A point cloud data management and web rapid recognition visualization method

Abstract

This application provides a method for point cloud data management and rapid Web identification and visualization, relating to the field of point cloud data processing technology. The method involves establishing a point cloud database; performing point cloud similarity clustering on the point cloud database, and then performing adaptive denoising processing based on the similarity clustering results to establish a denoised point cloud set; establishing a scale-space pyramid, using the similarity clustering results to call and match the scale-space pyramid, and reconstructing a position multi-scale model based on the call matching results; performing feature extraction on the denoised point cloud set using the position multi-scale model to establish a tunnel structure feature set; and using the tunnel structure feature set to perform tunnel visualization fitting and create a Web visualization model. This application solves the technical problem in the prior art of low Web visualization quality due to the large amount of tunnel point cloud data and the presence of noise and redundant information, significantly improving the processing efficiency and rendering performance accuracy of tunnel point cloud data on the Web.

Description

Technical Field

[0001] The present application relates to the technical field of point cloud data processing, and in particular to a point cloud data management and Web rapid identification and visualization method. Background Art

[0002] As tunnel projects continue to grow in scale and complexity, point cloud data, as high-precision three-dimensional spatial data, has become a crucial tool for tunnel structure inspection, monitoring, and analysis. Point cloud data, acquired through technologies like laser scanning and lidar, can accurately reflect the geometry and structural characteristics of tunnels.

[0003] Traditional point cloud data processing software mostly runs on a single machine, which has limitations in data sharing, collaborative work, and web-based use. With the development of 5G and WebGL technologies, 3D web applications have become a development trend. However, the massive scale of tunnel point cloud data creates performance bottlenecks for loading and rendering on the web side. Traditional rendering mechanisms often result in slow loading when processing millions or hundreds of millions of points, affecting the user experience. Secondly, the computing power of the web platform and the rendering performance of the browser are relatively weak. Especially when rendering complex tunnel structures, freezes or delays often occur, affecting real-time interaction and viewing effects. In addition, the noise and redundant information in the point cloud data also make the visualization effect unsatisfactory. Low-quality point cloud data may result in the inaccurate presentation of the details of the tunnel structure, which cannot meet the needs of modern tunnel engineering for efficient data processing and convenient visualization. Summary of the Invention

[0004] This application provides a point cloud data management and Web rapid identification and visualization method, which solves the technical problems in the existing technology such as slow Web loading and rendering jams and low visualization quality due to the huge amount of tunnel point cloud data and the presence of noise and redundant information in the data. It achieves the technical effect of significantly improving the processing efficiency and rendering performance accuracy of tunnel point cloud data on the Web, thereby enhancing the shareability and collaboration of tunnel point cloud data.

[0005] In view of the above problems, the present application provides a point cloud data management and Web rapid identification and visualization method, which includes: establishing a point cloud database, which is used to store tunnel point cloud data; performing point cloud similarity clustering on the point cloud database, and then performing adaptive denoising processing based on the similarity clustering results to establish a denoised point cloud set; establishing a scale space pyramid, using the similarity clustering results to call and match the scale space pyramid, and reconstructing a position multi-scale model based on the call matching results; performing feature extraction on the denoised point cloud set using the position multi-scale model to establish a tunnel structure feature set; and using the tunnel structure feature set to perform tunnel visualization fitting to create a Web visualization model.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] By clustering tunnel point cloud data stored in a point cloud database based on similarity, adaptively denoising the data using an appropriate denoising algorithm based on the similarity clustering results, and constructing a denoised point cloud, the team can reduce noise and irrelevant information, thereby improving the quality of the tunnel point cloud data. A scale-space pyramid is then constructed and matched using the similarity clustering results. A multi-scale positional model is then reconstructed based on the matching results. This allows for data analysis at different scales, ensuring that the point cloud data effectively represents the tunnel structure at various scales, providing more comprehensive information for subsequent operations such as feature extraction. Using the multi-scale positional model, feature extraction is performed on the denoised point cloud to construct a tunnel structural feature set, further refining the capture of tunnel structural features. This not only preserves the tunnel's geometric structure but also enhances the recognition of complex structural features. Based on the extracted tunnel structural feature set, visual fitting and modeling are performed, ultimately forming a web-based visualization model. This transforms complex tunnel point cloud data into an intuitive graphical display, making it easier for users to view and interact with the data through a web platform, understanding the details of the tunnel structure and improving the convenience and real-time nature of data display and interaction.

