A total station survey data acquisition method and system

Abstract

The present application relates to the field of data acquisition, and more particularly to a total station measurement data acquisition method and system, comprising: obtaining an original measurement data set with a time sequence label by performing three-dimensional coordinate calculation and arrangement on original total station measurement data; obtaining an adaptive neighborhood radius of each point by comprehensively analyzing the local neighborhood distance of the original measurement data points; obtaining a shared local structure tensor between point pairs by fusing and constructing the local structure tensor of the measurement points; obtaining a shape-adaptive symmetric anisotropic distance measurement by performing principal axis direction deviation analysis on the shared local structure tensor; and obtaining a gross error point identification result and eliminating abnormal measurement data by jointly judging the shape-adaptive distance and the neighborhood radius, thereby solving the problem of inaccurate gross error identification caused by uneven distribution density of measurement points and data anisotropy.

Description

Technical Field

[0001] This invention relates to the field of data acquisition, and in particular to a method and system for acquiring total station measurement data. Background Technology

[0002] Total stations, as high-precision measuring devices, are widely used in the long-term automated health monitoring of large infrastructure projects such as bridges, dams, and high-rise buildings. In these applications, the total station performs continuous, high-frequency coordinate measurements on multiple monitoring points deployed on the structure according to a preset cycle, generating massive amounts of temporal spatial data. To ensure the accuracy of subsequent deformation analysis, it is essential to effectively identify and remove the unavoidable gross errors (i.e., abnormal data points caused by factors such as atmospheric turbulence and instantaneous instrument reading errors) inevitably mixed in the raw data stream. In existing technologies, one method for achieving automated gross error identification is the DBSCAN algorithm. The basic principle of this algorithm is that in coordinate space, normal measurement data points naturally cluster due to their inherent repeatability, forming one or more high-density regions; while gross errors typically manifest as isolated noise points far from these dense regions. The DBSCAN algorithm automatically divides data points into core points, boundary points, and noise points using two core parameters set by the user: a globally fixed neighborhood radius and a minimum number of points. In this method, identified noise points are considered gross errors and are discarded, while core points and boundary points together constitute normal measurement data clusters and are retained for subsequent analysis. However, the gross error identification method using the standard DBSCAN algorithm has an unresolved issue regarding its algorithm mechanism and data characteristics when processing actual total station monitoring data, limiting the accuracy of its identification results. In a typical monitoring network, the measurement data of each monitoring point exhibits different distribution densities according to the error propagation law due to varying physical distances from the total station. Specifically, data clusters of nearby monitoring points are compact and dense, while data clusters of distant monitoring points are loose and low-density. The standard DBSCAN algorithm requires a globally uniform neighborhood radius to measure the neighborhood range of all data points, a setting that cannot simultaneously accommodate the diverse data densities present in the network. If EPS is set too small to accurately identify high-density clusters, a large number of normal low-density data points will be misclassified as gross errors due to overly strict scaling. Conversely, if EPS is set too large to accommodate low-density clusters, some genuine gross errors will be incorrectly classified into high-density data clusters due to overly lenient scaling, resulting in missed detections. This conflict between the global fixedness of the EPS parameter and the local non-uniformity of data density makes it difficult for the algorithm to achieve consistent and optimal gross error identification across the entire monitoring network when selecting parameters, which is a problem that urgently needs to be solved in current technology. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a total station measurement data acquisition method to solve the problem of inaccurate identification of gross errors caused by uneven distribution density of measuring points and data anisotropy.

[0004] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0005] Firstly, this application provides a method for acquiring total station measurement data, which includes the following steps:

[0006] Step S1: Obtain the original measurement dataset with time-series labels by performing three-dimensional coordinate calculation and processing on the original total station measurement data;

[0007] Step S2: Obtain the adaptive neighborhood radius of each point by comprehensively analyzing the local neighborhood distances of the original measurement data points;

[0008] Step S3: By fusing and constructing the local structure tensors of the measurement points, the shared local structure tensors between point pairs are obtained;

[0009] Step S4: Obtain a morphologically adaptive symmetric anisotropic distance metric by performing principal axis deviation analysis on the shared local structure tensor;

[0010] Step S5: By jointly judging the morphological adaptive distance and the neighborhood radius, obtain the gross error identification results and remove abnormal measurement data.

