-view image data of geological bodies are collected using UAV oblique photography technology, including vertical and oblique images.

[0016] Based on PIX4D software, the acquired multi-view image data is imported and processed. Dense point cloud data is generated through the software's automated workflow, and the spatial geometric information of the geological body is extracted to form three-dimensional point cloud data containing high-precision three-dimensional coordinates.

[0017] Collect and integrate borehole exploration data to establish a database of internal geological structure information, and associate it with the three-dimensional point cloud data. The borehole exploration data includes borehole location, stratum depth, lithological distribution, and physical properties.

[0018] Optionally, the step of preprocessing the three-dimensional point cloud data, and using the preprocessed three-dimensional point cloud data, image data, and borehole exploration data to construct a geological BIM model before excavation, includes:

[0019] The three-dimensional point cloud data is denoised by using a statistical filtering algorithm to remove isolated points, noise points, and irregular point clusters, ensuring the smoothness and accuracy of the point cloud data.

[0020] Based on a multi-layer dynamic graph convolutional neural network, the denoised point cloud data is classified, and geological features, vegetation, buildings and noise are extracted from the point cloud according to geometric features, providing a basis for annotation and classification for subsequent removal of non-geological feature point clouds.

[0021] Based on the classified 3D point cloud data, deep feature learning is used to segment it, extract feature regions related to geological bodies, and remove point cloud data that are not related to geological bodies.

[0022] By differentiating the lithology of the pure geological point cloud data generated after segmentation, and combining borehole exploration data and deep learning classification models, the point cloud data of the geological body is refined into different lithological regions.

[0023] The 3D point cloud data after lithological differentiation is registered, and the spatial alignment of the 3D point cloud data from multiple scans is performed by the iterative nearest point algorithm. At the same time, the coordinates are unified and fused with the borehole exploration data to ensure that the point cloud, image data and geological borehole information are unified in the same coordinate system, thereby eliminating spatial errors and integrating surface and subsurface information.

[0024] Using the preprocessed 3D point cloud data, image data, and borehole exploration data, a complete geological BIM model is constructed to represent the 3D morphology, internal structure, and attribute information of the geological body.

[0025] Optionally, the preprocessed 3D point cloud data, image data, and borehole exploration data are used to construct a complete pre-excavation geological BIM model to represent the 3D morphology, internal structure, and attribute information of the geological body, including:

[0026] Import the preprocessed 3D point cloud data and borehole exploration data into BIM modeling software;

[0027] Geometric features of the land surface, including slope, toe, and fault areas, are extracted from the three-dimensional point cloud data, and a digital elevation model (DEM) of the land surface is generated. A subsurface stratification model is established by combining the borehole exploration data and docking the stratum depth information in the borehole with the point cloud surface model to form a unified three-dimensional geological BIM model of the surface and subsurface.

[0028] Add necessary geological attributes to the generated BIM model to distinguish the excavation volume of different lithologies and strata. The geological attributes include the lithology of the strata and the density and volume parameters related to the excavation requirements.

[0029] Output a complete geological BIM model before excavation.

[0030] Optionally, after the excavation is completed, oblique photography is used to acquire point cloud data of the same area, and a geological model is constructed again. To address the issues of unclear and sparse point cloud identification in the slope toe area, an inverse distance weighted interpolation algorithm is used to optimize data accuracy, thus constructing a post-excavation geological BIM model, including:

[0031] After the excavation work was completed, UAV oblique photography technology was used again to collect multi-view images of the excavated area, and the three-dimensional point cloud data of the geological body after excavation was obtained by using the data from the newly collected multi-view images.

[0032] Based on the 3D point cloud data of the geological body after excavation, inverse distance weighted interpolation algorithm is used to fill in the elevation values ​​of missing points in areas with sparse point clouds. In addition, for the problem of unclear and sparse point cloud identification in the slope foot area, inverse distance interpolation combined with local trend fitting method is used to adjust the interpolation strategy and parameters.

[0033] The optimized 3D point cloud data was combined with the borehole exploration data before excavation, and the geological BIM model after excavation was constructed using Revit modeling software.

[0034] Optionally, based on the three-dimensional point cloud data of the excavated geological body, for areas with sparse point clouds, an inverse distance weighted interpolation algorithm is used to complete the elevation values ​​of missing points. Furthermore, to address the issues of unclear and sparse point cloud identification in the slope toe area, an inverse distance interpolation combined with a local trend fitting method is used to adjust the interpolation strategy and parameters, including:

[0035] Density analysis was performed on the 3D point cloud data of the excavated geological body to identify sparse regions. Specifically, for each point p... i If the number of points in its neighborhood is less than the preset number of points, then determine p. i The area in question is a sparse region.

[0036] For sparse regions, the inverse distance weighted interpolation algorithm is used to complete the elevation values ​​Z(x,y) of missing points, expressed as:

[0037]

[0038] In the formula, Z i It is a neighboring point p i The known elevation value, d i is the distance between the interpolation point and its neighboring points, and p is the distance weighting exponent;

[0039] From the point cloud data of the slope toe area obtained based on classification before excavation, key areas with sparse or unclear feature representation are marked.

[0040] Within the marked slope toe area, feature points in the neighborhood of sparse regions are selected, and a local trend equation is established using a linear regression fitting method.

[0041] Based on the fitted trend results obtained after solving the local trend equation, the parameters of the inverse distance interpolation are optimized, and the preliminary elevation values ​​of the trend fitting are combined with the interpolation results of the inverse distance weighted algorithm to generate the final interpolated elevation values ​​according to the weights.

[0042] After interpolation is completed, the neighborhood mean smoothing method is used to eliminate the abruptness of the interpolation region;

[0043] After interpolation and optimization are completed, the elevation values ​​of the interpolated area are compared with those of the original three-dimensional point cloud data of the excavated geological body. The accuracy of the interpolated area is verified by calculating the mean square error.

[0044] Optionally, the step of dividing the geological BIM model before excavation into regions based on lithology, and calculating the pixel-by-pixel elevation difference of the geological BIM model before and after excavation using a robust elevation difference calculation method based on neighborhood windows, includes:

[0045] Based on the three-dimensional point cloud data before excavation, which has been differentiated by lithology, different lithological regions are labeled.

[0046] Based on the labeled lithological area, the elevation difference is calculated pixel by pixel using the elevation data generated before and after excavation and the neighborhood window method. For each pixel, the elevation data of the previous and subsequent stages are compared within the neighborhood window, and the minimum absolute value is selected as the true elevation difference. At the same time, an elevation difference threshold is set to mark the changed points or unchanged points, and the pixel-level results are mapped to the lithological area.

[0047] Optionally, based on the labeled lithological area, using the elevation data generated before and after excavation, a neighborhood window method is used to calculate the elevation difference pixel by pixel. For each pixel, the elevation data before and after the excavation are compared within the neighborhood window, and the minimum absolute value is selected as the true elevation difference. Simultaneously, an elevation difference threshold is set to mark changed or unchanged points, and the pixel-level results are mapped to the lithological area, including...

[0048] Set an appropriate neighborhood window size based on the geological features, select the center pixel of the window, locate its coordinates in the post-excavation stage elevation data, and extract all pixel values ​​of the corresponding window range in the pre-excavation stage elevation data.

[0049] For each pixel within the window, calculate the elevation difference between the center pixel after excavation and the neighboring pixels before excavation, and take the minimum absolute value as the true elevation difference of the current pixel;

[0050] Each pixel is marked as changed by setting an elevation difference threshold. If the actual elevation difference of the current pixel is greater than or equal to the elevation difference threshold, it is marked as a changed point; otherwise, it is marked as an unchanged point.

[0051] The proportion of change points in each lithological region is statistically analyzed, and the proportion of change pixels to total pixels is calculated. If the proportion of change points in a certain lithological region reaches a preset threshold, the region is marked as a change region; otherwise, the region is marked as an unchanged region.

[0052] Optionally, the step of calculating the volume change of each lithological zone based on the elevation difference calculation results to obtain the total engineering volume of the entire excavation area includes:

[0053] Based on the elevation difference calculation results, each lithological region is divided into regular grid units;

[0054] For each grid cell, the volume change of a single grid cell is calculated based on the elevation difference and grid area;

[0055] The volume change of all grids within each lithological region is summed to obtain the volume change of that region. For regions with negative elevation differences, the volume change is recorded as excavation volume; for regions with positive elevation differences, the volume change is recorded as deposition volume.