[0008] In summary, this application combines multiple links such as data storage, denoising, scale analysis, feature extraction and visual display to improve the processing efficiency and accuracy of tunnel point cloud data. The point cloud database provides structured storage support, and similarity clustering and denoising processing ensure the quality of the data. The matching of scale space pyramid and similarity clustering optimizes the performance of point cloud data at different scales, and the position multi-scale model enhances the accuracy of feature extraction. Finally, through the Web visualization model, tunnel structure data can be presented intuitively, achieving efficient, accurate and easy-to-interactive data display, making the management, analysis and display process of tunnel point cloud data more efficient and accurate, and providing strong support for the intelligent management of tunnel projects.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A flowchart of a point cloud data management and Web rapid identification visualization method provided in an embodiment of the present application.

[0011] Figure 2A schematic diagram of a process for performing cluster search of all nodes in a point cloud image using initial cluster points in a point cloud data management and Web rapid identification visualization method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The present embodiment provides a method for point cloud data management and rapid web recognition and visualization. After performing point cloud similarity clustering on tunnel point cloud data, adaptive denoising is performed based on the similarity clustering results. A scale-space pyramid is then established, and the scale-space pyramid is called and matched using the similarity clustering results to reconstruct a multi-scale position model. Feature extraction of the denoised point cloud is performed using the multi-scale position model to establish a tunnel structure feature set. This tunnel structure feature set is used to perform tunnel visualization fitting and create a web visualization model. The present embodiment solves the technical problems in the prior art of slow web loading, lag rendering, and low visualization quality caused by the large amount of tunnel point cloud data and the presence of noise and redundant information in the data. This significantly improves the processing efficiency and rendering performance accuracy of tunnel point cloud data on the web, thereby enhancing the shareability and collaboration of tunnel point cloud data.

[0013] like Figure 1 As shown, the embodiment of the present application provides a point cloud data management and Web rapid identification visualization method, the method comprising:

[0014] Step S1: establishing a point cloud database, which is used to store tunnel point cloud data.

[0015] Specifically, tunnel point cloud data refers to data obtained through technologies such as laser scanning, describing the coordinates of individual points on the tunnel surface in three-dimensional space. For example, scanning a tunnel wall with a laser scanner generates coordinate data for a large number of discrete points, which is point cloud data. A point cloud database is a database system used to store and manage tunnel point cloud data.

[0016] A point cloud database is created using a database management system (relational databases like MySQL or PostgreSQL, or non-relational databases specifically designed for 3D data processing like MongoDB). Interactive 3D laser scanners or other 3D data acquisition devices acquire tunnel point cloud data and store this data in a specific format (for example, by acquisition time, acquisition area, etc.). This centralized storage of tunnel point cloud data facilitates subsequent data management, query, and analysis, providing a foundation for the entire point cloud data processing process.

[0017] Step S2: After performing point cloud similarity clustering on the point cloud database, adaptive denoising is performed based on the similarity clustering results to establish a denoised point cloud set.

[0018] Specifically, the tunnel point cloud data in the point cloud database is clustered by point cloud similarity, and point cloud data with similar features (such as spatial position, geometric shape, etc.) are classified into one category. For example, point clouds of adjacent and similarly shaped areas on the tunnel wall can be clustered into one category. Based on the similarity clustering results, isolated points or abnormal points that do not conform to the clustering results are identified, the noise characteristics of the point cloud data in each cluster are analyzed, and a suitable denoising algorithm is selected to remove these points as noise points and establish a denoised point cloud set. For example, a voxel grid method or a spatial neighborhood search can be used to analyze the noise distribution, intensity, and type in the point cloud data, and direct removal of noise points can be used to clean up isolated noise points, and weighted denoising can be used to process edge noise points.