[0011] Preferably, the step of obtaining a time-labeled original measurement dataset by performing three-dimensional coordinate calculation and processing on the original total station measurement data includes:

[0012] By setting multiple monitoring points in the structure or area to be measured, and using a total station automated measurement system, periodic three-dimensional coordinate measurements are performed on each monitoring point within a preset measurement cycle. The three-dimensional spatial coordinate values ​​of the monitoring point and its corresponding time label are obtained at each measurement time. The measurement results of all the monitoring points are organized in chronological order to form an original measurement dataset containing the three-dimensional coordinates and time labels of each monitoring point at multiple measurement times. The original measurement dataset consists of multiple measurement data points, and each measurement data point contains its three-dimensional coordinate values ​​and acquisition time information.

[0013] Preferably, the step of obtaining the adaptive neighborhood radius of each point by comprehensively analyzing the local neighborhood distances of the original measurement data points includes:

[0014] Set the number of nearest neighbor data points and obtain the set of nearest neighbor data points for any target measurement data point from all measurement data points. For any target measurement data point, use the Euclidean distance between the target measurement data point and the farthest data point in its nearest neighbor data point set as the local maximum neighborhood arm length of the target measurement data point. Use the average Euclidean distance between the target measurement data point and all data points in its nearest neighbor data point set as the local average neighborhood arm length of the target measurement data point. Use the square root of the sum of the squares of the local maximum neighborhood arm length and the local average neighborhood arm length of the target measurement data point as the adaptive neighborhood radius of the target measurement data point.

[0015] Preferably, the step of fusing and constructing the local structure tensor of the measurement points to obtain the shared local structure tensor between point pairs includes:

[0016] For any set of nearest neighbor data for a target measurement data point, the vector pointing from the geometric center of the set of nearest neighbor data to any target neighbor measurement data point in the set is taken as the position vector of the target measurement data point relative to the target neighbor measurement data point; the cross product of the position vector of the target measurement data point relative to the target neighbor measurement data point and itself is taken as the first vector cross product of the target measurement data point relative to the target neighbor measurement data point; for any set of nearest neighbor data for a target measurement data point, the sum of all first vector cross products in the set of nearest neighbor data is taken as the local structure tensor of the target measurement data point; for any two local structure tensors of the target measurement data points, the mean of the local structure tensors of the two target measurement data points is taken as the shared local structure tensor of the two target measurement data points.

[0017] Preferably, the step of obtaining a morphology-adaptive symmetric anisotropic distance metric by performing principal axis deviation analysis on the shared local structure tensor includes:

[0018] By performing eigenvector decomposition on the shared local structure tensor, a set of shared principal axis directions is obtained; by performing projection analysis on the connection direction of the measurement point pairs and the set of principal axis directions, a shared local principal axis deviation factor is obtained; by fusing and weighting the original Euclidean distance and the principal axis deviation factor, a morphology-adaptive symmetric anisotropic distance is obtained.

[0019] Preferably, the step of obtaining the set of shared principal axis directions by performing eigenvector decomposition on the shared local structure tensor includes:

[0020] Eigenvector decomposition is performed on the shared local structure tensor to obtain eigenvalue-eigenvector pairs corresponding to the number of dimensions of the shared local structure tensor. All eigenvalue-eigenvector pairs are sorted in descending order according to the size of the eigenvalues, and the eigenvector corresponding to the first ranked eigenvalue is taken as the principal axis direction of the shared local structure tensor. The eigenvectors corresponding to the remaining eigenvalues ​​are taken as the secondary axis directions. The set formed by the principal axis direction and the secondary axis direction is taken as the set of shared principal axis directions.

[0021] Preferably, the step of obtaining the shared local principal axis deviation factor by projecting the set of measurement point pairs connecting the line direction and the principal axis direction includes:

[0022] For any two target measurement data points, the direction vector between the two target measurement data points is obtained through their coordinates; the dot product of the direction vector between the two target measurement data points and any feature vector in the set of shared principal axis directions of the two target measurement data points is taken as the first vector dot product between the two target measurement data points.

[0023] The result of summing all the first vector dot products between the two target measurement data points is used as the denominator, and the result of summing all the first vector dot products between the two target measurement data points except for the principal axis direction is used as the numerator. The result of the corresponding fraction is used as the shared local principal axis deviation factor of the two target measurement data points.

