Regional deformation monitoring method and system based on total station

Abstract

The present invention relates to the field of construction engineering monitoring technology, and in particular to a regional deformation monitoring method and system based on a total station. The method comprises the following steps: arranging total station control points in a target area to obtain total station control network data; performing multi-source data acquisition based on the total station control network data to obtain an original observation data set; performing matching feature point analysis on the original observation data set to obtain matching feature point data; performing point cloud data set registration processing based on the matching feature point data and the original observation data set to obtain a registered point cloud data set; performing point cloud block processing on the registered point cloud data set to obtain block point cloud data; performing feature calculation on the block point cloud data to obtain point cloud feature data. The present invention improves the monitoring accuracy by realizing refined monitoring of deformation areas, overall morphological change analysis, and automatic identification and differentiation of deformation types, and overcomes the limitation that existing methods are difficult to identify deformation types.

Description

Technical Field

[0001] The present invention relates to the technical field of construction engineering monitoring, and in particular to a regional deformation monitoring method and system based on a total station. Background Art

[0002] As a high-precision measuring instrument, the total station has long been widely used in topographic mapping, engineering surveying and monitoring. With the advancement of technology, the functions of the total station have been continuously enhanced, and the point cloud data collection capabilities have been continuously improved, which has made it possible to apply it to regional deformation monitoring. Early deformation monitoring based on total stations mainly relied on manual observation and measurement. The staff measured the three-dimensional coordinates of the target point multiple times at different times, and then judged the deformation by comparing and analyzing the coordinate changes. This method has problems such as low efficiency, limited accuracy, and difficulty in fine-grained monitoring. In recent years, with the development of point cloud processing technology, some deformation monitoring methods based on total station point cloud data have emerged. These methods use total stations to collect point cloud data of the target area, and identify and quantify deformation through point cloud registration, comparative analysis and other means. However, these methods still have the following disadvantages:

[0003] 1. Limited accuracy, difficult to carry out refined monitoring: Existing methods usually only collect the three-dimensional coordinates of target points or feature points, and the point cloud density is low, which makes it difficult to reflect the detailed changes in the target area, especially when the deformation is small. It is difficult to accurately identify and quantify the deformation, lack of analysis of the overall morphological changes in the deformation area, and difficult to fully reflect the deformation situation.

[0004] 2. Difficulty in identifying deformation types: Existing methods usually only focus on the changes in the three-dimensional coordinates of points, lack the use of semantic information in point clouds, and can only identify whether there is deformation in the target area. It is difficult to distinguish different deformation types, such as settlement, horizontal displacement, tilt, etc. Summary of the invention

[0005] Based on this, it is necessary to provide a regional deformation monitoring method and system based on a total station to solve at least one of the above technical problems.

[0006] To achieve the above purpose, a regional deformation monitoring method based on a total station includes the following steps:

[0007] Step S1: Arrange total station control points in the target area to obtain total station control network data; perform multi-source data acquisition based on the total station control network data to obtain an original observation data set;

[0008] Step S2: performing matching feature point analysis on the original observation data set to obtain matching feature point data; performing point cloud data set registration processing according to the matching feature point data and the original observation data set to obtain a registered point cloud data set;

[0009] Step S3: performing point cloud block processing on the registered point cloud data set to obtain block point cloud data; performing feature calculation on the block point cloud data to obtain point cloud feature data; performing semantic segmentation model training based on the point cloud feature data to obtain a semantic segmentation model; performing point cloud semantic segmentation using the semantic segmentation model to obtain a semantic feature data set;

[0010] Step S4: performing deformation region identification according to the semantic feature data set to obtain deformation region refinement data; performing deformation region detection analysis on the deformation region refinement data to obtain deformation monitoring fusion result data;

[0011] Step S5: Generate target model parameters and initialize the model according to the registered point cloud data set to obtain target model parameters and an initial model; reconstruct the deformation area model of the deformation monitoring fusion result data using the target model parameters and the point cloud feature data to obtain reconstructed model fragment data; construct a three-dimensional model of the target area according to the reconstructed model fragment data and the initial model to obtain a three-dimensional model of the target area.

[0012] The present invention establishes a coordinate system of the target area by laying out control points, and uses the control point data to collect multi-source data, such as point cloud data, image data, reflection intensity data, etc. This lays the foundation for subsequent data processing and analysis and ensures the consistency and accuracy of the data. By matching feature points, the point cloud data from different data sources are aligned so that the data from different data sources can be uniformly processed in the same coordinate system. This provides the necessary conditions for subsequent point cloud analysis, semantic segmentation and model reconstruction. By processing the point cloud data in blocks, the complexity of data processing is reduced and the processing efficiency is improved. At the same time, through feature calculation and semantic segmentation, different objects and scenes in the point cloud data can be identified, and semantic information can be extracted, which provides important information for subsequent deformation monitoring and model reconstruction. By analyzing the semantic feature data, the deformed area of ​​the target area can be automatically identified, and the specific situation of the deformed area, such as deformation type, deformation degree, etc., can be further refined and analyzed, which provides important information for subsequent model reconstruction. By using the registration point cloud data, point cloud feature data and deformation monitoring result data to reconstruct the model of the target area, a refined three-dimensional model of the target area with rich deformation type information can be generated, providing important reference information for subsequent analysis, visualization and decision-making. Therefore, the present invention provides a regional deformation monitoring method based on a total station. High-density and high-precision three-dimensional information is acquired through multi-source data acquisition by the total station, and combined with semantic segmentation, deformation area extraction, and three-dimensional reconstruction technology, refined monitoring of the deformation area and overall morphological change analysis are achieved, which well solves the shortcomings of existing methods in terms of accuracy and refined monitoring capabilities. At the same time, through semantic feature analysis and time series data mining, automatic recognition and differentiation of deformation types are achieved, overcoming the limitation that existing methods are difficult to identify deformation types.

[0013] Preferably, step S1 comprises the following steps:

[0014] Step S11: Arrange total station control points in the target area to obtain total station control network data; set total station measurement parameters according to the total station control network data to obtain total station setting parameters, wherein the total station setting parameters include station position parameters, angle measurement mode parameters, scanning mode parameters and total station camera parameters;

[0015] Step S12: performing multi-angle observation and measurement data collection on the target area according to the measuring station position parameters and the angle measurement mode parameters to obtain multi-angle observation data;

[0016] Step S13: performing high-density point cloud scanning on the target area according to the scanning mode parameters to obtain high-density point cloud data;

[0017] Step S14: using the total station camera parameters to collect image data of the target area at different angles to obtain an image data set;

[0018] Step S15: recording reflection intensity information of the high-density point cloud data and the image data set to obtain reflection intensity data;

[0019] Step S16: synchronize and mark the multi-angle observation data, high-density point cloud data, image data set and reflection intensity data to obtain synchronized marked data; integrate the observation data on the synchronized marked data to obtain the original observation data set.

[0020] The present invention lays the foundation for subsequent multi-source data acquisition by arranging total station control points and setting measurement parameters for the target area. The total station control network data ensures the accuracy and reliability of the measurement, and the comprehensive measurement parameter setting provides the necessary technical support for different types of data acquisition, which helps to improve the efficiency and quality of the overall measurement. By using the station position parameters and angle measurement mode parameters for multi-angle observation, comprehensive geometric information of the target area can be obtained. This multi-angle observation method can effectively reduce the occlusion and shadow effects, improve the integrity and accuracy of the measurement, and provide high-quality basic data for subsequent point cloud processing and three-dimensional reconstruction. Detailed three-dimensional spatial information of the target area is obtained through high-density point cloud scanning. This high-density scanning method can capture the subtle features of the target surface, provide rich geometric information for subsequent feature extraction, semantic segmentation and deformation monitoring tasks, and is conducive to improving the accuracy and reliability of these analyses. Image data at different angles are collected using the total station camera to supplement the texture and color information of the target area. These image data can not only assist the semantic segmentation and feature extraction of the point cloud, but also provide realistic visual effects for the final three-dimensional model, greatly improving the quality and application value of the three-dimensional reconstruction results. Recording the reflection intensity information of high-density point clouds and image data increases the dimension and information volume of the data. Reflection intensity data can reflect the material characteristics of the target surface, which helps to improve the accuracy of point cloud classification and semantic segmentation, and also provides additional judgment basis for deformation monitoring. Through data synchronization, labeling and integration, multi-source heterogeneous data are unified into a coordinated and consistent data set. This data integration method not only ensures the spatiotemporal consistency between different types of data, but also facilitates subsequent data processing and analysis, effectively improving the data management and processing efficiency of the entire measurement system.