[0019] Through similarity clustering and denoising, the quality of point cloud data can be significantly improved. The denoised point cloud reduces the interference of irrelevant data, improves the accuracy and reliability of the data, and makes subsequent analysis and visualization more accurate.

[0020] Step S3: establishing a scale space pyramid, performing call matching on the scale space pyramid using the similarity clustering result, and reconstructing a position multi-scale model according to the call matching result.

[0021] Specifically, in point cloud processing, the scale-space pyramid can represent point cloud data at different levels of resolution, helping to process tunnel structure information at different levels of detail. The position multi-scale model is constructed based on the scale-space pyramid and combines the spatial position information of the point cloud. This model can reflect the changes in tunnel structure at different scales. A node distribution analysis is performed on the point cloud database to determine the spatial distribution of the point cloud data. Parameters such as the limiting scale and distribution coefficient are then determined to construct the scale-space pyramid. Based on similar clustering results, the point cloud data is matched to different scales of the scale-space pyramid. If a certain type of point cloud data in the clustering results spatially corresponds to a scale range in the scale-space pyramid, a matching relationship is established between the point cloud data and the scale-space pyramid based on this correspondence, generating a position multi-scale model.

[0022] The application of scale-space pyramids and multi-scale models enables tunnel point cloud data to be analyzed and visualized at different resolutions, accurately extracting information about the tunnel structure at all levels, preserving details while improving processing efficiency.

[0023] Step S4: performing feature extraction of the denoised point cloud using the position multi-scale model to establish a tunnel structure feature set.

[0024] Specifically, the tunnel structure feature set is a collection of features that can represent the tunnel structure (such as the tunnel's outline, the shape of key parts, etc.) extracted from the tunnel point cloud data after a series of processing. For example, the circular cross-sectional shape of the tunnel and the shape features of the arch can be included in this feature set. Based on the multi-scale position model, feature extraction algorithms (such as shape descriptor algorithms) are used to extract features from the denoised point cloud. For example, features such as the curvature of the tunnel wall and the cross-sectional shape of the tunnel are extracted, and these extracted features are organized into a tunnel structure feature set. Feature extraction can refine the key structural information of the tunnel, which helps to deeply analyze and understand the tunnel structure.

[0025] Step S5: Perform tunnel visualization fitting using the tunnel structure feature set to create a Web visualization model.

[0026] Specifically, the Web visualization model is a tunnel visualization model that can be viewed and interacted with on a web browser. Using a set of tunnel structural features, tunnel visualization is performed using technologies such as WebGL. For example, the 3D shape of the tunnel is drawn based on its cross-sectional shape characteristics, and the spatial layout of the model is determined based on the tunnel's orientation characteristics. These features are converted into a 3D model that can be displayed on a web browser, creating the Web visualization model. The Web visualization model allows users to conveniently view the 3D structure of the tunnel on a web browser without the need for specialized 3D software. This improves data sharing and interactivity, facilitating collaborative work and decision-making among tunnel engineering professionals.

[0027] Furthermore, in step S2 of the embodiment of the present application, point cloud similarity clustering is performed on the point cloud database, including:

[0028] Step S21: taking each point in the point cloud database as a node, calculating the similarity between nodes, and constructing a point cloud graph.

[0029] Step S22: Obtain the calibration segmentation result of the tunnel, use the calibration segmentation result to distribute initial clustering points in the point cloud map, use the initial clustering points to perform clustering search for all nodes in the point cloud map, and complete point cloud similarity clustering based on the clustering search results.

[0030] Specifically, when performing point cloud similarity clustering, each point in the point cloud database is first treated as a node. Methods such as Euclidean distance and cosine similarity are used to measure the similarity between points. The similarity between each node is then calculated, and a point cloud graph is constructed based on the similarity between nodes. In this graph, nodes represent points in the point cloud, and edges represent the similarity between points. The point cloud graph provides the topological structure of the point cloud data, providing an analytical basis for clustering point cloud data.