[0056] By summarizing the volume change data of all lithological areas, the total engineering volume of the entire excavation area is calculated.

[0057] To achieve the above objectives, a second aspect of this application proposes a geological BIM modeling and quantity calculation device based on point cloud sparsity optimization, comprising:

[0058] The data acquisition module is used to acquire geological body image data and three-dimensional point cloud data using oblique photography technology, and to associate and combine the three-dimensional point cloud data with borehole exploration data;

[0059] The geological model construction module before excavation is used to preprocess the three-dimensional point cloud data. Using the preprocessed three-dimensional point cloud data, image data and borehole exploration data, a geological BIM model before excavation is constructed. The data preprocessing includes denoising, classification, segmentation, lithology differentiation and registration.

[0060] The post-excavation geological model reconstruction module is used to acquire point cloud data of the same area using oblique photography technology after the excavation work is completed, and to rebuild the geological model again. In addition, it uses an inverse distance weighted interpolation algorithm to optimize the data accuracy and build a post-excavation geological BIM model to address the problems of unclear and sparse point cloud identification in the slope foot area.

[0061] The elevation difference calculation module is used to divide the geological BIM model before excavation into regions based on lithology, and to calculate the elevation difference pixel by pixel for the digital elevation models before and after excavation using a robust elevation difference calculation method based on neighborhood windows.

[0062] The quantity calculation module is used to calculate the volume change of each lithological zone based on the elevation difference calculation results, and obtain the quantity of work for the entire excavation area.

[0063] The technical solutions provided by the embodiments of this application bring at least the following beneficial effects:

[0064] By introducing point cloud sparsity optimization technology, combined with inverse distance weighted interpolation (IDW) algorithm and local trend fitting method, the problems of point cloud data sparsity and insufficient expression of slope foot area features in traditional methods are effectively solved, and the accuracy and detail representation of geological models are greatly improved.

[0065] By employing a robust elevation difference calculation method based on neighborhood windows and combining it with a classification-optimized digital elevation model (DEM), pixel-by-pixel elevation change fine detection is achieved, significantly improving the accuracy of excavation volume calculation, especially in complex terrain and non-stationary geological change scenarios.

[0066] By improving the data preprocessing workflow, including denoising, classification, segmentation, and lithology differentiation, the impact of point cloud data registration errors on engineering quantity calculations is effectively reduced, thereby improving the reliability and stability of the results.

[0067] By combining UAV oblique photogrammetry technology to acquire high-resolution point cloud data, data acquisition becomes more efficient and convenient. Compared to traditional methods, this solution provides more comprehensive and accurate geological information support for construction planning and management, significantly improving the overall efficiency and decision-making capabilities of the project.

[0068] This technical solution has been specifically optimized for slope toe areas and boundary feature areas under complex geological conditions, which significantly reduces the statistical errors of traditional methods in these key areas and ensures the practical application value of geological BIM models and excavation volume calculation results.

[0069] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0070] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0071] Figure 1A flowchart illustrating a geological BIM modeling and engineering quantity calculation method based on point cloud sparse optimization provided in this application embodiment;

[0072] Figure 2 This is a schematic diagram of a geological BIM modeling and engineering quantity calculation device based on point cloud sparse optimization, provided as an embodiment of this application. Detailed Implementation

[0073] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0074] To address the problems existing in the prior art, this application provides a geological BIM modeling and engineering quantity calculation method based on point cloud sparsity optimization. Figure 1 This is a flowchart illustrating a geological BIM modeling and quantity calculation method based on point cloud sparse optimization, provided as an embodiment of this application. Figure 1 As shown, the method includes the following steps:

[0075] S1. Oblique photography technology is used to acquire geological body image data and 3D point cloud data, and the 3D point cloud data is correlated and combined with borehole exploration data.

[0076] In this embodiment, oblique photogrammetry is used to acquire high-precision 3D point cloud and image data of the geological body. This is combined with borehole exploration data (including strata, lithology, depth, etc.) to provide complete multi-source input data for the construction of the geological BIM model. The combined application of oblique photogrammetry and borehole exploration ensures the comprehensiveness and accuracy of the data, laying a solid foundation for subsequent model generation.

[0077] Specifically, step S1 includes the following steps:

[0078] S101. Collect multi-view image data of geological bodies using UAV oblique photography technology. The multi-view image data includes vertical images and oblique images.

[0079] In this embodiment, a drone equipped with a high-resolution camera is used to perform multi-angle oblique photography of the target geological body. During the image acquisition process, a combination of vertical and oblique shooting is used to ensure that the acquired image data covers the entire range of the geological body, including key areas such as the slope, slope toe, and boundary features.

[0080] Understandably, multi-view acquisition effectively avoids the problem of traditional single-view image data failing to cover complex terrain, ensuring the integrity and accuracy of the generated 3D point cloud data. Meanwhile, to reduce the impact of external environmental factors (such as weather and lighting conditions) on data acquisition quality, it is recommended to conduct aerial photography in clear, windless weather, and to pre-plan the drone's flight path and shooting intervals to ensure data uniformity and coverage. After acquisition, an image dataset containing multiple angles and resolutions is generated, providing a reliable data foundation for subsequent 3D point cloud generation.

[0081] S102. Based on PIX4D software, the acquired multi-view image data is imported and processed. Dense point cloud data is generated through the software's automated workflow, and the spatial geometric information of the geological body is extracted to form three-dimensional point cloud data containing high-precision three-dimensional coordinates.

[0082] In this embodiment, the acquired multi-view image data is imported into PIX4D software, and its built-in automated workflow is used to complete image processing and point cloud data generation. The specific operations include the following key steps:

[0083] First, the software uses image feature matching algorithms (such as SIFT) to automatically register and stitch multi-view image data to construct a spatial geometric framework model of the geological body.

[0084] Secondly, after the initial registration is completed, dense point cloud data is generated based on the multi-view stereo reconstruction (MVS) algorithm. This point cloud data contains the three-dimensional coordinate information of the geological body and surface detail features, and has the characteristics of high resolution and high accuracy.

[0085] Finally, to further improve the quality of point cloud data, the built-in filtering tools of PIX4D software were used to remove outliers and noise points, while retaining the true point cloud information of the geological surface.

[0086] Through the above processing steps, a dense point cloud dataset containing high-precision three-dimensional coordinates is finally generated. This data can fully reflect the external geometric features of the geological body, laying the foundation for the subsequent construction of a geological BIM model.

[0087] S103. Collect and integrate borehole exploration data, establish a database of internal geological structure information, and associate it with three-dimensional point cloud data. The borehole exploration data includes borehole location, stratum depth, lithological distribution, and physical properties.

[0088] In this embodiment of the application, borehole exploration data is an important data source reflecting the internal structural characteristics of geological bodies, and its collection and integration include the following:

[0089] First, obtain the location information of boreholes within the target area and collect relevant attribute data, including physical characteristics such as formation depth, lithology distribution, porosity, density, and permeability. All collected data should originate from reliable exploration records to ensure their accuracy and usability.

[0090] Then, based on the collected borehole data, a digital information database of the internal structure of the geological body is established, and the data is managed hierarchically. For example, data can be stored according to different strata or lithologies for easy retrieval and analysis later.

[0091] Finally, the borehole exploration data is integrated with the 3D point cloud data generated by oblique photogrammetry. Specifically, a coordinate matching method is used to align the geographical location of the boreholes with the 3D coordinates of the point cloud data to form a complete representation model of the external geometric features and internal structural characteristics of the geological body. This step enables the organic fusion of multi-source data, providing comprehensive information support for geological BIM models.

[0092] In addition, the integrated data needs to be preliminarily sorted out to remove overlapping images and outliers in the point cloud data, and the accuracy of the borehole exploration data needs to be checked.

[0093] S2. Perform data preprocessing on the 3D point cloud data. Using the preprocessed 3D point cloud data, image data, and borehole exploration data, construct a geological BIM model before excavation. Data preprocessing includes denoising, classification, segmentation, lithology differentiation, and registration.

[0094] To ensure the integrity and accuracy of 3D point cloud data, comprehensive preprocessing is required before using it for geological BIM model construction.

[0095] The core objective of data preprocessing is to clean, optimize, and structure point cloud data, while organically integrating it with image data and borehole exploration data. Preprocessing steps include denoising, classification, segmentation, lithological differentiation, and registration to ensure the consistency and correlation of point cloud data in terms of spatial geometric information and internal lithological characteristics.