[0021] Preferably, step S2 comprises the following steps:

[0022] Step S21: performing original observation data denoising on the point cloud data in the original observation data set to obtain denoised point cloud data; performing data redundancy elimination on the denoised point cloud data to obtain simplified point cloud data;

[0023] Step S22: using the total station control network data to perform coordinate system conversion on the simplified point cloud data to obtain global coordinate point cloud data;

[0024] Step S23: extracting feature points from the global coordinate point cloud data and the original image data set in the original observation data set to obtain feature point data; performing feature point matching on the feature point data using a scale-invariant feature transformation algorithm to obtain matching feature point data;

[0025] Step S24: performing point cloud data set registration processing according to the matched feature point data and the original observation data set to obtain a registered point cloud data set.

[0026] The present invention significantly improves data quality and processing efficiency by denoising and eliminating redundancy of original point cloud data. Denoising operation can eliminate random errors and abnormal points generated during the measurement process, and improve the accuracy and reliability of point cloud data. Data redundancy elimination can significantly reduce the amount of data while retaining key geometric information, reduce the computational complexity of subsequent processing, and lay the foundation for efficient point cloud analysis and processing. The coordinate system conversion is performed using the total station control network data to achieve the unification of point cloud data from the local coordinate system to the global coordinate system. This step ensures the consistency of point cloud data collected by different measuring stations in the same coordinate system, provides accurate spatial reference for subsequent data fusion, deformation monitoring and three-dimensional reconstruction, and improves the accuracy and reliability of the entire measurement system. Feature point extraction and matching of point cloud data and image data lay the foundation for subsequent data registration. Feature point extraction can capture significant features in the data, and the application of the scale-invariant feature transform (SIFT) algorithm ensures the robustness of feature matching, and can effectively handle feature correspondences under different scales, rotations and viewing angles. This step greatly improves the accuracy and reliability of multi-source data fusion. By matching feature point data to perform point cloud data set registration processing, accurate alignment and fusion of multi-source data is achieved. This step not only eliminates the geometric deviation between different data sources, but also improves the consistency and integrity of the overall data. The registered point cloud data set provides high-quality input data for subsequent tasks such as semantic segmentation, deformation monitoring and 3D reconstruction, significantly improving the performance and accuracy of the entire measurement system.

[0027] Preferably, step S24 includes the following steps:

[0028] Step S241: performing initial change matrix calculation between different data according to the matching feature point data to obtain initial change matrix data; performing data coarse registration according to the initial change matrix data to obtain coarse registration point cloud data;

[0029] Step S242: performing spatial fine registration on the coarse registration point cloud data to obtain fine registration point cloud data;

[0030] Step S243: performing error analysis on the precise registration point cloud data to obtain error analysis data; performing error optimization on the precise registration point cloud data according to the error analysis data to obtain optimized registration point cloud data;

[0031] Step S244: generating a registered point cloud data set by performing registration on the optimized registered point cloud data and the original observation data set to obtain a registered point cloud data set.

[0032] The present invention lays the foundation for subsequent fine registration by calculating the initial change matrix and performing rough registration. This step can quickly estimate the approximate spatial relationship between different data sets, effectively reducing the search range and calculation amount of subsequent fine registration. The result of rough registration significantly improves the efficiency and stability of the entire registration process, which is particularly important when processing point cloud data of large-scale or complex scenes. The rough registration result is spatially fine-registered to further improve the alignment accuracy of the point cloud data. The fine registration process can carefully adjust the relative position between point clouds and eliminate slight deviations caused by factors such as measurement errors and equipment errors. This step significantly improves the accuracy of the final registration result and provides a high-quality data basis for subsequent deformation monitoring and three-dimensional reconstruction tasks. Through error analysis and optimization, the reliability and accuracy of the registration result are further improved. Error analysis can identify systematic errors and outliers in the registration process, and the optimization process based on the analysis results can adjust the registration parameters in a targeted manner to reduce cumulative errors. This step not only improves the overall quality of the registration, but also enhances the robustness of the results, so that the system can adapt to more complex measurement environments and data situations. The optimized registration point cloud data is integrated with the original observation data set to generate the final registration point cloud data set. This step realizes the effective fusion of multi-source data, retains the rich information of the original data, and ensures the consistency of spatial position. The generated registration point cloud data set provides a comprehensive, accurate and consistent data foundation for subsequent semantic segmentation, deformation monitoring and 3D model reconstruction tasks, significantly improving the performance and application value of the entire measurement system.

[0033] Preferably, step S3 comprises the following steps:

[0034] Step S31: performing point cloud data preprocessing on the registered point cloud data set to obtain preprocessed point cloud data; performing point cloud block processing on the preprocessed point cloud data according to the spatial position to obtain block point cloud data;

[0035] Step S32: performing feature calculation on the block point cloud data to obtain point cloud feature data;

[0036] Step S33: performing image texture information mapping on the point cloud feature data to obtain texture point cloud data; performing data enhancement processing on the texture point cloud data under different lighting conditions to obtain enhanced point cloud data;

[0037] Step S34: using the enhanced point cloud data to perform semantic segmentation model training on the pre-selected deep learning model to obtain a semantic segmentation model;

[0038] Step S35: using a semantic segmentation model to perform semantic segmentation on the texture point cloud data to obtain semantic segmentation point cloud data;

[0039] Step S36: performing feature fusion on the semantic segmentation point cloud data and the point cloud feature data to obtain semantic feature fusion data; performing data formatting on the semantic feature fusion data to obtain a semantic feature data set.

[0040] The present invention improves the efficiency and quality of subsequent processing through point cloud data preprocessing and spatial segmentation. Preprocessing can eliminate noise and abnormal points and improve data quality; while spatial segmentation divides large-scale point cloud data into manageable subsets, effectively reducing the computational complexity, enabling the system to efficiently process large-scale scene data, and laying the foundation for parallel processing. Feature calculation is performed on the segmented point cloud data to enrich the feature representation of the point cloud. These features include geometric features, statistical features, etc., which provide multi-dimensional information support for subsequent semantic segmentation and deformation monitoring, and significantly improve the accuracy and robustness of these tasks. The expressiveness of point cloud data is further enriched through image texture information mapping and data enhancement. The introduction of texture information adds rich visual features to the point cloud, while data enhancement improves the generalization ability of the model, enabling the system to better adapt to scenes under different lighting conditions, and enhancing the accuracy and stability of semantic segmentation. The deep learning model is trained using enhanced point cloud data to obtain a semantic segmentation model suitable for specific scenes. This training method based on actual data can enable the model to better adapt to the characteristics of the target area, improve the accuracy and efficiency of semantic segmentation, and provide high-quality semantic information for subsequent deformation monitoring and three-dimensional reconstruction. The texture point cloud data is processed using the trained semantic segmentation model to achieve automatic semantic annotation of the point cloud data. This step greatly improves the information content of the point cloud data, provides semantic support for subsequent deformation monitoring, 3D reconstruction and other tasks, and significantly improves the intelligence level of the entire system. Through feature fusion and data formatting, a comprehensive semantic feature dataset is generated. This dataset not only contains the geometric features of the point cloud, but also integrates semantic information, providing a rich and comprehensive data foundation for subsequent deformation monitoring and 3D model reconstruction. Data formatting ensures the consistency and availability of the data, facilitating subsequent processing and analysis.

[0041] Preferably, step S32 includes the following steps:

[0042] Step S321: determining neighboring points of the block point cloud data to obtain neighboring point set data; calculating normal vectors of the neighboring point set data to obtain point cloud data with normal vectors;

[0043] Step S322: performing curvature calculation based on the neighboring point set data and the point cloud data with normal vectors to obtain point cloud data with curvature;

[0044] Step S323: Calculate the roughness according to the point cloud data with curvature to obtain point cloud data with roughness;

[0045] Step S324: extracting radiation information from the block point cloud data to obtain point cloud data with radiation information;

[0046] Step S325: performing feature integration on the point cloud data with normal vectors, the point cloud data with curvature, the point cloud data with roughness, and the point cloud data with radiation information to obtain point cloud feature data.