[0031] The calibration segmentation results are obtained by calibrating and segmenting the point cloud data using the geometric features (such as shape and size), geological features, construction features, etc. of the tunnel, and are used to determine the distribution of initial cluster points in the point cloud map. The initial cluster points refer to the points pre-selected during the clustering process to start the clustering algorithm. The tunnel point cloud data is preliminarily segmented according to the tunnel characteristics to obtain the calibration segmentation results. Based on the calibration segmentation results, the positions of the initial cluster points are determined in the point cloud map. These initial cluster points are usually selected in prominent areas or on the boundaries of the point cloud data. With the initial cluster points as the center, a cluster search is performed on all nodes in the point cloud map to find points similar to the initial cluster points. Based on the clustering search results, similar points are classified into the same category to complete the point cloud similarity clustering.

[0032] By calibrating the segmentation results, the positions of the initial clustering points can be determined more accurately, thereby guiding the clustering algorithm to more effectively identify and group similar point cloud data, providing a more accurate data set for subsequent denoising and feature extraction, and improving the accuracy and efficiency of clustering.

[0033] Further, such as Figure 2 As shown, in step S22 of the embodiment of the present application, the initial clustering points are used to perform a cluster search of all nodes in the point cloud graph, including:

[0034] Step S22-21: establishing a clustering fitness function, where the clustering fitness function is a function for performing clustering fitness calculation.

[0035] Step S22-22: After taking the initial cluster point as the cluster center, the cluster fitness function is used to calculate the cluster fitness within the preset search range of the cluster center. The cluster fitness calculation includes the distance calculation between the node and the cluster center, the cluster size calculation corresponding to the cluster center, and the similarity calculation.

[0036] Step S22-23: perform cluster expansion according to the cluster fitness calculation result and reconstruct the cluster cluster.

[0037] Step S22-24: perform cluster center search and update on the reconstructed clusters to establish a reconstructed cluster center.

[0038] Step S22-25: Execute search iterations to complete the cluster search of all nodes in the point cloud graph.

[0039] Specifically, the clustering fitness function is a function that evaluates the quality of clustering. This function measures the effectiveness of clustering based on clustering feature parameters such as the distance between the node and the cluster center, the size of the cluster, and the similarity between the nodes. It also evaluates the similarity between a certain point and the cluster center and the degree of cohesion of the cluster. A clustering fitness function is predefined to evaluate the effectiveness of each cluster in the subsequent clustering process. Exemplarily, this function can contain multiple sub-functions, each of which calculates the distance between the node and the cluster center, the size of the cluster, and the similarity between the nodes. The clustering fitness function is constructed by weighted averaging of multiple sub-functions.

[0040] A preset search range is set, centered around the initial cluster point. This range can be a fixed distance threshold or a dynamic range determined by the point cloud distribution of the current cluster center. Within the preset search range, the cluster fitness function is used to calculate the distance between each node and the cluster center, the size of the clusters, and the similarity between nodes.

[0041] According to the results of the clustering fitness function, the point cloud data within the cluster is gradually expanded. If the fitness value of some points is high, it means that these points are similar to the current cluster center and can be added to the cluster. By expanding the cluster within the preset search range, the cluster is reconstructed to form a new clustering result. Then, the cluster center of the reconstructed cluster is searched and updated to establish a reconstructed cluster center to ensure that the new cluster center can better represent the nodes in the cluster. The cluster is reorganized according to the reconstructed cluster center. The search iteration is performed to optimize the cluster center through multiple iterations until the cluster no longer changes significantly or the distribution of all nodes no longer changes, and the iteration is stopped.

[0042] Through this clustering process, the nodes in the point cloud data are effectively divided into multiple clusters with similar structures, each of which more accurately represents the structural characteristics of the tunnel. The newly added cluster center establishment and search iterations ensure that the clustering process is not limited to the initial cluster points, but can dynamically adapt to the distribution changes of unclustered points. This further improves the accuracy and robustness of clustering, further ensures the precise segmentation of tunnel point cloud data, improves the integrity and efficiency of data processing, and provides more reliable data support for subsequent denoising, feature extraction, and visual modeling.

[0043] Furthermore, steps S22-25 of the embodiment of the present application also include:

[0044] Step S22-251: Determine whether a cluster search stop event occurs. If a cluster search stop event occurs, generate a new instruction.