[0024] Preferably, the step of obtaining the morphology-adaptive symmetric anisotropic distance by fusing and weighting the original Euclidean distance and the principal axis deviation factor includes:

[0025] The result of adding the constant 1 to the shared local principal axis deviation factor is used as the first distance optimization weight. The result of multiplying the first distance optimization weight of any two target measurement data points with the Euclidean distance between the two target measurement data points is used as the morphologically adaptive symmetric anisotropic distance between the two target measurement data points.

[0026] Preferably, the step of jointly judging the morphological adaptive distance and the neighborhood radius to obtain the gross error identification result and remove abnormal measurement data includes:

[0027] Set a minimum data point threshold, and perform DBSCAN clustering on the measurement data points in the original measurement dataset based on the adaptive neighborhood radius of the measurement data points and the morphologically adaptive symmetric anisotropic distance between the measurement data points. Complete the clustering process and obtain the core points, boundary points and noise points marked during the clustering process. The measurement data points marked as noise points are used as gross data points and deleted from the original dataset to obtain the measurement dataset with outlier measurement data removed.

[0028] Secondly, this application provides a total station measurement data acquisition system, including: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a total station measurement data acquisition method is implemented.

[0029] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:

[0030] This invention effectively solves the problem of uneven point cloud density caused by ranging differences by constructing an adaptive neighborhood radius that adapts to local density variations at measurement points. This technique combines the local average neighborhood arm length and the maximum arm length to form a composite diffusion radius. It automatically tightens the neighborhood judgment scale in high-density areas and dynamically widens the discrimination boundary in low-density areas, thereby ensuring that normal measurement points within different density clusters are accurately identified and retained, effectively suppressing erroneous deletions caused by fixed neighborhood parameters.

[0031] Meanwhile, this invention introduces a shared local structure tensor and constructs an anisotropic distance metric mechanism based on principal axis deviation, significantly improving the adaptability of data morphology to environmental disturbances. By analyzing the deviation between the direction of the connection between point pairs and the shared principal extension direction, the clustering process is guided to extend along the natural stretching direction of the data clusters, avoiding misidentification of trajectory changes caused by physical disturbances as gross errors. This technique demonstrates excellent direction discrimination and structure recognition capabilities in actual monitoring scenarios, enabling high-precision outlier removal even under complex environmental conditions, thus improving the overall quality of total station monitoring data and the reliability of subsequent analysis. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a flowchart of a total station measurement data acquisition method provided in an embodiment of the present invention. Detailed Implementation

[0034] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.

[0035] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0036] See Figure 1 This is a flowchart of a total station measurement data acquisition method provided in Embodiment 1 of the present invention, as shown below. Figure 1 As shown, the method may include:

[0037] Step S1: Obtain the original measurement dataset with time-series labels by performing three-dimensional coordinate calculation and processing on the original total station measurement data.

[0038] In a total station measurement data acquisition method of the present invention, step S1 is used to acquire a raw measurement dataset with time-series labels, which serves as the basic data source for subsequent gross error identification and processing. Specifically, multiple total station measurement monitoring points are first set up in the structure to be measured or the monitoring area. The monitoring points can be spatially deployed in a targeted manner according to the key parts of the structure, deformation-sensitive areas, or differences in distance measurement conditions.

[0039] By setting up multiple monitoring points in the structure or area to be measured, and using a total station automated measurement system, periodic three-dimensional coordinate measurements are performed on each monitoring point within a preset measurement cycle. The three-dimensional spatial coordinate values ​​of the monitoring point and its corresponding time label are obtained at each measurement time. The measurement results of all the monitoring points are organized in chronological order to form an original measurement dataset containing the three-dimensional coordinates and time labels of each monitoring point at multiple measurement times. The original measurement dataset consists of multiple measurement data points, and each measurement data point contains its three-dimensional coordinate values ​​and acquisition time information.

[0040] During the measurement process, a total station automated monitoring system was used to periodically measure the three-dimensional spatial coordinates of each monitoring point. Each measurement operation recorded the measurement time information and measurement result data, including the three-dimensional coordinate values ​​of each monitoring point at that moment, and the corresponding time label. As time progressed, the system organized the multiple measurement data points obtained from continuous measurements in chronological order to form an original measurement dataset, where each data point contained spatial three-dimensional coordinate values ​​and time label information.