[0047] The present invention can effectively describe the local geometric shape of point cloud data by calculating the normal vectors of neighboring points around each point. Normal vector information can be used in subsequent curvature calculation, roughness analysis, semantic segmentation and other links to provide key information for more accurate understanding of point cloud data. Curvature reflects the degree of curvature of point cloud data and can be used to identify features such as sharp edges, smooth surfaces and curved surfaces in point clouds. Curvature information can better describe the geometric shape of point cloud data and provide accurate geometric information for subsequent semantic segmentation and model reconstruction. Roughness can reflect the smoothness of the surface of point cloud data, which is very useful for identifying objects of different materials, such as smooth walls and rough bricks. Roughness information can provide important texture feature information for subsequent semantic segmentation and improve the accuracy of semantic segmentation. Radiation information can reflect the intensity of light reflected by each point in point cloud data, which can be used to distinguish objects of different materials, such as dark asphalt pavement and light cement pavement. Radiation information can provide important color feature information for subsequent semantic segmentation and improve the accuracy of semantic segmentation. Integrating the geometric feature information (normal vector, curvature, roughness) and radiation information calculated above can form richer and more complete point cloud feature data, provide more comprehensive information for subsequent semantic segmentation and model reconstruction, and improve the accuracy and reliability of the final result.

[0048] Preferably, step S4 comprises the following steps:

[0049] Step S41: constructing a temporal semantic feature dataset according to the semantic feature dataset to obtain a temporal semantic feature dataset;

[0050] Step S42: performing preliminary deformation region recognition on the temporal semantic feature data set to obtain preliminary deformation region recognition data; performing deformation region refinement on the preliminary deformation region recognition data to obtain refined deformation region data;

[0051] Step S43: using the semantic segmentation model to perform semantic analysis on the deformation region refinement data to obtain semantic segmentation data of the deformation region;

[0052] Step S44: performing deformation type identification based on the deformation region semantic segmentation data to obtain deformation type identification data; performing deformation quantization index calculation on the deformation region refinement data to obtain deformation quantization index;

[0053] Step S45: fusing the deformation type identification data and the deformation quantification index into deformation monitoring results to obtain deformation monitoring fusion result data.

[0054] The present invention integrates the semantic feature data of different time periods to establish a time series data set containing time information and semantic information. This provides the necessary data basis for subsequent deformation monitoring and can effectively analyze the deformation conditions of different time periods. By comparing the semantic feature data of different time periods, the deformation area is preliminarily identified. Subsequently, the preliminarily identified area is refined to more accurately locate the deformation area, providing more accurate information for subsequent deformation analysis. By using the trained semantic segmentation model, the deformation area is semantically analyzed to identify the specific categories in the deformation area, such as piers, bridge decks, bridges, etc. This provides important information for a deeper understanding of the deformation situation. By further analyzing the specific types of the deformation area, such as the pier settlement area, the bridge deck tilt area, the bridge expansion joint change area, etc. At the same time, the quantitative indicators such as the area and volume of the deformation area are calculated to more intuitively reflect the degree of deformation. The type information and quantitative indicators of the deformation area are integrated to form a more comprehensive and accurate deformation monitoring result. This provides more reliable information for subsequent model reconstruction and decision analysis.

[0055] Preferably, step S5 comprises the following steps:

[0056] Step S51: extracting real-time point cloud data and real-time image data sets from the registered point cloud data sets to obtain real-time point cloud data and real-time image data sets;

[0057] Step S52: generating target model parameters according to the real-time point cloud data to obtain target model parameters; initializing the model according to the target model parameters to obtain an initial model;

[0058] Step S53: reconstructing a deformation region model of the deformation region in the deformation monitoring fusion result data using the target model parameters, the point cloud feature data and the real-time point cloud data to obtain reconstructed model fragment data;

[0059] Step S54: merging the models of the undeformed regions of the reconstructed model fragment data and the initial model to obtain a fused model; adding model semantic details to the fused model according to the target model parameters and the semantic feature data set to obtain a detail enhanced model;

[0060] Step S55: using the real-time image data set to perform model texture mapping on the detail enhancement model to obtain a texture mapping model; performing model quality inspection on the texture mapping model and performing model repair to obtain a repaired three-dimensional model;

[0061] Step S56: updating the model attributes of the repaired three-dimensional model according to the semantic feature data set and the deformation monitoring fusion result data to obtain a three-dimensional model of the target area.

[0062] The present invention obtains the latest point cloud data and image data to provide real-time information for subsequent model reconstruction, which can ensure the timeliness and accuracy of the reconstructed model. The parameters of the target model are generated according to the real-time point cloud data, and the model is initialized using these parameters, which lays the foundation for the subsequent reconstruction of the deformation area model. The model parameters, point cloud feature data and real-time point cloud data are used to reconstruct the model of the deformation area to generate new model fragments. This enables the model to reflect the latest deformation and maintain the integrity and consistency of the model. The reconstructed model fragments of the deformation area are merged with the initial model, and the model details are added using semantic feature data, making the model more complete, rich and closer to the real scene. The model is texture mapped using real-time image data to make the model more realistic, and the model is quality checked and repaired to ensure the integrity and quality of the model. The model attributes are updated using semantic feature data and deformation monitoring result data, so that the model contains richer information, such as material, category, deformation, etc., and finally a refined three-dimensional model of the target area with rich deformation type information is obtained.

[0063] Preferably, step S53 includes the following steps:

[0064] Step S531: performing plane segmentation on the deformation area in the deformation monitoring fusion result data using the point cloud data with normal vectors in the point cloud feature data to obtain plane segmentation data;

[0065] Step S532: performing plane fitting on the plane segmentation data to obtain plane equation data;

[0066] Step S533: using the plane equation data to perform gridding processing on the deformation monitoring fusion result data of the deformation area point cloud data to obtain initial grid model data;

[0067] Step S534: performing grid semantic segmentation on the initial grid model data using the target model parameters and the semantic feature data set to obtain semantic segmentation grid data;

[0068] Step S535: reconstructing different types of deformed region models according to the semantic segmentation grid data to obtain a deformed region reconstruction model data set; and performing reconstruction model fusion on the deformed region reconstruction model data set to obtain fused reconstruction model data;

[0069] Step S536: Use the real-time point cloud data to perform geometric detail enhancement on the fused reconstructed model data to obtain reconstructed model fragment data.

[0070] The present invention segments the plane part in the deformation area by using normal vector information, providing a more accurate geometric basis for subsequent model reconstruction. Fitting the segmented plane part to obtain the plane equation can more accurately describe the plane shape and provide necessary data for subsequent meshing processing. Using the plane equation data, the point cloud data of the deformation area is converted into a mesh model, providing a more convenient and efficient geometric structure for subsequent model reconstruction. Using semantic feature data, the mesh model is semantically segmented to identify different categories of objects in the mesh model, such as piers, bridge decks, bridges, etc., providing more accurate semantic information for subsequent model reconstruction. According to the semantic segmentation results, different types of deformation areas are model reconstructed, and the reconstructed model fragments are fused to obtain a complete deformation area model, which more accurately reflects the geometric shape and semantic information of the deformation area. Using the latest real-time point cloud data, the reconstructed model is enhanced in detail, which can more accurately reflect the geometric details of the deformation area and improve the realism and accuracy of the model.