[0096] Preprocessing effectively improves data quality, eliminates noise and redundancy, and accurately represents the geometric morphology and structural characteristics of geological bodies, thus providing a solid foundation for the subsequent construction of geological BIM models. Finally, using the preprocessed 3D point cloud data, image data, and borehole exploration data, a geological BIM model is constructed before excavation, providing an accurate basic model for geological analysis and engineering planning.

[0097] Specifically, step S2 also includes:

[0098] S201. Denoise the 3D point cloud data by using statistical filtering algorithms to remove isolated points, noise points, and irregular point clusters, ensuring the smoothness and accuracy of the point cloud data.

[0099] First, noise points are identified and cleaned from the raw point cloud data. Since point cloud data generated by UAV oblique photography may contain redundant environmental points (such as trees, equipment, etc.) or anomalous noise points, isolated and error points need to be removed using filtering algorithms based on distance, density, and normal vector distribution. This process significantly improves the realism and usability of the point cloud data, providing clean foundation data for subsequent processing.

[0100] In this embodiment, the statistical filtering (SOR, Statistical Outlier Removal) algorithm is used to denoise the original point cloud data in order to remove isolated points, noisy points and irregular point clusters.

[0101] Statistical filtering analyzes the neighborhood distribution of each point in the point cloud, calculates the distance from each point to its neighborhood center, and compares this distance with global statistical values ​​(such as the mean and standard deviation) to identify and remove outliers that deviate from the normal range. Simultaneously, it preserves key geometric feature points to ensure data integrity. Denoising significantly improves the smoothness and accuracy of point cloud data, effectively eliminating interference from external environments (such as noise or equipment errors) on data quality. The processed point cloud data has a higher signal-to-noise ratio, laying a high-quality data foundation for subsequent classification, segmentation, and registration operations.

[0102] Furthermore, the specific operation of step S201 includes the following detailed process:

[0103] S2011, First, represent the point cloud data as a point set P = {p i |p i =(x i ,y i ,z i )}, where each point p i It includes its three-dimensional coordinate information (x, y, z). To identify isolated and noisy points, embodiments of this application specify the three-dimensional coordinate information (x, y, z) for each point p. i Includes defining a local neighborhood N i The neighborhood range is determined by the set search radius r:

[0104] N i ={p j ∣∥p i -p j ∥≤r,p j ∈P}

[0105] Among them, ∥p i -p j∥ represents point p i and p j The Euclidean distance between points is given by r, which is the radius parameter used for neighborhood search. Adjusting r controls the size of the neighborhood. The purpose of local neighborhood partitioning is to associate each point with a set of surrounding points for subsequent statistical analysis.

[0106] S2012, For each point p i Calculate the average distance of its neighboring points.

[0107]

[0108] Where, |N i |for point p i The total number of points in a neighborhood. That is, the total number of points in the neighborhood. By calculating the average distance from each point to its neighborhood points, the spatial distribution characteristics of a point in its local environment can be obtained.

[0109] S2013. Perform statistical analysis on the average distance of all points, and calculate the global mean μ and standard deviation σ:

[0110]

[0111]

[0112] Where n is the total number of points in the point cloud data. The global mean μ reflects the average neighborhood distance of all points, while the standard deviation σ represents the dispersion of the distance distribution between points.

[0113] S2014. Based on the calculated global mean μ and standard deviation σ, set filtering conditions to remove outliers and noise points. The filtering conditions are:

[0114]

[0115] Here, k is usually set to 2 to define a range. Any point p i Average neighborhood distance If the value is outside the specified range, the point is considered an isolated point or a noise point and is removed.

[0116] After the above steps, outliers in the point cloud data are removed, and the remaining point cloud data is smoother and more accurate, providing high-quality basic data for subsequent classification, segmentation and modeling operations.

[0117] S202. Based on a multi-layer dynamic graph convolutional neural network, the denoised point cloud data is classified. Geological features, vegetation, buildings and noise are extracted from the point cloud according to geometric features, providing a basis for annotation and classification for subsequent removal of non-geological feature point clouds.

[0118] In this embodiment, a multi-layer dynamic graph convolutional neural network (DGCNN) is used to classify point cloud data. The DGCNN can capture local and global features of point clouds through dynamic neighborhood feature maps, adapting to complex geometric shapes. The classification process extracts key information from the geometric features of the point cloud, dividing the point cloud data into categories such as geological feature points, vegetation points, building points, and noise points.

[0119] Specifically, the multi-layer dynamic graph convolutional neural network first constructs a neighborhood feature map, extracts local geometric relationships through convolution operations, and then captures global spatial information through the overlay of multiple deep networks. This method can dynamically adjust the classification criteria to ensure the robustness and accuracy of the classification results. The classified data provides a clear basis for subsequent removal of non-geological feature point clouds and extraction of key geological regions.

[0120] Furthermore, the specific operation of step S202 includes the following detailed process:

[0121] S2021. Using the denoised point cloud data obtained in step S201 as input, the edge convolution function is used to extract features of the geometric relationship between each point in the point cloud and its neighboring points. First, the edge feature e between a point and its neighboring points is defined. ij The calculation formula is as follows:

[0122] e ij =ReLU(W(p) j -p i )+b)

[0123] Where, p i p is the current point j Let p be a neighborhood point, W and b be learnable parameters, and ReLU be the activation function used to extract nonlinear local features. Edge features describe the features of point p. i With neighboring point p j The relative relationships between them can capture the local geometric properties of point clouds.

[0124] S2022, For each point p i It aggregates edge features within its neighborhood through max pooling operation to extract local features.

[0125]

[0126] Where l represents the number of convolutional layers. Through this process, point p iIt can obtain high-dimensional local geometric feature information from its neighboring points. This operation is repeated multiple times, extracting local features of different scales of the point cloud layer by layer by stacking multiple layers of convolutional networks, ensuring that the details and global geometric information of the point cloud are captured.

[0127] S2023. Using a multi-layer Dynamic Graph Convolutional Network (DGCNN), local features are extracted layer by layer into deep geometric features. The features of the (l+1)th layer are represented as:

[0128] H (l+1) =f(H (l) )

[0129] Among them, H (l) Let f be the feature of the l-th layer, and f be the ReLU activation function. The multi-layer DGCNN structure can dynamically adjust the edge features between a point and its neighbors, thereby extracting richer geometric information.

[0130] S024. Based on the extraction of local features, global max-pooling is used to perform global feature aggregation on the deep features of the point cloud data, generating a global feature vector h. global :

[0131]

[0132] Here, H(L) is the feature matrix of the last layer of the network, and the max operation takes the maximum value of the features at each point to ensure the unordered nature of the point cloud. The global feature vector h... global It can represent the overall geometric features of point cloud data, including global geometric information.

[0133] S2025, Transfer the global feature vector h global Input to a fully connected layer, and use the softmax function to output the classification probability distribution.

[0134]

[0135] The softmax function is used to calculate the probability distribution of each point belonging to different categories (such as geological bodies, vegetation, buildings, noise, etc.), while W and b are the weights and bias parameters of the fully connected layer. The classification result can assign a category label to each point in the point cloud data, providing a clear basis for subsequent point cloud segmentation.

[0136] Through the above steps, DGCNN can effectively capture the local and global features of point cloud data, and achieve high-precision classification of categories such as geological bodies, vegetation, buildings and noise in point clouds, providing important data support for further processing of geological models.

[0137] S203. Based on the classified 3D point cloud data, deep feature learning is used to segment it, extract feature regions related to geological bodies, and remove point cloud data that are unrelated to geological bodies.

[0138] In this embodiment of the application, based on the classified point cloud data, a deep feature learning algorithm is used to segment the geologically related areas.

[0139] During the segmentation process, initial seed points related to geological features are first selected and expanded to the entire target area using a region growing algorithm to extract key geological areas such as slopes, slope toes, and fault planes. Simultaneously, high-dimensional features extracted using a deep learning model are used to further optimize the region boundaries, ensuring the integrity and consistency of the segmentation results. Irrelevant point cloud data (such as trees and buildings) are removed during segmentation to improve the purity of geological body-related data. The segmented data better represents the three-dimensional structural characteristics of geological bodies and reduces the interference of redundant points on subsequent modeling and analysis, thus providing high-quality data support for geological BIM model construction.

[0140] Furthermore, the specific operation of step S203 includes the following detailed process:

[0141] S2031. Construct a feature relation matrix R using the deep features H of the point cloud data to describe the similarity relationships between each point in the point cloud. Feature relation matrix R ij The calculation is based on the cosine similarity function:

[0142]

[0143] Among them, R ij Point p i and point p j Feature similarity, h i and h j They are points p i and point p j The deep features are represented by ∥·∥, which indicates the L2 norm of the feature vector. Cosine similarity can effectively measure the similarity of features in a point cloud, providing support for subsequent classification and segmentation.