[0045] Step S22-252: Perform distribution analysis on the unclustered nodes according to the newly added instructions and establish a new cluster center.

[0046] Step S22-253: Continue searching iteratively based on the newly added cluster center.

[0047] Specifically, the cluster search stop event means that during the clustering process, the cluster center of some clusters has stabilized and no significant points are assigned to the cluster, or the search range of the cluster has reached the set threshold.

[0048] After each cluster expansion, the changes in the current cluster are evaluated to determine whether the conditions for stopping the search have been met. Specifically, if the distribution of the center point and points within a cluster has stabilized, and no new points have been clustered into the cluster, the search for the cluster has been completed, the cluster has reached a stable state, and its clustering quality cannot be improved by adding new nodes. At this point, the search for the current cluster can be stopped, the status of the currently completed cluster is recorded, and new instructions are generated for the unclustered nodes in preparation for subsequent clustering operations. This new instruction is used to guide further clustering of other unclustered nodes to add new clusters.

[0049] Unclustered nodes are nodes that have not yet been assigned to any clusters during the clustering process. After generating new instructions, a distribution analysis is performed on these unclustered points. Based on their spatial location, point cloud density, similarity, and other characteristics, the distribution characteristics of these points are determined to determine whether these unclustered nodes are sufficient to form one or more new clusters. Based on the distribution analysis results, one or more new cluster centers are selected. These new cluster centers are usually the center of gravity of the unclustered point cloud data or areas with high density.

[0050] Based on the newly added cluster center, the search continues for nodes similar to that center, performing cluster expansion. These unclusters are compared with the new cluster center using the cluster fitness function to expand the new clusters. After each cluster expansion, the cluster center is updated based on the location of the newly added points, and the cluster fitness is recalculated. This iterative cluster search process is repeated until all nodes in the point cloud are clustered.

[0051] The above steps can effectively process unclustered points by introducing cluster stop search events and new instructions, avoiding missing certain structural parts in the tunnel point cloud data, ensuring the integrity of the clustering process, and avoiding omissions or misclassifications.

[0052] Furthermore, in step S22 of the embodiment of the present application, obtaining the calibration and segmentation results of the tunnel includes:

[0053] Step S22-11: extracting geometric features of the tunnel and establishing a first segmentation constraint. The geometric features include shape and size features of the tunnel.

[0054] Step S22-12: Extract geological features of the tunnel and establish a second segmentation constraint.

[0055] Step S22-13: Perform temporal extraction of tunnel construction features and establish a third segmentation constraint.

[0056] Step S22 - 14 : establishing the calibrated segmentation result using the first segmentation constraint, the second segmentation constraint, and the third segmentation constraint.

[0057] Specifically, in point cloud data, geometric features usually include structural information such as the shape, size, curvature, and flatness of the tunnel. These features help identify the geometric boundaries of the tunnel, the shape of the pipe, etc. Geometric features are extracted from the tunnel point cloud data, and based on the extracted geometric features, the first segmentation constraint is established to guide how to segment the tunnel point cloud data into different regions. For example, different regions can be segmented based on features such as the diameter and curvature of the tunnel. Assuming that the diameter of the tunnel is stable within a certain range, the points in the point cloud data can be preliminarily segmented based on these stable geometric features to distinguish between areas such as the tunnel wall, ground, and top.

[0058] Geological characteristics refer to geological features of a tunnel, such as rock type, soil properties, and structural layers. The geological information in the point cloud data is analyzed to identify different geological structures and features. Based on the extracted geological features, a second segmentation constraint is established to effectively segment the tunnel point cloud data according to different geological layers. For example, a tunnel section may pass through different rock layers with varying density and hardness. By extracting geological features from the tunnel point cloud data, the different characteristics of hard rock layers and soft soil layers can be identified, thereby demarcating these areas in the point cloud data.

[0059] During construction, the tunnel's morphology changes over time. By extracting time-series data, we can derive tunnel characteristics at different stages of construction. For example, some areas may represent exposed rock after excavation, while others may represent areas with completed support structures. Based on the different construction stages, a third segmentation constraint is established to segment the point cloud data into regions that match the construction stage, effectively distinguishing changes in the point cloud data at different stages.