[0071] Preferably, the present invention further provides a regional deformation monitoring system based on a total station, which is used to execute the regional deformation monitoring method based on a total station as described above. The regional deformation monitoring system based on a total station comprises:

[0072] The multi-source observation data acquisition module is used to arrange the total station control points in the target area and obtain the total station control network data; multi-source data acquisition is performed based on the total station control network data to obtain the original observation data set;

[0073] The data preprocessing and registration module is used to perform matching feature point analysis on the original observation data set to obtain matching feature point data; perform point cloud data set registration processing based on the matching feature point data and the original observation data set to obtain a registered point cloud data set;

[0074] The semantic segmentation and feature extraction module is used to perform point cloud segmentation processing on the registered point cloud data set to obtain segmented point cloud data; perform feature calculation on the segmented point cloud data to obtain point cloud feature data; perform semantic segmentation model training based on the point cloud feature data to obtain a semantic segmentation model; perform point cloud semantic segmentation using the semantic segmentation model to obtain a semantic feature data set;

[0075] The multi-temporal deformation monitoring module is used to identify the deformation area according to the semantic feature data set to obtain the deformation area refinement data; perform deformation area detection analysis on the deformation area refinement data to obtain the deformation monitoring fusion result data;

[0076] The three-dimensional model reconstruction module is used to generate target model parameters and initialize the model according to the registered point cloud data set to obtain the target model parameters and the initial model; use the target model parameters and the point cloud feature data to reconstruct the deformation area model of the deformation monitoring fusion result data to obtain the reconstructed model fragment data; construct the three-dimensional model of the target area according to the reconstructed model fragment data and the initial model to obtain the three-dimensional model of the target area.

[0077] The present invention establishes a coordinate system of the target area by laying out control points, and uses the control point data to collect multi-source data, such as point cloud data, image data, reflection intensity data, etc. This lays the foundation for subsequent data processing and analysis and ensures the consistency and accuracy of the data. By matching feature points, the point cloud data from different data sources are aligned so that the data from different data sources can be uniformly processed in the same coordinate system. This provides the necessary conditions for subsequent point cloud analysis, semantic segmentation and model reconstruction. By processing the point cloud data in blocks, the complexity of data processing is reduced and the processing efficiency is improved. At the same time, through feature calculation and semantic segmentation, different objects and scenes in the point cloud data can be identified, and semantic information can be extracted, which provides important information for subsequent deformation monitoring and model reconstruction. By analyzing the semantic feature data, the deformed area of ​​the target area can be automatically identified, and the specific situation of the deformed area, such as deformation type, deformation degree, etc., can be further refined and analyzed, which provides important information for subsequent model reconstruction. By using the registration point cloud data, point cloud feature data and deformation monitoring result data to reconstruct the model of the target area, a complete three-dimensional model containing rich details and attribute information can be generated, providing important reference information for subsequent analysis, visualization and decision-making. Therefore, the present invention provides a regional deformation monitoring method based on a total station. High-density and high-precision three-dimensional information is acquired through multi-source data acquisition by the total station. Combined with semantic segmentation, deformation area extraction, and three-dimensional reconstruction technology, refined monitoring of the deformation area and overall morphological change analysis are achieved, which well solves the shortcomings of existing methods in terms of accuracy and refined monitoring capabilities. At the same time, through semantic feature analysis and time series data mining, automatic recognition and differentiation of deformation types are achieved, overcoming the limitation that existing methods are difficult to identify deformation types. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 A schematic diagram of the steps of the regional deformation monitoring method based on the total station;

[0079] Figure 2 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0080] Figure 3 for Figure 1 Detailed implementation steps of step S4 in FIG.

[0081] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0082] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0083] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0084] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0085] To achieve this, please refer to Figures 1 to 3 , a regional deformation monitoring method based on a total station, comprising the following steps:

[0086] Step S1: Arrange total station control points in the target area to obtain total station control network data; perform multi-source data acquisition based on the total station control network data to obtain an original observation data set;

[0087] Step S2: performing matching feature point analysis on the original observation data set to obtain matching feature point data; performing point cloud data set registration processing according to the matching feature point data and the original observation data set to obtain a registered point cloud data set;

[0088] Step S3: performing point cloud block processing on the registered point cloud data set to obtain block point cloud data; performing feature calculation on the block point cloud data to obtain point cloud feature data; performing semantic segmentation model training based on the point cloud feature data to obtain a semantic segmentation model; performing point cloud semantic segmentation using the semantic segmentation model to obtain a semantic feature data set;

[0089] Step S4: performing deformation region identification according to the semantic feature data set to obtain deformation region refinement data; performing deformation region detection analysis on the deformation region refinement data to obtain deformation monitoring fusion result data;

[0090] Step S5: Generate target model parameters and initialize the model according to the registered point cloud data set to obtain target model parameters and an initial model; reconstruct the deformation area model of the deformation monitoring fusion result data using the target model parameters and the point cloud feature data to obtain reconstructed model fragment data; construct a three-dimensional model of the target area according to the reconstructed model fragment data and the initial model to obtain a three-dimensional model of the target area.

[0091] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of the regional deformation monitoring method based on the total station of the present invention. In this example, the regional deformation monitoring method based on the total station includes the following steps:

[0092] Step S1: Arrange total station control points in the target area to obtain total station control network data; perform multi-source data acquisition based on the total station control network data to obtain an original observation data set;

[0093] In the embodiment of the present invention, firstly, the total station control points are arranged in the target area to establish a control network. Then, the total station parameters are set according to the control network data, including the station position, angle measurement mode, scanning mode and camera parameters. Next, the total station is used to perform multi-angle observation, high-density point cloud scanning, image acquisition and reflection intensity recording. Finally, these data are synchronously marked and integrated to form the original observation data set.

[0094] Step S2: performing matching feature point analysis on the original observation data set to obtain matching feature point data; performing point cloud data set registration processing according to the matching feature point data and the original observation data set to obtain a registered point cloud data set;

[0095] In the embodiment of the present invention, the system extracts and matches feature points of the point cloud data and image data in the original observation data set to obtain matching feature point data. Then, the matching feature point data is used in combination with the original observation data set to perform point cloud data registration processing. This process includes steps such as initial transformation matrix calculation, rough registration, fine registration, and error optimization, and finally obtains a precisely registered point cloud data set.

[0096] Step S3: performing point cloud block processing on the registered point cloud data set to obtain block point cloud data; performing feature calculation on the block point cloud data to obtain point cloud feature data; performing semantic segmentation model training based on the point cloud feature data to obtain a semantic segmentation model; performing point cloud semantic segmentation using the semantic segmentation model to obtain a semantic feature data set;

[0097] In the embodiment of the present invention, the system first performs spatial block processing on the registered point cloud dataset, dividing the large-scale point cloud data into manageable small blocks. Then, feature calculations are performed on each block, including normal vectors, curvature, roughness, and radiation information. Next, these feature data are used to train a deep learning model to obtain a semantic segmentation model. Finally, the trained model is used to perform semantic segmentation on the point cloud, assigning semantic labels to each point to form a semantic feature dataset.

[0098] Step S4: performing deformation region identification according to the semantic feature data set to obtain deformation region refinement data; performing deformation region detection analysis on the deformation region refinement data to obtain deformation monitoring fusion result data;

[0099] In the embodiment of the present invention, the system uses the semantic feature data set to construct a temporal semantic feature data set to perform preliminary identification and refinement of the deformation area. Then, the refined deformation area is semantically analyzed to identify the deformation type. At the same time, the quantitative index of the deformation is calculated. Finally, the deformation type and the quantitative index are fused to obtain comprehensive deformation monitoring fusion result data.

[0100] Step S5: Generate target model parameters and initialize the model according to the registered point cloud data set to obtain target model parameters and an initial model; reconstruct the deformation area model of the deformation monitoring fusion result data using the target model parameters and the point cloud feature data to obtain reconstructed model fragment data; construct a three-dimensional model of the target area according to the reconstructed model fragment data and the initial model to obtain a three-dimensional model of the target area;

[0101] In the embodiment of the present invention, the system first generates the target model parameters according to the registered point cloud data set and constructs the initial model. Then, the deformed area is reconstructed using the target model parameters, point cloud feature data and deformation monitoring fusion result data to obtain a reconstructed model fragment. Next, the reconstructed model fragment is merged with the initial model and semantic details are added. Finally, model texture mapping, quality inspection and attribute update are performed to finally obtain a complete three-dimensional model of the target area.