[0144] S2032. Based on the similarity matrix R and the classification results, classify each point p in the point cloud data. i Assign category tags The formula for assigning category labels is as follows:

[0145]

[0146] Here, argmax represents selecting the category with the highest similarity. Category labels include geological feature areas (such as slopes, slope toes, fault planes, etc.) and non-geological feature areas (such as trees, houses, noise points). Through this process, each point in the point cloud is explicitly labeled with its corresponding category.

[0147] S2033. Based on the assigned category labels, extract the point cloud data related to the geological body as feature regions. Specifically, extract slope features based on local normal vector consistency and curvature information, extract slope toe features using boundary points in areas with significant geometric feature changes, and extract fault feature points based on areas with high local feature gradients. The formula for feature region extraction is:

[0148]

[0149] in, The local gradient is used to identify different feature regions. The combination of curvature information and gradient changes ensures the accuracy of the extraction results and can effectively distinguish geological feature regions such as slope surfaces, slope toes, and faults.

[0150] S2034. Based on the classification results, remove points in the point cloud that are not classified as geological bodies (such as vegetation, buildings, and noise points) to generate a clean point cloud dataset. Clean point cloud dataset P geo The definition is as follows:

[0151]

[0152] The final point cloud dataset retains only point clouds relevant to geological bodies, removing irrelevant points such as trees, buildings, and other noisy points. This segmentation process generates high-purity geological feature point clouds, providing high-quality data support for subsequent geological BIM model construction and geological analysis.

[0153] S204. By differentiating the lithology of the pure geological point cloud data generated after segmentation, and combining borehole exploration data and deep learning classification models, the point cloud data of the geological body is refined into different lithological regions.

[0154] In this embodiment, borehole exploration data and a deep learning classification model are combined to differentiate lithology in the segmented geological point cloud data. The lithological distribution characteristics (such as sandstone, shale, and clay) in the borehole exploration data are matched with the geometric features of the point cloud, refining the geological point cloud into multiple lithological regions. This lithological differentiation endows the geological BIM model with physical and chemical attribute information, enabling the model to not only possess three-dimensional morphological features but also reflect the diversity and distribution patterns of lithology.

[0155] Furthermore, the specific operation of step S204 includes the following detailed process:

[0156] S2041. Extract the clean point cloud data P generated in step S203. geo Lithological distribution information is obtained from borehole exploration data. The borehole data includes the three-dimensional coordinates (x, y, z) of the borehole location and the corresponding lithological category label y. rock (e.g., soil layers, sandstone, shale, etc.). For each point p in the point cloud... i Calculate the distance d from the borehole location, assign preliminary lithological labels to the point cloud points, and classify them according to the following rules:

[0157]

[0158] Where r is the borehole influence radius, used to define the effective range of influence of the borehole on the point cloud points. Unclassified points represent point clouds that are far from all boreholes and will be further processed in subsequent steps. This step assigns preliminary lithological labels to the point cloud data based on spatial distribution.

[0159] S2042. Reconstruct the point cloud feature similarity matrix using the feature relation matrix R from step S2031. Use this matrix as input and extract the local and global features required for lithological classification using a multi-layer dynamic graph convolutional network (DGCNN). Features of each point cloud point are dynamically extracted through edge convolutional layers and combined with the geometric and attribute information of the surrounding neighborhood to generate deep features for classification. For each point, DGCNN uses the classification probability distribution P... i Predict the probability of a point belonging to each lithology category. Finally, assign the final lithology category to the point cloud point based on the maximum value of the probability distribution:

[0160]

[0161] Among them, P i For point p i The probability distribution of different lithological categories is used, and argmax is used to select the category with the highest probability value as the lithological label for the point cloud points. DGCNN ensures the accuracy of classification results by dynamically adjusting neighborhood features, and performs particularly well in complex geological environments.

[0162] S2043. Map the lithology classification results to point cloud data points and output the three-dimensional volume and boundary information of each lithological region. Specifically, use the classified point cloud to calculate the spatial distribution range of each lithology category, extract boundary points to define the region boundaries, and calculate the volume information of each lithological region based on the classification results.

[0163] Ultimately, the output lithology classification results and regional volume provide accurate data support for subsequent engineering quantity calculations, and also provide an important basis for assigning attribute values ​​to the geological BIM model.

[0164] S205. Register the 3D point cloud data after lithological differentiation. Use the iterative nearest point algorithm to spatially align the 3D point cloud data from multiple scans. At the same time, unify and fuse the coordinates with the borehole exploration data to ensure that the point cloud, image data and geological borehole information are unified in the same coordinate system, eliminate spatial errors and integrate surface and subsurface information.

[0165] In this embodiment, an iterative nearest-point algorithm is used to perform precise registration of point cloud data and borehole exploration data from multiple scans. The registration process first employs a global feature matching method (such as coarse registration based on feature points) to initially align the point clouds from multiple scans. Then, the iterative nearest-point algorithm is used to gradually minimize the distance error between point clouds, achieving high-precision registration through multiple iterations.

[0166] Furthermore, to ensure the consistency of multi-source data, this application embodiment also unifies the coordinate system and spatially fuses point cloud data and borehole exploration data, integrating surface geometric features with subsurface lithological information. Registration effectively eliminates spatial errors between multi-source data, making the three-dimensional information of the geological body more complete and providing an accurate data foundation for the subsequent construction of geological BIM models.

[0167] Furthermore, the specific operation of step S205 includes the following detailed process:

[0168] S2051. Select the set of control points P with significant geological features from the multiple scanned point cloud data. control And set a reference point cloud P ref and target point cloud P target Initial alignment is achieved through rigid transformation of the control points, and the coarse transformation matrix T is calculated. init The transformation matrix is ​​solved by minimizing the distance error between the control points of the reference point cloud and the target point cloud, as shown in the formula:

[0169]

[0170] Wherein, t(p i ) represents the transformation matrix t applied to point p in the reference point cloud. i Position after initial rigid transformation, q i This step positions the corresponding point in the target point cloud. This step completes the rough alignment of the point cloud, providing a foundation for subsequent precise registration.

[0171] S2052. Based on the coarse alignment results, the registration accuracy of the point cloud is further optimized using the Iterative Closest Point (ICP) algorithm. The ICP algorithm iteratively optimizes the transformation matrix T by repeatedly matching the nearest neighbor pairs in the point cloud. optUltimately, accurate registration of the two point clouds is achieved using the following formula:

[0172]

[0173] Wherein, T(p i ) is the target point after transformation, q i It is in the reference point cloud and p i The nearest neighbor points are used. Through multiple iterations, the errors between point clouds are gradually reduced to ensure high-precision registration of the point clouds.

[0174] S2053. Map the image data generated by oblique photography onto the point cloud surface, and calculate the two-dimensional coordinates (u,v) of each point in the image using the three-dimensional coordinates (x,y,z) of the point cloud and the projection matrix M of the image:

[0175]

[0176] Where M is the projection matrix of the image, and (u,v) are the corresponding positions of the point cloud points on the image. Through this process, the texture information of the image is mapped onto the surface of the point cloud, thereby enhancing the image texture of the point cloud surface and forming a textured point cloud, which provides support for the visualization of geological BIM models.

[0177] S2054. Using borehole location data as control points, align the point cloud data with the borehole exploration data to a unified coordinate system. Calculate the rigid transformation matrix E by minimizing the error between the point cloud control points and the borehole control points. align :

[0178]

[0179] Wherein, T(p j ) represents the transformed point cloud control point position, p j ′ These are the corresponding control point locations in the borehole data. Through error optimization, the coordinate consistency between the point cloud data and the borehole data can be ensured, further improving the registration accuracy of multi-source data.

[0180] S2055. Integrate the registered point cloud data, image textures, and borehole data into a unified 3D dataset.

[0181] The final integrated dataset contains surface geometric features, image texture information, and borehole lithology information, forming a complete three-dimensional geological information representation. This integrated 3D dataset provides comprehensive data support for subsequent geological BIM model construction and analysis.

[0182] S206. Using the pre-processed 3D point cloud data, image data, and borehole exploration data, construct a complete geological BIM model before excavation to express the 3D morphology, internal structure, and attribute information of the geological body.