[0041] This raw measurement dataset comprehensively reflects the spatial distribution and temporal evolution of each measuring point during the monitoring period, providing complete input data for subsequent implementation of gross error identification algorithms based on density and structural features. In actual deployment, the sampling period of the raw measurement data can be set according to the engineering monitoring requirements. In this invention, the sampling period is set to once every 10 minutes to ensure the timeliness and continuity of the data.

[0042] This completes the process of calculating and organizing the three-dimensional coordinates of the original total station measurement data to obtain the original measurement dataset with time-series labels.

[0043] Step S2: By comprehensively analyzing the local neighborhood distances of the original measurement data points, the adaptive neighborhood radius of each point is obtained.

[0044] The limitation of the existing DBSCAN algorithm lies in the global fixedness of its neighborhood radius. This setting conflicts with the uneven density characteristics of total station measurement data, which varies in density due to different distances, preventing the algorithm from finding an optimal neighborhood radius value that is applicable to all monitoring points simultaneously. To solve this problem, this invention should not use a uniform radius, but rather dynamically evaluate the most suitable local neighborhood radius for each data point in the dataset, based on the data distribution pattern of its region. To achieve this goal, it is necessary to assess whether the region where the data point is located is dense or sparse. The local neighborhood radius of a data point should be uniquely determined by the actual distribution of the surrounding data points. Specifically, it should simultaneously reflect the average clustering scale and the maximum dispersion range of points within its neighborhood. For a high-density data cluster, the local average neighborhood arm length and the local maximum neighborhood arm length of any point within it will be very small. Conversely, for a low-density data cluster, the values ​​of these two indicators will be significantly larger. Based on this principle, this step fuses these two indicators that can jointly characterize the local scale to calculate an adaptive neighborhood radius that can accurately match the local data density. Specifically, the number of nearest neighbor data points is set. In this embodiment of the invention, the number of nearest neighbor data points is set. The number of nearest neighbor data points can be adjusted according to the specific scenario. The set of nearest neighbor data points for any target measurement data point among all measurement data points is obtained. For any target measurement data point, the Euclidean distance between the target measurement data point and the farthest data point in its set of nearest neighbor data points is taken as the local maximum neighborhood arm length of the target measurement data point. The average Euclidean distance between the target measurement data point and all data points in its set of nearest neighbor data points is taken as the local average neighborhood arm length of the target measurement data point. The square root of the sum of the squares of the local maximum neighborhood arm length and the local average neighborhood arm length of the target measurement data point is taken as the adaptive neighborhood radius of the target measurement data point.

[0045] In one implementation, assume the first The measurement data point and the first The Euclidean distance between the measurement data points is The number of nearest neighbor data points is , No. The set of nearest neighbor data points for each data point is: Then the first The formula for calculating the adaptive neighborhood radius of a data point is:

[0046]

[0047] in, Indicates the first Adaptive neighborhood radius for each data point; Indicates the first The measurement data point and the first Euclidean distance between measurement data points; Represents the maximum value function; Indicates the first The set of nearest neighbor data points for each data point; This indicates the number of nearest neighbor data points.

[0048] It should be noted that the adaptive neighborhood radius designed in this invention aims to solve the problem caused by the globally fixed neighborhood radius parameter in existing technologies. By constructing an adaptive neighborhood radius, it achieves adaptation to the local data scale. In actual total station surveying scenarios, for a nearby monitoring point, the measurement data is highly concentrated due to small errors, resulting in a local maximum neighborhood arm length. With local average neighborhood arm length The value of the value is relatively small, and the adaptive neighborhood radius calculated by the formula is also a relatively small value. This small neighborhood radius can form a fine judgment scale, thereby effectively identifying the small gross errors existing in the compact data cluster. In contrast, for a distant monitoring point, its measurement data is loosely distributed due to error accumulation, and its local maximum neighborhood arm length is small. With local average neighborhood arm length The value of the coefficient increases significantly, and the calculated adaptive neighborhood radius is also magnified accordingly. This magnified coefficient forms a judgment scale that matches the actual dispersion range of the loose data cluster, thus ensuring that all normal far-distance measurement points are fully included and avoiding misjudgments caused by overly strict scales. Furthermore, the fusion evaluation using the local maximum neighborhood arm length and the local average neighborhood arm length is superior to using either one alone. If only the local maximum neighborhood arm length is used for the adaptive neighborhood radius evaluation, the neighborhood radius will be entirely determined by the farthest neighbor, easily influenced by the accidental location of that point. By introducing the local average neighborhood arm length, the calculation result can comprehensively reflect the average distribution of all measurement data points within the neighborhood, reducing the sensitivity to the location of a single farthest point. Simultaneously, if only the local average neighborhood arm length is used, the defined radius may be smaller than the actual distance to the farthest neighbor, resulting in the inability to include all neighborhood data points. By introducing the local maximum neighborhood arm length, it is ensured that the final generated adaptive neighborhood radius is necessarily greater than or equal to the distance to the farthest neighbor, guaranteeing the integrity of the neighborhood definition.