[0102] Preferably, step S1 comprises the following steps:

[0103] Step S11: Arrange total station control points in the target area to obtain total station control network data; set total station measurement parameters according to the total station control network data to obtain total station setting parameters, wherein the total station setting parameters include station position parameters, angle measurement mode parameters, scanning mode parameters and total station camera parameters;

[0104] Step S12: performing multi-angle observation and measurement data collection on the target area according to the measuring station position parameters and the angle measurement mode parameters to obtain multi-angle observation data;

[0105] Step S13: performing high-density point cloud scanning on the target area according to the scanning mode parameters to obtain high-density point cloud data;

[0106] Step S14: using the total station camera parameters to collect image data of the target area at different angles to obtain an image data set;

[0107] Step S15: recording reflection intensity information of the high-density point cloud data and the image data set to obtain reflection intensity data;

[0108] Step S16: synchronize and mark the multi-angle observation data, high-density point cloud data, image data set and reflection intensity data to obtain synchronized marked data; integrate the observation data on the synchronized marked data to obtain the original observation data set.

[0109] In the embodiment of the present invention, a number of control points are arranged in the target area, and the control points are observed by a total station to obtain the three-dimensional coordinate data of the control points, thereby forming the total station control network data. Then, according to the control network data, the station position parameters of the total station are set to the coordinates of the control point A, the angle measurement mode parameters are set to the angle measurement of two surfaces, the scanning mode parameters are set to the high-density point cloud scanning, and the total station camera parameters are set to the wide-angle lens. According to the station position parameters, the total station is set at the control point A, and then the target area is observed at multiple angles according to the angle measurement mode parameters to obtain observation data at different angles, thereby forming multi-angle observation data. According to the scanning mode parameters, the target area is scanned with a high-density point cloud by the total station, and the scanning density is 100 points per square meter to obtain high-density point cloud data of the target area. According to the total station camera parameters, the digital camera on the total station is used to collect images of the target area at different angles to obtain a series of image data at different viewing angles, thereby forming an image data set. Analyze the high-density point cloud data and image data set, extract and record the reflection intensity information of each point and pixel, and obtain the reflection intensity data of the point cloud and image. Synchronize the multi-angle observation data, high-density point cloud data, image data set, and reflection intensity data according to the acquisition time, and add data tags to obtain synchronized marked data. Then integrate multiple data in the synchronized marked data, remove redundant and erroneous data, and obtain the original observation data set.

[0110] Preferably, step S2 comprises the following steps:

[0111] Step S21: performing original observation data denoising on the point cloud data in the original observation data set to obtain denoised point cloud data; performing data redundancy elimination on the denoised point cloud data to obtain simplified point cloud data;

[0112] Step S22: using the total station control network data to perform coordinate system conversion on the simplified point cloud data to obtain global coordinate point cloud data;

[0113] Step S23: extracting feature points from the global coordinate point cloud data and the original image data set in the original observation data set to obtain feature point data; performing feature point matching on the feature point data using a scale-invariant feature transformation algorithm to obtain matching feature point data;

[0114] Step S24: performing point cloud data set registration processing according to the matched feature point data and the original observation data set to obtain a registered point cloud data set.

[0115] The present invention performs statistical filtering on the point cloud data in the original observation data set, removes noise points, and obtains denoised point cloud data. Then, the denoised point cloud data is voxelized, the space is divided into a number of cubic grids, each grid retains one point, and redundant points are removed to obtain simplified point cloud data. The control point coordinates provided by the total station control network data are used to establish a coordinate conversion equation, and the points in the simplified point cloud data are converted from the local coordinate system to the global coordinate system to obtain global coordinate point cloud data. SIFT feature point extraction is performed on the global coordinate point cloud data and the original image data set respectively to obtain feature point data of the point cloud and the image. Then, the SIFT feature matching algorithm is used to match the two feature point data to obtain matching feature point pairs and form matching feature point data. According to the matching feature point data, the ICP algorithm is used to align the point cloud data set, and the point cloud data obtained from different measuring stations are aligned to the same coordinate system. At the same time, the point cloud and image data in the original observation data set are used for joint alignment to further improve the alignment accuracy, and finally the aligned point cloud data set is obtained.

[0116] Preferably, step S24 comprises the following steps:

[0117] Step S241: performing initial change matrix calculation between different data according to the matching feature point data to obtain initial change matrix data; performing data coarse registration according to the initial change matrix data to obtain coarse registration point cloud data;

[0118] Step S242: performing spatial fine registration on the coarse registration point cloud data to obtain fine registration point cloud data;

[0119] Step S243: performing error analysis on the precise registration point cloud data to obtain error analysis data; performing error optimization on the precise registration point cloud data according to the error analysis data to obtain optimized registration point cloud data;

[0120] Step S244: generating a registered point cloud data set by performing registration on the optimized registered point cloud data and the original observation data set to obtain a registered point cloud data set.

[0121] The present invention calculates the initial rotation matrix R and translation vector t between different data based on the matching feature point data to form the initial change matrix data. Then, the point cloud data is roughly registered using the initial change matrix so that the different data are roughly aligned to obtain the roughly registered point cloud data. Based on the roughly registered point cloud data, the ICP algorithm is used for fine registration. The ICP algorithm iteratively optimizes the distance between the nearest points and continuously updates the transformation matrix until the error threshold is met or the maximum number of iterations is reached to obtain the precisely registered point cloud data. The precisely registered point cloud data is error analyzed to calculate the distance error and angle error between the corresponding points before and after the registration to form the error analysis data. According to the error analysis data, the precisely registered point cloud data is error optimized, and the transformation matrix is ​​optimized by the weighted least squares method to minimize the distance error and angle error to obtain the optimized registered point cloud data. The optimized registered point cloud data is fused with the original observation data set to generate a complete registered point cloud data set. During the fusion process, the image data in the original observation dataset is used to perform texture mapping on the point cloud data, and the control point data in the original observation dataset is used to perform geometric constraints on the registration results to ensure the accuracy and integrity of the registered point cloud dataset.

[0122] Preferably, step S3 comprises the following steps:

[0123] Step S31: performing point cloud data preprocessing on the registered point cloud data set to obtain preprocessed point cloud data; performing point cloud block processing on the preprocessed point cloud data according to the spatial position to obtain block point cloud data;

[0124] Step S32: performing feature calculation on the block point cloud data to obtain point cloud feature data;

[0125] Step S33: performing image texture information mapping on the point cloud feature data to obtain texture point cloud data; performing data enhancement processing on the texture point cloud data under different lighting conditions to obtain enhanced point cloud data;

[0126] Step S34: using the enhanced point cloud data to perform semantic segmentation model training on the pre-selected deep learning model to obtain a semantic segmentation model;

[0127] Step S35: using a semantic segmentation model to perform semantic segmentation on the texture point cloud data to obtain semantic segmentation point cloud data;

[0128] Step S36: performing feature fusion on the semantic segmentation point cloud data and the point cloud feature data to obtain semantic feature fusion data; performing data formatting on the semantic feature fusion data to obtain a semantic feature data set.

[0129] As an example of the present invention, refer to Figure 2 As shown, in this example, step S3 includes:

[0130] Step S31: performing point cloud data preprocessing on the registered point cloud data set to obtain preprocessed point cloud data; performing point cloud block processing on the preprocessed point cloud data according to the spatial position to obtain block point cloud data;

[0131] In the embodiment of the present invention, the registration point cloud data set is downsampled to reduce the point cloud density to obtain pre-processed point cloud data. Then, the pre-processed point cloud data is spatially divided using an octree algorithm to divide the point cloud into a number of spatial blocks, each block containing a certain number of points, to obtain block point cloud data.

[0132] Step S32: performing feature calculation on the block point cloud data to obtain point cloud feature data;

[0133] In the embodiment of the present invention, feature calculation is performed on the block point cloud data, including normal vector feature, curvature feature, roughness feature, etc. The normal vector feature calculates the normal vector of each point by PCA algorithm; the curvature feature calculates the curvature value of each point and its neighboring points; the roughness feature calculates the distance variance of each point in the neighborhood to the fitting plane. Finally, point cloud feature data containing multiple features is obtained.

[0134] Step S33: performing image texture information mapping on the point cloud feature data to obtain texture point cloud data; performing data enhancement processing on the texture point cloud data under different lighting conditions to obtain enhanced point cloud data;

[0135] In the embodiment of the present invention, the point cloud feature data is mapped to the corresponding image texture information, and RGB color information is assigned to each point cloud point to obtain texture point cloud data. Then, by changing the illumination direction and intensity, the texture point cloud data is enhanced to simulate the point cloud data under different illumination conditions to obtain enhanced point cloud data.