[0183] After preprocessing the point cloud data, a complete geological BIM model is constructed by combining the processed point cloud data, image data, and borehole exploration data. During the construction process, the three-dimensional geometry of the geological body is first generated using the point cloud data to ensure high-precision representation of the model's surface features. Then, based on lithological differentiation results, lithological distribution and related physical properties (such as density and permeability) are assigned to the model to enrich its internal attribute information. Finally, by integrating surface point cloud data and underground borehole data, a geological BIM model with complete three-dimensional spatial representation is formed. The final geological BIM model not only displays the external morphology of the geological body but also achieves accurate representation of the geological body by combining lithological distribution and internal structure, providing comprehensive data support for subsequent engineering planning, construction management, and dynamic monitoring.

[0184] Specifically, the detailed operation of step S206 includes the following process:

[0185] S2061. Import the preprocessed point cloud data and borehole exploration data into BIM modeling software (such as Revit). This application does not make specific limitations on this.

[0186] Point cloud data is primarily used to construct 3D surface models, containing geometric information of characteristic areas such as slopes and toes, while borehole exploration data provides layered information on subsurface strata, particularly the depth and spatial distribution of lithology. The combination of point cloud and borehole exploration data ensures that the geological model simultaneously encompasses complete information from both the surface and subsurface, providing comprehensive data support for further model construction.

[0187] S2062. Extract the geometric features of the land surface, including slope, toe, and fault areas, from point cloud data, and generate a digital elevation model (DEM). Combine this with the stratigraphic depth information from borehole data to establish a subsurface stratification model.

[0188] Specifically, borehole data provides the lithological boundaries of each stratum, while the point cloud surface model provides the geometric reference for the surface. These two are then docked and merged within a coordinate system to form a unified three-dimensional geological model that integrates the surface and subsurface. This model can fully represent the surface geometry and subsurface geological structure, providing a reliable geometric and geological foundation for engineering planning and analysis.

[0189] S2063. Add necessary geological attribute information to the generated BIM model to distinguish the excavation volume of different lithologies and strata.

[0190] First, the lithological properties of the strata (such as soil, sandstone, shale, etc.) are bound to each layer of the model, ensuring that each layer has a clear lithological definition. Second, based on the geological exploration report, physical parameters (such as density, volume, permeability, etc.) related to the excavation project requirements are added. This attribute information can intuitively reflect the engineering characteristics of different strata, ensuring that the BIM model has high accuracy and practicality in subsequent engineering quantity calculations and analyses.

[0191] S2064. Output a complete geological BIM model before excavation for subsequent engineering quantity calculations and construction planning.

[0192] The final model not only includes geometric information of the surface and subsurface but also integrates geological attributes and engineering parameters, providing comprehensive data support for excavation projects. This model possesses clear surface features and accurate subsurface lithological distribution, enabling direct use for detailed calculations of engineering quantities, construction simulation, and engineering decision-making. The application of geological BIM models can significantly improve the planning efficiency and construction management level of geological engineering projects.

[0193] S3. After the excavation work is completed, point cloud data of the same area is obtained using oblique photography technology, and the geological model is constructed again. In order to address the problem of unclear and sparse point cloud identification in the slope foot area, the data accuracy is optimized using the inverse distance weighted interpolation algorithm to construct the geological BIM model after excavation.

[0194] In this embodiment, after the excavation is completed, oblique photogrammetry is used to collect data on the excavated area again to obtain updated point cloud data, and the geological model is reconstructed. Furthermore, to address the issues of unclear and sparse point cloud data in the slope toe area, an inverse distance weighted interpolation algorithm is used to optimize the point cloud data. Finally, using the optimized point cloud data and combining it with the pre-excavation geological BIM model, the post-excavation geological BIM model is reconstructed.

[0195] The final post-excavation geological BIM model accurately displays the three-dimensional topographic changes after excavation, including regional features such as slopes, toes, and faults, while also containing high-precision surface and subsurface information. Through dynamic model updates, changes before and after excavation can be compared in real time, providing reliable support for project construction management, excavation effect evaluation, and geological analysis.

[0196] Specifically, step S3 includes the following steps:

[0197] S301. After the excavation work is completed, UAV oblique photogrammetry technology is used again to acquire multi-view images of the same area, including vertical and oblique images, to ensure that the image data covers the key features of the entire excavation area. During the acquisition process, special attention is paid to the slope toe, slope surface, and other complex geological areas to ensure high resolution and comprehensive coverage of the image data. By importing the multi-view image data into modeling software (such as PIX4D), and after image stitching and reconstruction processing, high-precision 3D point cloud data of the excavated area is generated.

[0198] For detailed procedures, please refer to steps S101-S103, which will not be repeated in this application.

[0199] S302. Based on the three-dimensional point cloud data of the geological body after excavation obtained in step S101, for areas with sparse point clouds, the elevation values ​​of missing points are filled in using the inverse distance weighted interpolation algorithm. For the problem of unclear point cloud identification in the slope foot area, the interpolation strategy and parameters are adjusted by combining inverse distance interpolation with local trend fitting method.

[0200] Furthermore, the specific operation of step S302 includes the following detailed process:

[0201] S3021. Perform density analysis on the three-dimensional point cloud data of the excavated geological body to identify sparse areas.

[0202] For each point p in the point cloud data i Based on the number of points in its neighborhood |N i Perform statistical analysis:

[0203] |N i |={p j ∣∥p i -p j ∥≤r}

[0204] Where, |N i | Represents point p i The number of points within the neighborhood of p, where r is the user-defined neighborhood radius, typically 0.5 meters. i If the number of points in its neighborhood is less than the preset number of points, then determine p. i The region in question is sparse. This analysis ensures the accurate identification of sparse regions, providing a target area for subsequent interpolation processing.

[0205] In one possible embodiment, if |N i If the value is less than 10, then p is determined. i The area in question is a sparse region.

[0206] 3022. For sparse regions, this application embodiment uses the inverse distance weighted interpolation (IDW) algorithm to complete the elevation values ​​Z(x,y) of missing points, expressed as:

[0207]

[0208] In the formula, Z i It is a neighboring point p i The known elevation value, d i Let be the distance between the interpolation point and its neighboring points, and p be the distance weighting exponent with a value of 2. This algorithm performs continuous completion of the point cloud in sparse regions, ensuring that the generated elevation model has high accuracy.

[0209] 3023. From the point cloud data of the slope toe region obtained in step S2033, mark key areas with unclear feature representations, especially boundary points or high curvature points. These areas are then extracted and marked to support subsequent trend analysis and interpolation optimization. These feature points contain the geometric characteristics of the region boundaries and are key data points reflecting important geometric changes in the slope toe region.

[0210] 3024. Within the marked slope toe region, select feature points (x) within the neighborhood of the sparse region. i ,y i ,z i And a local trend equation is established using linear regression fitting method, the formula is:

[0211] z = ax + by + c

[0212] Where a, b, and c are the fitting parameters. The linear trend equation can provide the overall height change trend of the region, providing additional reference information for the interpolation algorithm and improving the interpolation accuracy.

[0213] 3025. Based on the local trend fitting results, the parameters of the inverse distance weighted interpolation algorithm are optimized. In sparse regions, the interpolation radius r is adjusted to 0.3 meters, and the number of interpolation neighborhood points is increased to 20. Simultaneously, the trend fitting results are combined with the IDW interpolation results to generate the final interpolated elevation value according to the weights.

[0214] In this embodiment, the elevation values ​​fitted by the trend and the elevation values ​​from the inverse distance weighted interpolation are assigned weights of 0.6 and 0.4, respectively. This step, combining the trend fitting and interpolation results, can improve the expressive power of sparse region boundary features, especially significantly optimizing model accuracy in the slope toe region.

[0215] 3026. After interpolation, the neighborhood mean smoothing method is used to optimize the interpolation region, thereby eliminating abruptness in the interpolation region, reducing the dispersion between interpolation points, and making the terrain transition more natural. The smoothing formula is:

[0216]

[0217] in, The smoothed point position, p j Points within the neighborhood, |N i | represents the number of neighboring points. The mean smoothing method significantly reduces abruptness caused by computational errors in the interpolation region, resulting in a smoother and more continuous optimized terrain.

[0218] 3027. After completing interpolation and optimization, compare the elevation values ​​of the interpolated region with those of the original 3D point cloud data of the excavated geological body. Verify the accuracy of the interpolated region by calculating the mean square error (MSE). The formula for calculating the mean square error is:

[0219]

[0220] in, The elevation value of the interpolation point. represents the actual elevation value in the original point cloud data, and n is the number of verification points. By calculating the mean square error, it can be ensured that the accuracy of the interpolation optimization area meets the requirements of geological modeling, providing reliable elevation data support for the subsequent construction of geological BIM models.