[0049] Thus, the adaptive neighborhood radius of each point is obtained by comprehensively analyzing the local neighborhood distances of the original measurement data points.

[0050] Step S3: By fusing and constructing the local structure tensors of the measurement points, the shared local structure tensors between point pairs are obtained.

[0051] In practical total station surveying scenarios, especially in monitoring areas traversing complex terrain or experiencing atmospheric disturbances, such as river crossings, long-distance measurements in valleys, or the deployment of measuring points in narrow urban streets, measurement data are often significantly affected by spatial directionality. Due to environmental factors such as water evaporation, surface heat radiation, or building obstruction, the temperature, humidity, and density distribution of the atmospheric medium becomes uneven in the horizontal or vertical directions, leading to varying degrees of lateral or vertical refraction along the laser ranging path. This atmospheric refraction effect has a clear directional characteristic, typically causing a systematic shift of the measuring point along a specific axis, resulting in the originally isotropic measurement data cluster exhibiting an ellipsoidal shape with significant stretching along the principal axis.

[0052] In such data structures, if standard Euclidean distance or single-point neighborhood structures are still used for data analysis, the overall structural trend reflected by directional deformation between measurement data is easily overlooked, thus affecting the accuracy of gross error identification. To address this, this invention proposes a method of constructing a shared local structure tensor to accurately describe the common neighborhood structural characteristics of any two measurement data points. This structure tensor not only reflects the geometric shape and directional consistency of the measurement point's neighborhood but also provides a unified morphological benchmark for subsequent construction of anisotropic distance metrics, ensuring good symmetry and structural adaptability in distance calculations across different point pairs.

[0053] To achieve this structural representation, it is first necessary to establish a mathematical representation of the local structure tensor based on the spatial relative position between each measurement data point and its nearest neighbor data points. Furthermore, the tensor information of each point pair is fused to generate a unified representation of the shared local structure of that point pair. Specifically, for any target measurement data point's nearest neighbor data set, the vector pointing from the geometric center of the nearest neighbor data set to any target neighbor measurement data point in the nearest neighbor data set is taken as the position vector of the target measurement data point relative to the target neighbor measurement data point. The outer product of the position vector of the target measurement data point relative to the target neighbor measurement data point and itself is taken as the first vector outer product of the target measurement data point relative to the target neighbor measurement data point. For any target measurement data point's nearest neighbor data set, the sum of all first vector outer products in the nearest neighbor data set is taken as the local structure tensor of the target measurement data point. For any two target measurement data points' local structure tensors, the mean of the two target measurement data points' local structure tensors is taken as the shared local structure tensor of the two target measurement data points.

[0054] In one implementation, assume the first The center of the set of nearest neighbor data points of the measured data point points points to the set of the nearest neighbor data points of the i-th data point. The position vector of each data point is ;No. The center of the set of nearest neighbor data points of the measured data point points points to the set of the nearest neighbor data points of the i-th data point. The position vector of each data point is Then the first The measurement data point and the first The expression for calculating the shared local structure tensor of all measurement data points is:

[0055]

[0056] in, Indicates the first The measurement data point and the first A shared local structure tensor for each measurement data point; Indicates the first The center of the set of nearest neighbor data points of the measured data point points points to the set of the nearest neighbor data points of the i-th data point. The position vector of each data point; Indicates the first The center of the set of nearest neighbor data points of the measured data point points points to the set of the nearest neighbor data points of the i-th data point. The position vector of each data point; Indicates the first The set of nearest neighbor data points for each data point; Indicates the first The set of nearest neighbor data points for each data point; This represents the outer product operation between vectors.

[0057] This completes the process of fusing and constructing the local structure tensors of the measurement points to obtain the shared local structure tensors between point pairs.

[0058] Step S4: By performing principal axis deviation analysis on the shared local structure tensor, a morphology-adaptive symmetric anisotropic distance metric is obtained.