[0136] Step S34: using the enhanced point cloud data to perform semantic segmentation model training on the pre-selected deep learning model to obtain a semantic segmentation model;

[0137] In the embodiment of the present invention, PointNet++ is selected as the deep learning model, and the model is trained using enhanced point cloud data. During the training process, the point cloud data and the corresponding semantic labels are used as input, and the model parameters are optimized through the back propagation algorithm so that the model can accurately perform semantic segmentation on the point cloud. After multiple rounds of iterative training, a semantic segmentation model is obtained.

[0138] Step S35: using a semantic segmentation model to perform semantic segmentation on the texture point cloud data to obtain semantic segmentation point cloud data;

[0139] In an embodiment of the present invention, a trained semantic segmentation model is used to perform semantic segmentation on texture point cloud data, and each point cloud point is divided into different semantic categories, such as bridges, lamp posts, bridge deck pavement, etc., to obtain semantic segmentation point cloud data.

[0140] Step S36: performing feature fusion on the semantic segmentation point cloud data and the point cloud feature data to obtain semantic feature fusion data; performing data formatting on the semantic feature fusion data to obtain a semantic feature data set;

[0141] In the embodiment of the present invention, the semantic segmentation point cloud data is fused with the point cloud feature data, and semantic attribute information is added to each point cloud point to obtain semantic feature fusion data. Finally, the semantic feature fusion data is formatted and converted into a data format that is easy to store and apply, such as LAS format, to obtain a semantic feature data set.

[0142] Preferably, step S32 includes the following steps:

[0143] Step S321: determining neighboring points of the block point cloud data to obtain neighboring point set data; calculating normal vectors of the neighboring point set data to obtain point cloud data with normal vectors;

[0144] Step S322: performing curvature calculation based on the neighboring point set data and the point cloud data with normal vectors to obtain point cloud data with curvature;

[0145] Step S323: Calculate the roughness according to the point cloud data with curvature to obtain point cloud data with roughness;

[0146] Step S324: extracting radiation information from the block point cloud data to obtain point cloud data with radiation information;

[0147] Step S325: performing feature integration on the point cloud data with normal vectors, the point cloud data with curvature, the point cloud data with roughness, and the point cloud data with radiation information to obtain point cloud feature data.

[0148] In an embodiment of the present invention, for each point in the block point cloud data, the nearest neighbor point is searched by the KD tree algorithm to construct the neighbor point set data. Then, the PCA algorithm is used to perform principal component analysis on the neighbor point set, and the normal vector feature is extracted. The normal vector information is attached to the original point cloud to obtain the point cloud data with normal vector. According to the neighbor point set data, the distance and angle between each point and its neighbor point are calculated, the local surface is fitted by the least squares method, the principal curvature and Gaussian curvature are calculated, and the curvature information is attached to the point cloud data with normal vector to obtain the point cloud data with curvature. According to the point cloud data with curvature, the distance variance from the point in the neighborhood of each point to the fitting plane is calculated to characterize the roughness of the local point cloud, and the roughness information is attached to the point cloud data with curvature to obtain the point cloud data with roughness. The block point cloud data is subjected to radiation correction to eliminate the influence of atmospheric and geometric factors on the radiation intensity, and the radiation intensity information of each point is extracted. The radiation information is attached to the original point cloud to obtain the point cloud data with radiation information. Feature integration is performed on point cloud data with normal vectors, point cloud data with curvature, point cloud data with roughness, and point cloud data with radiation information. Different feature data are merged into the same point cloud. Each point contains multiple attribute information such as XYZ coordinates, normal vector, curvature, roughness, radiation intensity, etc., to obtain point cloud feature data.

[0149] Preferably, step S4 comprises the following steps:

[0150] Step S41: constructing a temporal semantic feature dataset according to the semantic feature dataset to obtain a temporal semantic feature dataset;

[0151] Step S42: performing preliminary deformation region recognition on the temporal semantic feature data set to obtain preliminary deformation region recognition data; performing deformation region refinement on the preliminary deformation region recognition data to obtain refined deformation region data;

[0152] Step S43: using the semantic segmentation model to perform semantic analysis on the deformation region refinement data to obtain semantic segmentation data of the deformation region;

[0153] Step S44: performing deformation type identification based on the deformation region semantic segmentation data to obtain deformation type identification data; performing deformation quantization index calculation on the deformation region refinement data to obtain deformation quantization index;

[0154] Step S45: fusing the deformation type recognition data and the deformation quantification index into deformation monitoring results to obtain deformation monitoring fusion result data.

[0155] As an example of the present invention, refer to Figure 3 As shown, in this example, step S4 includes:

[0156] Step S41: constructing a temporal semantic feature dataset according to the semantic feature dataset to obtain a temporal semantic feature dataset;

[0157] In the embodiment of the present invention, the semantic feature data acquired at different times are organized into a time series according to the timestamp information of the semantic feature data set to construct a temporal semantic feature data set. For example, the semantic feature data at three moments T1, T2, and T3 are arranged in chronological order to form a temporal semantic feature data set.

[0158] Step S42: performing preliminary deformation region recognition on the temporal semantic feature data set to obtain preliminary deformation region recognition data; performing deformation region refinement on the preliminary deformation region recognition data to obtain refined deformation region data;

[0159] In the embodiment of the present invention, the temporal semantic feature data set is compared time by time, and the deformed area is preliminarily identified through the deformation of semantic information and geometric information to obtain preliminary identification data of the deformed area. Then, the preliminary identification result is refined by using morphological filtering and region growing algorithm, small deformed areas are removed, and adjacent deformed areas are merged to obtain refined deformed area data.

[0160] Step S43: using the semantic segmentation model to perform semantic analysis on the deformation region refinement data to obtain semantic segmentation data of the deformation region;

[0161] In an embodiment of the present invention, a semantic segmentation model is used to perform semantic analysis on the refined data of the deformation area, the point cloud in the deformation area is reclassified, and the semantic information of the deformation area is identified, such as the displacement changes of the bridge piers, the inclination changes of the bridge deck, the changes of the bridge expansion joints, etc., to obtain the semantic segmentation data of the deformation area.

[0162] Step S44: performing deformation type identification based on the deformation region semantic segmentation data to obtain deformation type identification data; performing deformation quantization index calculation on the deformation region refinement data to obtain deformation quantization index;

[0163] In the embodiment of the present invention, based on the semantic segmentation data of the deformation area, the deformation type is identified by the deformation type of semantic information (such as pier settlement, bridge deck inclination, bridge expansion joint change), and deformation type identification data is obtained. At the same time, geometric quantitative analysis is performed on the deformation area refinement data, and indicators such as the area, volume, height difference, etc. of the deformation area are calculated to obtain deformation quantitative indicators.

[0164] Step S45: fusing the deformation type recognition data and the deformation quantification index into deformation monitoring results to obtain deformation monitoring fusion result data;

[0165] In the embodiment of the present invention, the deformation type identification data and the deformation quantification index are fused to generate a complete deformation monitoring result. During the fusion process, a decision tree or random forest algorithm is used to comprehensively judge the deformation attribute and deformation degree of the deformation area according to the deformation type and deformation quantification index to obtain the deformation monitoring fusion result data.

[0166] Preferably, step S5 comprises the following steps:

[0167] Step S51: extracting real-time point cloud data and real-time image data sets from the registered point cloud data sets to obtain real-time point cloud data and real-time image data sets;

[0168] Step S52: generating target model parameters according to the real-time point cloud data to obtain target model parameters; initializing the model according to the target model parameters to obtain an initial model;

[0169] Step S53: reconstructing a deformation region model of the deformation region in the deformation monitoring fusion result data using the target model parameters, the point cloud feature data and the real-time point cloud data to obtain reconstructed model fragment data;

[0170] Step S54: merging the models of the undeformed regions of the reconstructed model fragment data and the initial model to obtain a fused model; adding model semantic details to the fused model according to the target model parameters and the semantic feature data set to obtain a detail enhanced model;

[0171] Step S55: using the real-time image data set to perform model texture mapping on the detail enhancement model to obtain a texture mapping model; performing model quality inspection on the texture mapping model and performing model repair to obtain a repaired three-dimensional model;

[0172] Step S56: updating the model attributes of the repaired three-dimensional model according to the semantic feature data set and the deformation monitoring fusion result data to obtain a three-dimensional model of the target area.