[0060] The aforementioned geometric, geological, and construction constraints are combined to generate a calibrated segmentation result. Each point cloud data point is assigned to an appropriate category based on these three constraints. This segmentation method not only reflects the tunnel's geometry but also distinguishes between different geological layers and construction stages, making the processing of tunnel point cloud data more scientific and rational, and better reflecting the tunnel's actual structure and construction status.

[0061] Furthermore, in step S3 of the embodiment of the present application, the scale space pyramid is established, including:

[0062] Step S31: performing node distribution analysis on the point cloud database and configuring limit scales, wherein the limit scales include a minimum scale and a maximum scale.

[0063] Step S32: Acquire the accuracy constraint of point cloud data management, and establish a distribution coefficient according to the accuracy constraint.

[0064] Step S33: constructing a scale space pyramid using the distribution coefficient and the limit scale.

[0065] Specifically, the spatial distribution of each node in the point cloud data is analyzed to evaluate its density, sparsity and other characteristics. By calculating the distance statistics between points, such as the average distance, the standard deviation of the distance, etc., the changes in the density of nodes in different areas can be evaluated. Through this analysis, we can understand which areas of the point cloud data have more concentrated points and which areas have more sparse points. According to the results of the node distribution analysis, the limit scale is configured. The limit scale is the upper and lower limits of the scale range considered in the multi-scale analysis process, including the minimum scale and the maximum scale. Among them, the minimum scale is the smallest scale in the processing process, which represents the scale of the smallest details or local structures in the point cloud. The maximum scale is the largest scale in the processing process, which represents the scale of the coarsest and most macroscopic structure in the point cloud.

[0066] Precision constraints for point cloud data management are predefined by the user based on specific application needs and data quality requirements. For example, certain high-precision measurement applications require point cloud data accuracy to the millimeter level; whereas, for more coarse-grained analyses, accuracy requirements may be reduced to the centimeter level. Based on the needs of the tunnel project, the corresponding point cloud data accuracy requirements are determined. A distribution coefficient is established based on the accuracy constraints. The distribution coefficient is a parameter determined by the accuracy constraints during the construction of the scale-space pyramid and determines the detail retention and distribution of the point cloud data at different scales. The distribution coefficient is inversely proportional to the accuracy constraint: the higher the accuracy requirement, the smaller the distribution coefficient; conversely, the lower the accuracy requirement, the larger the distribution coefficient.

[0067] A scale-space pyramid is constructed using the determined distribution coefficients and limiting scales. The scale-space pyramid processes point cloud data at different scales to create multiple layers, each corresponding to a different scale range. This hierarchical structure allows tunnel point cloud data to be analyzed at different levels, thereby obtaining feature information at different scales. The specific construction process involves resampling and processing the point cloud data at multiple scale levels using the distribution coefficients and limiting scales. For example, specific values ​​at different scale levels can be determined based on the proportional relationship between scales, such as using a uniform or logarithmic scale. Then, corresponding filtering or sampling methods are applied to process the point cloud data at each scale level. The bottom layer of the pyramid corresponds to smaller scales, preserving more detail, while the top layer corresponds to larger scales, focusing on more macroscopic features. Each scale level uses a different resampling strategy to reduce the number of points while preserving key structural features, ensuring a suitable point cloud representation at each scale, ultimately constructing a complete scale-space pyramid.

[0068] By establishing a scale-space pyramid, tunnel point cloud data can be effectively organized at different scale levels, preserving fine local structural features while also capturing macroscopic, overall structural information. This multi-scale analysis method improves the flexibility and efficiency of point cloud data processing, balancing detail and integrity during point cloud data processing and providing more refined and comprehensive support for subsequent visualization.

[0069] Furthermore, in step S3 of the embodiment of the present application, the scale space pyramid is called and matched using the similar clustering result, and the position multi-scale model is reconstructed according to the call and matching result, including:

[0070] Step S34: obtaining the optimal matching scale between the scale space pyramid and the similar clustering result.