[0221] S303. The optimized point cloud data is fused with the borehole data before excavation, and the surface point cloud and underground borehole information are integrated through a unified coordinate system. The point cloud data provides the latest geometric morphology of the surface after excavation, while the borehole data reflects the lithological distribution and depth information of the underground strata.

[0222] Using Revit modeling software, point cloud data is used to generate a 3D surface model, including the excavated slope, toe, and other key areas. Simultaneously, a subsurface geological stratification model is constructed using borehole data, comprehensively representing both surface and subsurface geological features within the same model. By combining the model's geometric and geological attribute information, accurate representation of the excavated geological features is achieved.

[0223] The resulting post-excavation geological BIM model not only visually displays the three-dimensional morphological changes of the excavation area but also accurately reflects geological characteristics and their attributes. This model provides high-precision data support for construction management, excavation volume calculation, and geological change analysis, helping to achieve more efficient and scientific management in engineering planning and construction decisions.

[0224] S4. Divide the geological BIM model before excavation into regions based on lithology, and calculate the elevation difference pixel by pixel for the digital elevation models before and after excavation using a robust elevation difference calculation method based on neighborhood windows.

[0225] In this embodiment, the geological BIM model before excavation is divided into regions based on lithological information. Combining lithological attributes (such as soil layers, sandstone, shale, etc.) from borehole data and spatial geometric features in the model, the model is divided into different lithological regions. Each region is divided based on geological attributes and spatial distribution characteristics, ensuring that the model accurately reflects the layered structure and lithological boundaries of the geological body, providing precise basic data for subsequent analysis. Then, for the digital elevation models before and after excavation, a robust elevation difference calculation method based on neighborhood windows is used to calculate the elevation difference pixel-by-pixel between the two stages of the model.

[0226] By calculating the elevation difference pixel by pixel, an elevation change map of the excavation area can be generated, clearly showing the spatial distribution characteristics of the excavation volume. This step provides accurate basic data support for subsequent excavation volume calculation and geological change analysis, and also provides important reference for construction planning and decision optimization.

[0227] Specifically, step S4 also includes:

[0228] S401. Based on the three-dimensional point cloud data before excavation, which has been differentiated by lithology, mark the different lithological areas.

[0229] In this embodiment, based on the lithology classification results of step S204, different lithological regions in the geological BIM model are labeled. By combining borehole exploration data and point cloud classification results, clear labels are assigned to each lithological region (such as sandstone, shale, soil layers, etc.) in the model. These labels not only distinguish different geological regions but also provide accurate classification basis for subsequent elevation difference calculations and excavation volume statistics, ensuring that the elevation difference calculation results can be classified and summarized according to lithology, and providing data support for analyzing the excavation characteristics of each lithological region.

[0230] S402. Based on the labeled lithological area, using the elevation data generated before and after excavation, the elevation difference is calculated pixel by pixel using the neighborhood window method. For each pixel, the elevation data of the previous and subsequent stages are compared within the neighborhood window, and the minimum absolute value is selected as the true elevation difference. At the same time, an elevation difference threshold is set to mark the changed points or unchanged points, and the pixel-level results are mapped to the lithological area.

[0231] In this embodiment, based on the labeled lithological area, the elevation difference is calculated pixel-by-pixel using the digital elevation model (DEM) generated before and after excavation, employing a neighborhood window method. For each pixel, a neighborhood window is defined, and the elevation data before and after excavation are compared within the window range. The minimum absolute value of the elevation difference within the neighborhood is calculated and used as the true elevation difference for that pixel.

[0232] This method avoids erroneous elevation difference calculations caused by noise or outliers. It sets elevation difference thresholds and marks changed and unchanged points; for example, pixels with elevation differences exceeding the threshold are considered changed, otherwise marked as unchanged. Finally, the pixel-level elevation difference results are mapped back to lithological regions, forming elevation change detection results based on lithology classification, achieving high-precision geological change detection. This step clearly reflects the elevation change characteristics of different lithological regions, providing reliable data support for excavation volume calculation and geological analysis.

[0233] Furthermore, the specific operation of step S402 includes the following detailed process:

[0234] S4021. Set an appropriate neighborhood window size according to the geological features, select the center pixel of the window, locate its coordinates in the post-excavation stage elevation data, and extract all pixel values ​​of the corresponding window range in the pre-excavation stage elevation data.

[0235] Specifically, based on the local variation characteristics of the geological body, an appropriate neighborhood window size (2w+1) is set, with a window side length of w=3, corresponding to a 7×7 grid range. When selecting the window size, it is necessary to ensure that it can cover the local feature variations of the geological body while avoiding interference from too much irrelevant data. For each window's center pixel (i,j), its coordinates are located in the post-excavation stage elevation data, and all pixel values ​​within the corresponding window range are extracted from the pre-excavation stage elevation data. This process provides data support for pixel-by-pixel elevation difference calculation while ensuring accurate capture of local variations.

[0236] S4022. For each pixel within the window, calculate the elevation difference |d| between the center pixel (i,j) of the window after excavation and its neighboring pixels (p,q) before excavation. Arter (i,j)-d Before (p,q)|. Within the window, compare all elevation differences one by one, and take the smallest absolute value as the true elevation difference d of the current pixel. D (i,j):

[0237] d D (i,j)=min{|d After (i,j)-d Before (p,q)|}

[0238] Where, d D (i,j) is the elevation difference of pixel (i,j), d After (i,j) is the elevation value of the center pixel in the later stage DEM, d Before(p,q) represents the elevation value of pixel (p,q) within the window in the previous stage DEM, where (p,q) are the pixel coordinates within the window. This method can effectively reduce interference from noise or abnormal data, and obtain more robust elevation difference values.

[0239] S4023, By setting the elevation difference threshold T Height (Taking a value of 0.5 meters) Mark changes in the elevation difference for each pixel. If the actual elevation difference of the current pixel is greater than or equal to the elevation difference threshold, it is marked as a changed point; otherwise, it is marked as an unchanged point. The formula for the change marking rule is as follows:

[0240]

[0241] This marking method can intuitively distinguish between changed and unchanged points within the excavation area, providing a basis for subsequent statistical analysis of lithological areas.

[0242] S4024. Calculate the proportion of change points within each lithological region and the percentage of changed pixels to total pixels. If the proportion of change points within a lithological region reaches a preset threshold (e.g., 30%), the region is marked as a changed region; if the proportion is below the threshold, it is marked as an unchanged region. Unchanged regions are not included in subsequent excavation volume calculations. This method effectively reduces computational load while focusing on areas of significant change, improving analysis efficiency and accuracy.

[0243] Through the above steps, the pixel-by-pixel elevation difference calculation and lithological region change marking are completed, providing high-precision basic data support for subsequent excavation volume calculation and change area analysis.

[0244] S5. Based on the elevation difference calculation results, calculate the volume change of each lithological zone to obtain the total engineering volume of the entire excavation area.

[0245] In order to accurately calculate the volume change within the excavation area, this application combines the elevation difference results to precisely divide each lithological area, and calculates the volume change of each area step by step based on regular grid cells, and finally calculates the total engineering volume of the entire excavation area.

[0246] The specific steps of step S5 are as follows:

[0247] S501. Based on the elevation difference calculation results, each lithological region is divided into regular grid units.

[0248] In this embodiment of the application, based on the elevation difference calculation results of steps S401 and S402, different lithological regions (such as sandstone, shale, etc.) are divided into uniform grid cells, and the side length a of the grid cell is defined.

[0249] The grid size needs to be set according to the complexity of the terrain and the required computational accuracy; this application does not impose specific limitations on it. For example, for more complex terrain areas, the grid side length can be chosen to be smaller (e.g., 0.5 meters) to capture terrain variation features more precisely; while for flatter areas, the grid side length can be appropriately increased to simplify calculations. The main purpose of gridding is to discretize continuous terrain data, facilitating the calculation of volume changes in each cell, while ensuring that the complexity of the terrain within the region is fully represented.

[0250] S502. For each grid cell, calculate the volume change of the single grid cell based on the elevation difference and grid area.