[0173] In an embodiment of the present invention, the point cloud data and image data at the latest time are extracted from the registered point cloud data set as real-time point cloud data and real-time image data set. For example, the point cloud data and image data at time T3 are extracted for subsequent three-dimensional model reconstruction and updating. According to the real-time point cloud data, the geometric parameters of the target model, such as the length, width, height, wall equation, etc. of the building, are generated by plane fitting, edge extraction and other methods to obtain the target model parameters. Then, according to the target model parameters, the initial geometric framework of the target model is generated in three-dimensional space to obtain the initial model. The target model parameters and point cloud feature data are used to perform three-dimensional reconstruction of the deformed area in the deformation monitoring fusion result data. By extracting the point cloud of the deformed area, the real-time point cloud data is used to complete the geometric details, and the model is optimized in combination with semantic information, a three-dimensional model of the deformed area is reconstructed to obtain the reconstructed model fragment data. The reconstructed model fragment data is geometrically spliced ​​with the initial model, and the non-deformed area is merged to obtain a complete fusion model. Then, according to the target model parameters and semantic feature data sets, the fusion model is semantically processed, for example, detail elements such as pier settlement area, bridge deck tilt area, and bridge expansion joint change area are added to the model to obtain a detail enhanced model. Using the real-time image data set, the image information is mapped to the detail enhanced model through the texture mapping algorithm, and a real texture map is added to the model to obtain a texture mapping model. Then, the texture mapping model is quality checked to identify and repair geometric errors and texture errors to obtain a repaired 3D model. According to the semantic feature data set and the deformation monitoring fusion result data, the attributes of the repaired 3D model are updated. The semantic information and deformation quantification indicators of the deformation area are associated with the 3D model, and attribute labels such as material, category, and size are added to the model elements to obtain a 3D model of the target area with complete geometry, texture, and attribute information.

[0174] Preferably, step S53 includes the following steps:

[0175] Step S531: performing plane segmentation on the deformation area in the deformation monitoring fusion result data using the point cloud data with normal vectors in the point cloud feature data to obtain plane segmentation data;

[0176] Step S532: performing plane fitting on the plane segmentation data to obtain plane equation data;

[0177] Step S533: using the plane equation data to perform gridding processing on the deformation monitoring fusion result data of the deformation area point cloud data to obtain initial grid model data;

[0178] Step S534: performing grid semantic segmentation on the initial grid model data using the target model parameters and the semantic feature data set to obtain semantic segmentation grid data;

[0179] Step S535: reconstructing different types of deformed region models according to the semantic segmentation grid data to obtain a deformed region reconstruction model data set; and performing reconstruction model fusion on the deformed region reconstruction model data set to obtain fused reconstruction model data;

[0180] Step S536: Use the real-time point cloud data to perform geometric detail enhancement on the fused reconstructed model data to obtain reconstructed model fragment data.

[0181] In an embodiment of the present invention, the system first uses the point cloud data with normal vectors in the point cloud feature data to perform plane segmentation on the deformed area identified in the deformation monitoring fusion result data. This process divides the deformed area into multiple plane areas by analyzing the spatial relationship between the normal vector of each point and the surrounding points to form plane segmentation data. The plane segmentation data obtained in the previous step is subjected to plane fitting processing. The system uses the least squares method or RANSAC and other algorithms to fit the best plane equation for each segmented plane area, and obtains a set of plane equation data describing the geometric characteristics of the deformed area. Using the obtained plane equation data, the system performs grid processing on the deformed area point cloud in the deformation monitoring fusion result data. This process converts continuous point cloud data into a discrete grid structure to form initial grid model data, preparing for subsequent semantic segmentation and model reconstruction. The system combines the target model parameters and the semantic feature data set to perform grid semantic segmentation on the initial grid model data. This process assigns specific semantic labels to each unit in the grid model, such as bridge deck, abutment, support, etc., to form semantic segmentation grid data. According to the semantic segmentation grid data, the system performs model reconstruction on different types of deformation areas. For different semantic categories, corresponding reconstruction algorithms, such as plane reconstruction and surface fitting, are used to obtain the deformed area reconstruction model data set. Then, these independently reconstructed models are fused to form a complete fused reconstruction model data. Finally, the system uses the point cloud data collected in real time to enhance the geometric details of the fused reconstruction model data. This process compares the reconstructed model with the real-time point cloud, supplements the missing detail information in the model, improves the accuracy and realism of the model, and finally obtains high-quality reconstructed model fragment data.

[0182] Preferably, the present invention further provides a regional deformation monitoring system based on a total station, which is used to execute the regional deformation monitoring method based on a total station as described above. The regional deformation monitoring system based on a total station comprises:

[0183] The multi-source observation data acquisition module is used to arrange the total station control points in the target area and obtain the total station control network data; multi-source data acquisition is performed based on the total station control network data to obtain the original observation data set;

[0184] The data preprocessing and registration module is used to perform matching feature point analysis on the original observation data set to obtain matching feature point data; perform point cloud data set registration processing based on the matching feature point data and the original observation data set to obtain a registered point cloud data set;

[0185] The semantic segmentation and feature extraction module is used to perform point cloud segmentation processing on the registered point cloud data set to obtain segmented point cloud data; perform feature calculation on the segmented point cloud data to obtain point cloud feature data; perform semantic segmentation model training based on the point cloud feature data to obtain a semantic segmentation model; perform point cloud semantic segmentation using the semantic segmentation model to obtain a semantic feature data set;

[0186] The multi-temporal deformation monitoring module is used to identify the deformation area according to the semantic feature data set to obtain the deformation area refinement data; perform deformation area detection analysis on the deformation area refinement data to obtain the deformation monitoring fusion result data;

[0187] The three-dimensional model reconstruction module is used to generate target model parameters and initialize the model according to the registered point cloud data set to obtain the target model parameters and the initial model; use the target model parameters and the point cloud feature data to reconstruct the deformation area model of the deformation monitoring fusion result data to obtain the reconstructed model fragment data; construct the three-dimensional model of the target area according to the reconstructed model fragment data and the initial model to obtain the three-dimensional model of the target area.

[0188] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is intended that all modifications falling within the meaning and scope of equivalent elements of the application documents are included in the present invention.

[0189] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those 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 invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A regional deformation monitoring method based on a total station, characterized in that: The following steps are involved: Step S1: Arrange total station control points in the target area to obtain total station control network data; Multi-source data collection is performed based on the total station control network data to obtain the original observation data set; Step S2: performing matching feature point analysis on the original observation data set to obtain matching feature point data; Perform point cloud data set registration processing according to the matched feature point data and the original observation data set to obtain a registered point cloud data set; Step S3: performing point cloud block processing on the registered point cloud data set to obtain block point cloud data; Perform feature calculation on the block point cloud data to obtain point cloud feature data; A semantic segmentation model is trained based on the point cloud feature data to obtain a semantic segmentation model; the point cloud semantic segmentation is performed using the semantic segmentation model to obtain a semantic feature data set; Step S4: identifying the deformation region according to the semantic feature data set to obtain the deformation region refinement data; Perform deformation area detection and analysis on the deformation area refinement data to obtain deformation monitoring fusion result data; Step S5: Generate target model parameters and initialize the model according to the registered point cloud data set to obtain target model parameters and an initial model; reconstruct the deformation area model of the deformation monitoring fusion result data using the target model parameters and the point cloud feature data to obtain reconstructed model fragment data; construct a three-dimensional model of the target area according to the reconstructed model fragment data and the initial model to obtain a three-dimensional model of the target area.