[0071] Step S35: taking the optimal matching scale as the central scale, and performing multi-scale adaptive reconstruction at the position of the scale space pyramid according to the similarity clustering result and the central scale.

[0072] Step S36: establishing a position multi-scale model according to the multi-scale adaptive reconstruction result and the central scale.

[0073] Specifically, the similarity between certain features at each scale in the scale-space pyramid and corresponding features in the similar clustering results is calculated, and the scale at each scale level of the scale-space pyramid that best matches the similar clustering results is determined, namely the optimal matching scale. For example, Euclidean distance, Mahalanobis distance, and other methods can be used to calculate the similarity between the density and shape features of the point cloud at different scales and the relevant features of the clusters in the similar clustering results. By comparing the similarity values ​​at different scales, the scale with the greatest similarity is found, which is the optimal matching scale.

[0074] The optimal matching scale is used as the central scale. This central scale is the core of subsequent multi-scale adaptive reconstruction. Using the optimal matching scale as the central scale can make the reconstruction result better reflect the characteristics of the point cloud data. Multi-scale adaptive reconstruction is performed based on the similar clustering results and the position of the central scale in the scale space pyramid, and the representation of point cloud data or clusters at different scales in the scale space pyramid is adjusted. For example, at scales around the central scale, the point cloud at the corresponding scale is adjusted based on the similar clustering results, and the cluster information is integrated into the point cloud representation. Alternatively, the point cloud at different scales is resampled or filtered based on the boundaries and structure of the clusters to achieve multi-scale adaptation.

[0075] A multiscale location model is established based on the multiscale adaptive reconstruction results and the central scale. This model integrates various information from the multiscale adaptive reconstruction process, including adjusted point cloud data at different scales, cluster information, and information related to the central scale. This ensures that the point cloud data at each scale accurately reflects its actual location and structural characteristics in the tunnel, thereby more comprehensively and accurately describing the location characteristics of the point cloud data at different scales.

[0076] By establishing a multi-scale location model, tunnel point cloud data can be fully displayed at multiple scale levels, ensuring the accuracy of fine structure and macroscopic morphology, and enabling a comprehensive and accurate representation of the complex structure of the tunnel.

[0077] Furthermore, the method described in the embodiment of the present application also includes:

[0078] Step S61: performing abnormal feature recognition on the tunnel structure feature set, and creating abnormal annotations based on the abnormal feature recognition results.

[0079] Step S62: Visually display the abnormal annotation through a Web visualization model.

[0080] Specifically, anomaly detection algorithms (such as those based on statistical analysis) are used to identify abnormal features within the tunnel structure feature set. For example, if the curvature of the tunnel wall suddenly changes significantly, exceeding the normal range, this is considered an abnormal feature. Based on the identified abnormal features, anomaly annotations are created, such as marking the abnormal points or areas with special symbols or colors. These anomaly annotations are then integrated into a web visualization model and displayed in a web browser, allowing users to intuitively identify the abnormal areas.

[0081] Through abnormal feature recognition and visual display, abnormal conditions in the tunnel structure can be discovered and displayed in a timely manner, making it easier for engineering personnel to quickly locate and analyze problems, thereby improving tunnel safety and maintenance efficiency.

[0082] In summary, the point cloud data management and Web rapid identification and visualization method provided by the embodiments of the present application has the following technical effects:

[0083] The embodiment of the present application first stores and manages the tunnel point cloud data through a point cloud database to provide data support for subsequent processing and analysis. Then, through similarity clustering and adaptive denoising processing, the noise is effectively removed, the quality of the point cloud data is improved, and the foundation is laid for accurate analysis. On this basis, a scale space pyramid is constructed, and scale matching and multi-scale adaptive reconstruction are performed in combination with the similarity clustering results, so that the tunnel structure at different scales can be fully displayed, ensuring the consistency and accuracy of the data at all levels. Finally, the position multi-scale model is used to extract the tunnel structure features, and a Web visualization model is created to achieve accurate fitting and efficient visualization of the tunnel point cloud data. Through this technical solution, not only the accuracy and efficiency of tunnel point cloud data processing are improved, but also the expressiveness and consistency of point cloud data at different scales are significantly enhanced, providing strong support for data analysis, monitoring and visualization in tunnel engineering, and solving the shortcomings of the existing technology in rapid Web recognition and large-scale data processing.