[0251] In this embodiment of the application, for each grid cell, the elevation difference d is used. D (x,y) and the area A of the grid cells grid Calculate the volume change of a single grid cell. The volume calculation formula is:

[0252] V grid =A grid ·d D (x,y)

[0253] Among them, V grid V represents the volume change of a single mesh cell. grid =a 2 d is the area of ​​the grid cell. D (x, y) represents the elevation difference between the center points of the grid cells. A negative elevation difference indicates that excavation has occurred in the grid cell; a positive elevation difference indicates that accumulation has occurred in the grid cell. By calculating grid by grid, the distribution characteristics of volume changes within the region can be accurately captured.

[0254] S503. Accumulate the volume changes of all grid cells within each lithological region to calculate the total volume change for that region. The accumulation formula is:

[0255]

[0256] Among them, V rocktype This represents the total volume change of a specific lithological region (such as sandstone or clay), where n is the total number of grid cells within that region. For grid cells with negative elevation differences, the volume change is counted as excavation volume; for grid cells with positive elevation differences, the volume change is counted as accumulation volume.

[0257] This method of regional cumulative calculation can refine the volume change to the level of each lithological region, helping to analyze the specific contribution of different geological regions to the excavation project, while providing accurate volume change results to support subsequent excavation management and analysis.

[0258] S504. Summarize the volume changes of all lithological zones and calculate the total excavation volume for the entire excavation area. The final total excavation volume is obtained by summing the volume changes of each lithological zone:

[0259]

[0260] Among them, V total This represents the total volume of work in the entire excavation area, including the sum of excavated and stockpiled volumes. The summarized data comprehensively reflects the volume changes throughout the excavation project, facilitating overall assessment of the project volume and planning and management of the excavation work.

[0261] Through the above step-by-step calculations, volume changes can be precisely refined to the grid unit, gradually accumulated across lithological regions, and ultimately resulting in a total project volume statistic for the entire excavation area. This project volume assessment method based on elevation difference calculations not only ensures the accuracy of the calculation results but also provides detailed information on the specific variation characteristics of different lithological regions, offering comprehensive data support for construction cost estimation, resource allocation, and project schedule planning.

[0262] To achieve the above embodiments, this application also proposes a geological BIM modeling and engineering quantity calculation device based on point cloud sparse optimization. Figure 2 This is a structural schematic diagram of a geological BIM modeling and engineering quantity calculation device based on point cloud sparse optimization, provided as an embodiment of this application. Figure 2 As shown, the device includes:

[0263] The data acquisition module 100 is used to acquire geological body image data and three-dimensional point cloud data using oblique photography technology, and to associate and combine the three-dimensional point cloud data with borehole exploration data.

[0264] The geological model building module 200 before excavation is used to preprocess the three-dimensional point cloud data. Using the preprocessed three-dimensional point cloud data, image data and borehole exploration data, a geological BIM model before excavation is built. The data preprocessing includes denoising, classification, segmentation, lithology differentiation and registration.

[0265] The post-excavation geological model reconstruction module 300 is used to acquire point cloud data of the same area using oblique photography technology after the excavation work is completed, and to rebuild the geological model again. In addition, it uses the inverse distance weighted interpolation algorithm to optimize the data accuracy and build the post-excavation geological BIM model to address the problems of unclear and sparse point cloud identification in the slope foot area.

[0266] The elevation difference calculation module 400 is used to divide the geological BIM model before excavation into regions according to lithology, and to calculate the elevation difference of the geological BIM model before and after excavation pixel by pixel using a robust elevation difference calculation method based on neighborhood windows.

[0267] The 500 module for calculating engineering quantities is used to calculate the volume change of each lithological zone based on the elevation difference calculation results, and to obtain the engineering quantity of the entire excavation area.

[0268] In the foregoing descriptions of the embodiments, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0269] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0270] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0271] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0272] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this application can be achieved, and this is not limited herein.

[0273] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A geological BIM modeling and quantity calculation method based on point cloud sparsity optimization, characterized in that, Includes the following steps: Oblique photography technology is used to acquire geological body image data and three-dimensional point cloud data, and the three-dimensional point cloud data is correlated and combined with borehole exploration data; The three-dimensional point cloud data is preprocessed. Using the preprocessed three-dimensional point cloud data, image data and borehole exploration data, a geological BIM model is constructed before excavation. The data preprocessing includes denoising, classification, segmentation, lithology differentiation and registration. After the excavation work was completed, oblique photography technology was used to obtain point cloud data of the same area, and the geological model was reconstructed. In order to address the problem of unclear and sparse point cloud identification in the slope foot area, the inverse distance weighted interpolation algorithm was used to optimize the data accuracy and construct the geological BIM model after excavation. Based on lithology, the geological BIM model before excavation is divided into regions. For the digital elevation models before and after excavation, a robust elevation difference calculation method based on neighborhood windows is used to calculate the elevation difference of the geological BIM models before and after excavation pixel by pixel. Based on the elevation difference calculation results, the volume change of each lithological zone is calculated to obtain the total engineering volume of the entire excavation area; After the excavation is completed, oblique photogrammetry is used to acquire point cloud data of the same area, and a geological model is constructed again. To address the issues of unclear and sparse point cloud identification in the slope toe area, an inverse distance weighted interpolation algorithm is used to optimize data accuracy, thus constructing a post-excavation geological BIM model, including: After the excavation work was completed, the same area was once again captured by UAV oblique photography technology from multiple perspectives, and the three-dimensional point cloud data of the geological body after excavation was obtained by using the data from the newly acquired multi-perspective images. Based on the 3D point cloud data of the geological body after excavation, inverse distance weighted interpolation algorithm is used to fill in the elevation values ​​of missing points in areas with sparse point clouds. In addition, for the problem of unclear and sparse point cloud identification in the slope foot area, inverse distance interpolation combined with local trend fitting method is used to adjust the interpolation strategy and parameters. The optimized 3D point cloud data was combined with the borehole exploration data before excavation, and the geological BIM model after excavation was constructed using Revit modeling software. Based on the 3D point cloud data of the excavated geological body, for areas with sparse point clouds, an inverse distance weighted interpolation algorithm is used to complete the elevation values ​​of missing points. Furthermore, to address the issues of unclear and sparse point cloud identification in the slope toe area, an inverse distance interpolation combined with a local trend fitting method is used to adjust the interpolation strategy and parameters, including: Density analysis was performed on the 3D point cloud data of the excavated geological body to identify sparse areas. Specifically, for each point... If the number of points in its neighborhood is less than the preset number of points, then determine... The area in question is a sparse region. For sparse regions, inverse distance weighted interpolation algorithm is used to complete the elevation values ​​of missing points. The expression is: In the formula, Neighboring points The known elevation value, The distance between the interpolation point and its neighboring points. Distance weighting index; From the point cloud data of the slope toe area obtained based on classification before excavation, key areas with sparse or unclear feature representation are marked. Within the marked slope toe area, feature points in the neighborhood of sparse regions are selected, and a local trend equation is established using a linear regression fitting method. Based on the fitted trend results obtained after solving the local trend equation, the parameters of the inverse distance interpolation are optimized, and the preliminary elevation values ​​of the trend fitting are combined with the interpolation results of the inverse distance weighted algorithm to generate the final interpolated elevation values ​​according to the weights. After interpolation is completed, the neighborhood mean smoothing method is used to eliminate the abruptness of the interpolation region; After interpolation and optimization are completed, the elevation values ​​of the interpolated area are compared with those of the original three-dimensional point cloud data of the excavated geological body. The accuracy of the interpolated area is verified by calculating the mean square error.

2. The method according to claim 1, characterized in that, The process of acquiring geological body image data and 3D point cloud data using oblique photography technology, and then correlating and combining the 3D point cloud data with borehole exploration data, includes: Multi-view image data of geological bodies are collected using UAV oblique photography technology, including vertical and oblique images. Based on PIX4D software, the acquired multi-view image data is imported and processed. Dense point cloud data is generated through the software's automated workflow, and the spatial geometric information of the geological body is extracted to form three-dimensional point cloud data containing high-precision three-dimensional coordinates. Collect and integrate borehole exploration data to establish a database of internal geological structure information, and associate it with the three-dimensional point cloud data. The borehole exploration data includes borehole location, stratum depth, lithological distribution, and physical properties.