2. The regional deformation monitoring method based on total station according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: Arrange total station control points in the target area to obtain total station control network data; set total station measurement parameters according to the total station control network data to obtain total station setting parameters, wherein the total station setting parameters include station position parameters, angle measurement mode parameters, scanning mode parameters and total station camera parameters; Step S12: collecting multi-angle observation data of the target area according to the measuring station position parameters and the angle measurement mode parameters to obtain multi-angle observation data; Step S13: performing high-density point cloud scanning on the target area according to the scanning mode parameters to obtain high-density point cloud data; Step S14: using the total station camera parameters to collect image data of the target area at different angles to obtain an image data set; Step S15: recording reflection intensity information of the high-density point cloud data and the image data set to obtain reflection intensity data; Step S16: synchronize and mark the multi-angle observation data, high-density point cloud data, image data set and reflection intensity data to obtain synchronized marked data; integrate the observation data on the synchronized marked data to obtain the original observation data set.

3. The regional deformation monitoring method based on total station according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing original observation data denoising on the point cloud data in the original observation data set to obtain denoised point cloud data; performing data redundancy elimination on the denoised point cloud data to obtain simplified point cloud data; Step S22: using the total station control network data to perform coordinate system conversion on the simplified point cloud data to obtain global coordinate point cloud data; Step S23: extracting feature points from the global coordinate point cloud data and the original image data set in the original observation data set to obtain feature point data; performing feature point matching on the feature point data using a scale-invariant feature transformation algorithm to obtain matching feature point data; Step S24: performing point cloud data set registration processing according to the matched feature point data and the original observation data set to obtain a registered point cloud data set.

4. The regional deformation monitoring method based on total station according to claim 3 is characterized in that: Step S24 includes the following steps: Step S241: performing initial change matrix calculation between different data according to the matching feature point data to obtain initial change matrix data; performing data coarse registration according to the initial change matrix data to obtain coarse registration point cloud data; Step S242: performing spatial fine registration on the coarse registration point cloud data to obtain fine registration point cloud data; Step S243: performing error analysis on the precise registration point cloud data to obtain error analysis data; performing error optimization on the precise registration point cloud data according to the error analysis data to obtain optimized registration point cloud data; Step S244: generating a registered point cloud data set by performing registration on the optimized registered point cloud data and the original observation data set to obtain a registered point cloud data set.

5. The regional deformation monitoring method based on total station according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing point cloud data preprocessing on the registered point cloud data set to obtain preprocessed point cloud data; performing point cloud block processing on the preprocessed point cloud data according to the spatial position to obtain block point cloud data; Step S32: performing feature calculation on the block point cloud data to obtain point cloud feature data; Step S33: performing image texture information mapping on the point cloud feature data to obtain texture point cloud data; performing data enhancement processing on the texture point cloud data under different lighting conditions to obtain enhanced point cloud data; Step S34: using the enhanced point cloud data to perform semantic segmentation model training on the pre-selected deep learning model to obtain a semantic segmentation model; Step S35: using a semantic segmentation model to perform semantic segmentation on the texture point cloud data to obtain semantic segmentation point cloud data; Step S36: performing feature fusion on the semantic segmentation point cloud data and the point cloud feature data to obtain semantic feature fusion data; performing data formatting on the semantic feature fusion data to obtain a semantic feature data set.

6. The regional deformation monitoring method based on total station according to claim 5 is characterized in that: Step S32 includes the following steps: Step S321: determining neighboring points of the block point cloud data to obtain neighboring point set data; calculating normal vectors of the neighboring point set data to obtain point cloud data with normal vectors; Step S322: performing curvature calculation based on the neighboring point set data and the point cloud data with normal vectors to obtain point cloud data with curvature; Step S323: Calculate the roughness according to the point cloud data with curvature to obtain point cloud data with roughness; Step S324: extracting radiation information from the block point cloud data to obtain point cloud data with radiation information; Step S325: performing feature integration on the point cloud data with normal vectors, the point cloud data with curvature, the point cloud data with roughness, and the point cloud data with radiation information to obtain point cloud feature data.

7. The regional deformation monitoring method based on total station according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: constructing a temporal semantic feature dataset according to the semantic feature dataset to obtain a temporal semantic feature dataset; Step S42: performing preliminary deformation region recognition on the temporal semantic feature data set to obtain preliminary deformation region recognition data; performing deformation region refinement on the preliminary deformation region recognition data to obtain refined deformation region data; Step S43: using the semantic segmentation model to perform semantic analysis on the deformation region refinement data to obtain semantic segmentation data of the deformation region; Step S44: performing deformation type identification based on the deformation region semantic segmentation data to obtain deformation type identification data; performing deformation quantization index calculation on the deformation region refinement data to obtain deformation quantization index; Step S45: fusing the deformation type identification data and the deformation quantification index into deformation monitoring results to obtain deformation monitoring fusion result data.

8. The regional deformation monitoring method based on total station according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: extracting real-time point cloud data and real-time image data sets from the registered point cloud data sets to obtain real-time point cloud data and real-time image data sets; Step S52: generating target model parameters according to the real-time point cloud data to obtain target model parameters; initializing the model according to the target model parameters to obtain an initial model; Step S53: reconstructing a deformation region model of the deformation region in the deformation monitoring fusion result data using the target model parameters, the point cloud feature data and the real-time point cloud data to obtain reconstructed model fragment data; Step S54: merging the models of the undeformed regions of the reconstructed model fragment data and the initial model to obtain a fused model; adding model semantic details to the fused model according to the target model parameters and the semantic feature data set to obtain a detail enhanced model; Step S55: using the real-time image data set to perform model texture mapping on the detail enhancement model to obtain a texture mapping model; performing model quality inspection on the texture mapping model and performing model repair to obtain a repaired three-dimensional model; Step S56: updating the model attributes of the repaired three-dimensional model according to the semantic feature data set and the deformation monitoring fusion result data to obtain a three-dimensional model of the target area.

9. The regional deformation monitoring method based on total station according to claim 8, characterized in that: Step S53 includes the following steps: Step S531: performing plane segmentation on the deformation area in the deformation monitoring fusion result data using the point cloud data with normal vectors in the point cloud feature data to obtain plane segmentation data; Step S532: performing plane fitting on the plane segmentation data to obtain plane equation data; Step S533: using the plane equation data to perform gridding processing on the deformation monitoring fusion result data of the deformation area point cloud data to obtain initial grid model data; Step S534: performing grid semantic segmentation on the initial grid model data using the target model parameters and the semantic feature data set to obtain semantic segmentation grid data; Step S535: reconstructing different types of deformed region models according to the semantic segmentation grid data to obtain a deformed region reconstruction model data set; and performing reconstruction model fusion on the deformed region reconstruction model data set to obtain fused reconstruction model data; Step S536: Use the real-time point cloud data to perform geometric detail enhancement on the fused reconstructed model data to obtain reconstructed model fragment data.

10. A regional deformation monitoring system based on a total station, characterized in that: Used to execute the regional deformation monitoring method based on total station as claimed in claim 1, the regional deformation monitoring system based on total station comprises: The multi-source observation data acquisition module is used to arrange the total station control points in the target area and obtain the total station control network data; multi-source data acquisition is performed based on the total station control network data to obtain the original observation data set; The data preprocessing and registration module is used to perform matching feature point analysis on the original observation data set to obtain matching feature point data; perform point cloud data set registration processing based on the matching feature point data and the original observation data set to obtain a registered point cloud data set; The semantic segmentation and feature extraction module is used to perform point cloud segmentation processing on the registered point cloud data set to obtain segmented point cloud data; perform feature calculation on the segmented point cloud data to obtain point cloud feature data; perform semantic segmentation model training based on the point cloud feature data to obtain a semantic segmentation model; perform point cloud semantic segmentation using the semantic segmentation model to obtain a semantic feature data set; The multi-temporal deformation monitoring module is used to identify the deformation area according to the semantic feature data set to obtain the deformation area refinement data; perform deformation area detection analysis on the deformation area refinement data to obtain the deformation monitoring fusion result data; The three-dimensional model reconstruction module is used to generate target model parameters and initialize the model according to the registered point cloud data set to obtain the target model parameters and the initial model; use the target model parameters and the point cloud feature data to reconstruct the deformation area model of the deformation monitoring fusion result data to obtain the reconstructed model fragment data; construct the three-dimensional model of the target area according to the reconstructed model fragment data and the initial model to obtain the three-dimensional model of the target area.

Images