[0084] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A point cloud data management and Web rapid identification visualization method, characterized in that: The method comprises: Establishing a point cloud database, wherein the point cloud database is used to store tunnel point cloud data; After performing point cloud similarity clustering on the point cloud database, adaptive denoising is performed based on the similarity clustering results to establish a denoised point cloud set; Establishing a scale space pyramid, performing call matching on the scale space pyramid using the similarity clustering result, and reconstructing a position multi-scale model according to the call matching result; Perform feature extraction of denoised point clouds using the position multi-scale model to establish a tunnel structure feature set; Using the tunnel structure feature set to perform tunnel visualization fitting and create a Web visualization model; The performing point cloud similarity clustering on the point cloud database includes: Taking each point in the point cloud database as a node, calculating the similarity between nodes, and constructing a point cloud graph; Obtaining a calibration segmentation result of the tunnel, distributing initial clustering points in the point cloud using the calibration segmentation result, performing a cluster search of all nodes in the point cloud using the initial clustering points, and completing point cloud similarity clustering based on the clustering search results; The performing cluster search of all nodes in the point cloud graph using the initial cluster points includes: Establishing a clustering fitness function, wherein the clustering fitness function is a function for performing clustering fitness calculation; After taking the initial cluster point as the cluster center, the cluster fitness calculation within the preset search range of the cluster center is performed through the cluster fitness function, and the cluster fitness calculation includes the distance calculation between the node and the cluster center, the cluster size calculation corresponding to the cluster center, and the similarity calculation; Cluster expansion is performed based on the cluster fitness calculation results to reconstruct the cluster clusters; Perform cluster center search and update on the reconstructed cluster clusters to establish the reconstructed cluster center; Performing search iterations to complete the cluster search of all nodes in the point cloud graph; The execution of the search iteration comprises: Determine whether there is a cluster search stop event, and if there is a cluster search stop event, generate a new instruction; Perform distribution analysis on unclustered nodes according to the newly added instructions and establish new cluster centers; The search iteration is continued based on the newly added cluster centers.

2. A point cloud data management and Web rapid identification and visualization method according to claim 1, characterized in that: The obtaining of the calibration and segmentation results of the tunnel includes: Extract geometric features of the tunnel and establish the first segmentation constraint. The geometric features include the shape and size features of the tunnel. Extract geological features of the tunnel and establish the second segmentation constraint; Perform temporal extraction of tunnel construction features and establish the third segmentation constraint; The calibrated segmentation result is established using the first segmentation constraint, the second segmentation constraint, and the third segmentation constraint.

3. The point cloud data management and Web rapid identification and visualization method according to claim 1, characterized in that: The step of establishing a scale space pyramid includes: Performing node distribution analysis on the point cloud database and configuring limit scales, wherein the limit scales include a minimum scale and a maximum scale; Obtaining precision constraints for point cloud data management, and establishing distribution coefficients based on the precision constraints; A scale space pyramid is constructed using the distribution coefficient and the limit scale.

4. The point cloud data management and Web rapid identification and visualization method according to claim 3, characterized in that: The method of performing call matching on the scale space pyramid using the similar clustering result and reconstructing a position multi-scale model according to the call matching result includes: Obtaining an optimal matching scale between the scale space pyramid and the similarity clustering result; Taking the optimal matching scale as the central scale, and performing multi-scale adaptive reconstruction at the position of the scale space pyramid according to the similarity clustering result and the central scale; A position multi-scale model is established according to the multi-scale adaptive reconstruction result and the central scale.

5. The point cloud data management and Web rapid identification and visualization method according to claim 1, characterized in that: The method further comprises: performing abnormal feature recognition on the tunnel structure feature set, and creating an abnormal annotation based on the abnormal feature recognition result; The anomaly annotation is visualized through a Web visualization model.

Images