3. The method according to claim 2, characterized in that, The process of preprocessing the 3D point cloud data, and using the preprocessed 3D point cloud data, image data, and borehole exploration data to construct a geological BIM model before excavation, includes: The three-dimensional point cloud data is denoised by using a statistical filtering algorithm to remove isolated points, noise points, and irregular point clusters, ensuring the smoothness and accuracy of the point cloud data. Based on a multi-layer dynamic graph convolutional neural network, the denoised point cloud data is classified, and geological features, vegetation, buildings and noise are extracted from the point cloud according to geometric features, providing a basis for annotation and classification for subsequent removal of non-geological feature point clouds. Based on the classified 3D point cloud data, deep feature learning is used to segment it, extract feature regions related to geological bodies, and remove point cloud data that are not related to geological bodies. By differentiating the lithology of the pure geological point cloud data generated after segmentation, and combining borehole exploration data and deep learning classification models, the point cloud data of the geological body is refined into different lithological regions. The 3D point cloud data after lithological differentiation is registered, and the spatial alignment of the 3D point cloud data from multiple scans is performed by the iterative nearest point algorithm. At the same time, the coordinates are unified and fused with the borehole exploration data to ensure that the point cloud, image data and geological borehole information are unified in the same coordinate system, thereby eliminating spatial errors and integrating surface and subsurface information. Using the preprocessed 3D point cloud data, image data, and borehole exploration data, a complete geological BIM model is constructed to represent the 3D morphology, internal structure, and attribute information of the geological body.

4. The method according to claim 3, characterized in that, The preprocessed 3D point cloud data, image data, and borehole exploration data are used to construct a complete pre-excavation geological BIM model to represent the 3D morphology, internal structure, and attribute information of the geological body, including: Import the preprocessed 3D point cloud data and borehole exploration data into BIM modeling software; Geometric features of the land surface, including slope, toe, and fault areas, are extracted from the three-dimensional point cloud data, and a digital elevation model (DEM) of the land surface is generated. A subsurface stratification model is established by combining the borehole exploration data and docking the stratum depth information in the borehole with the point cloud surface model to form a unified three-dimensional geological BIM model of the surface and subsurface. Add necessary geological attributes to the generated BIM model to distinguish the excavation volume of different lithologies and strata. The geological attributes include the lithology of the strata and the density and volume parameters related to the excavation requirements. Output a complete geological BIM model before excavation.

5. The method according to claim 4, characterized in that, The process involves dividing the geological BIM model before excavation into regions based on lithology, and then calculating the pixel-by-pixel elevation difference of the digital elevation models before and after excavation using a robust elevation difference calculation method based on neighborhood windows. This includes: Based on the three-dimensional point cloud data before excavation, which has been differentiated by lithology, different lithological regions are labeled. Based on the labeled lithological area, the elevation difference is calculated pixel by pixel using the elevation data generated before and after excavation and the neighborhood window method. For each pixel, the elevation data of the previous and subsequent stages are compared within the neighborhood window, and the minimum absolute value is selected as the true elevation difference. At the same time, an elevation difference threshold is set to mark the changed points or unchanged points, and the pixel-level results are mapped to the lithological area.

6. The method according to claim 5, characterized in that, Based on the labeled lithological area, using elevation data generated before and after excavation, the elevation difference is calculated pixel-by-pixel using a neighborhood window method. For each pixel, the elevation data before and after the excavation are compared within the neighborhood window, and the minimum absolute value is selected as the true elevation difference. Simultaneously, an elevation difference threshold is set to mark changed or unchanged points, and the pixel-level results are mapped to the lithological area, including... Set an appropriate neighborhood window size based on the geological features, select the center pixel of the window, locate its coordinates in the post-excavation stage elevation data, and extract all pixel values ​​of the corresponding window range in the pre-excavation stage elevation data. For each pixel within the window, calculate the elevation difference between the center pixel after excavation and the neighboring pixels before excavation, and take the minimum absolute value as the true elevation difference of the current pixel; Each pixel is marked as changed by setting an elevation difference threshold. If the actual elevation difference of the current pixel is greater than or equal to the elevation difference threshold, it is marked as a changed point; otherwise, it is marked as an unchanged point. The proportion of change points in each lithological region is statistically analyzed, and the proportion of change pixels to total pixels is calculated. If the proportion of change points in a certain lithological region reaches a preset threshold, the region is marked as a change region; otherwise, the region is marked as an unchanged region.

7. The method according to claim 6, characterized in that, The volume change of each lithological zone is calculated based on the elevation difference to obtain the total engineering volume of the entire excavation area, including: Based on the elevation difference calculation results, each lithological region is divided into regular grid units; For each grid cell, the volume change of a single grid cell is calculated based on the elevation difference and grid area; The volume change of all grids within each lithological region is summed to obtain the volume change of that region. For regions with negative elevation differences, the volume change is recorded as excavation volume; for regions with positive elevation differences, the volume change is recorded as deposition volume. By summarizing the volume change data of all lithological areas, the total engineering volume of the entire excavation area is calculated.

8. A geological BIM modeling and quantity calculation device based on point cloud sparse optimization, characterized in that, include: The data acquisition module is used to acquire geological body image data and three-dimensional point cloud data using oblique photography technology, and to associate and combine the three-dimensional point cloud data with borehole exploration data; The geological model construction module before excavation is used to preprocess the three-dimensional point cloud data. Using the preprocessed three-dimensional point cloud data, image data and borehole exploration data, a geological BIM model before excavation is constructed. The data preprocessing includes denoising, classification, segmentation, lithology differentiation and registration. The post-excavation geological model reconstruction module is used to acquire point cloud data of the same area using oblique photography technology after the excavation work is completed, and to rebuild the geological model again. In addition, it uses an inverse distance weighted interpolation algorithm to optimize the data accuracy and build a post-excavation geological BIM model to address the problems of unclear and sparse point cloud identification in the slope foot area. The elevation difference calculation module is used to divide the geological BIM model before excavation into regions based on lithology, and to calculate the elevation difference pixel by pixel for the digital elevation models before and after excavation using a robust elevation difference calculation method based on neighborhood windows. The quantity calculation module is used to calculate the volume change of each lithological zone based on the elevation difference calculation results, and obtain the quantity of work for the entire excavation area. After the excavation is completed, oblique photogrammetry is used to acquire point cloud data of the same area, and a geological model is constructed again. To address the issues of unclear and sparse point cloud identification in the slope toe area, an inverse distance weighted interpolation algorithm is used to optimize data accuracy, thus constructing a post-excavation geological BIM model, including: After the excavation work was completed, the same area was once again captured by UAV oblique photography technology from multiple perspectives, and the three-dimensional point cloud data of the geological body after excavation was obtained by using the data from the newly acquired multi-perspective images. Based on the 3D point cloud data of the geological body after excavation, inverse distance weighted interpolation algorithm is used to fill in the elevation values ​​of missing points in areas with sparse point clouds. In addition, for the problem of unclear and sparse point cloud identification in the slope foot area, inverse distance interpolation combined with local trend fitting method is used to adjust the interpolation strategy and parameters. The optimized 3D point cloud data was combined with the borehole exploration data before excavation, and the geological BIM model after excavation was constructed using Revit modeling software. Based on the 3D point cloud data of the excavated geological body, for areas with sparse point clouds, an inverse distance weighted interpolation algorithm is used to complete the elevation values ​​of missing points. Furthermore, to address the issues of unclear and sparse point cloud identification in the slope toe area, an inverse distance interpolation combined with a local trend fitting method is used to adjust the interpolation strategy and parameters, including: Density analysis was performed on the 3D point cloud data of the excavated geological body to identify sparse areas. Specifically, for each point... If the number of points in its neighborhood is less than the preset number of points, then determine... The area in question is a sparse region. For sparse regions, inverse distance weighted interpolation algorithm is used to complete the elevation values ​​of missing points. The expression is: In the formula, Neighboring points The known elevation value, The distance between the interpolation point and its neighboring points. Distance weighting index; From the point cloud data of the slope toe area obtained based on classification before excavation, key areas with sparse or unclear feature representation are marked. Within the marked slope toe area, feature points in the neighborhood of sparse regions are selected, and a local trend equation is established using a linear regression fitting method. Based on the fitted trend results obtained after solving the local trend equation, the parameters of the inverse distance interpolation are optimized, and the preliminary elevation values ​​of the trend fitting are combined with the interpolation results of the inverse distance weighted algorithm to generate the final interpolated elevation values ​​according to the weights. After interpolation is completed, the neighborhood mean smoothing method is used to eliminate the abruptness of the interpolation region; After interpolation and optimization are completed, the elevation values ​​of the interpolated area are compared with those of the original three-dimensional point cloud data of the excavated geological body. The accuracy of the interpolated area is verified by calculating the mean square error.

Images