[0059] After constructing the shared local structure tensor between pairs of measurement data points, this invention further introduces a morphology-adaptive anisotropic distance metric mechanism to replace the traditional Euclidean distance calculation method, specifically addressing the directional structural characteristics of measurement data clusters in actual monitoring scenarios. In standard clustering methods, the distance between measurement points is typically calculated using Euclidean distance. This method implicitly assumes that the data cluster exhibits an isotropic structure in space, meaning that the distribution characteristics of the point cloud are essentially consistent in all directions. However, in real total station measurements, especially in scenarios with strong directional disturbances, such as crossing rivers, reservoirs, canyons, or urban alleyways along streets, measurement data points are often affected by atmospheric temperature gradients, water surface evaporation effects, or refraction disturbances caused by building obstruction. This results in a systematic stretching of the point cloud in a certain principal direction, while it remains relatively convergent or unchanged in other directions. Taking river-crossing measurements as an example, when a laser signal passes through the atmosphere above the river, the light path is horizontally deflected due to water evaporation and temperature differences between the two banks, causing a directional shift in the planar coordinates and forming a measurement point trajectory stretched along the river direction. The refraction in the vertical direction is weaker, making the fluctuation characteristics of elevation changes significantly inconsistent with the horizontal coordinates. Similarly, in urban streets, the alternating hot walls and shadows on both sides of buildings at different times cause uneven temperatures across the street, leading to a shift in the measurement point set along the street direction. In this asymmetric data structure stretched along the principal axis, continuing to use Euclidean distance for neighborhood discrimination between measurement points presents two serious problems: firstly, the distance calculation ignores the actual shape shift caused by the principal axis morphology of the point cloud, leading to misjudgments of directional connectivity; secondly, the independent distance calculation for different measurement points due to local structural differences leads to asymmetry in the distance metric, thus undermining the stability and repeatability that clustering algorithms should possess. To overcome these problems, this invention proposes an anisotropic distance metric method based on a shared local structure tensor. The core idea of ​​this method is as follows: First, perform feature decomposition on the shared structural tensor between point pairs to extract the principal axis vector that can represent the main stretching direction of the point cloud; then, combine the direction of the line connecting the measured point pairs to analyze the degree of deviation from the principal axis direction; finally, construct a directional penalty mechanism based on the degree of deviation to dynamically adjust the original Euclidean distance between point pairs, thereby forming a morphological adaptive distance measurement method that is more adaptable to the real structural morphology.

[0060] First, by performing eigenvector decomposition on the shared local structure tensor, a set of shared principal axis directions is obtained. Specifically, the shared local structure tensor is decomposed into eigenvalue pairs corresponding to the number of dimensions of the shared local structure tensor. All eigenvalue pairs are sorted in descending order according to the size of the eigenvalues, and the eigenvector corresponding to the first ranked eigenvalue is taken as the principal axis direction of the shared local structure tensor. The eigenvectors corresponding to the remaining eigenvalues ​​are taken as the secondary axis directions. The set formed by the principal axis directions and the secondary axis directions is taken as the set of shared principal axis directions.

[0061] Subsequently, a shared local principal axis deviation factor is obtained by projecting the direction of the line connecting the measurement points to the set of principal axis directions. Specifically, for any two target measurement data points, the direction vector between the two target measurement data points is obtained through their coordinates. The dot product of the direction vector between the two target measurement data points and any feature vector in the set of shared principal axis directions of the two target measurement data points is used as the first vector dot product between the two target measurement data points. The sum of all the first vector dot products between the two target measurement data points is used as the denominator, and the sum of all the first vector dot products between the two target measurement data points except for the principal axis direction is used as the numerator. The result of the calculation of the corresponding fraction is used as the shared local principal axis deviation factor between the two target measurement data points.

[0062] In one implementation, assume the first The measurement data point points to the first The direction vector of each measurement data point is ;No. The measurement data point and the first The first shared local structure tensor of the measurement data points The eigenvectors are The data dimension of the measurement data is Then the first The measurement data point and the first The formula for calculating the shared local principal axis deviation factor for all measurement data points is:

[0063]

[0064] in, Indicates the first The measurement data point and the first Shared local principal axis deviation factor for all measurement data points; Indicates the first The measurement data point points to the first The direction vector of each measurement data point; Indicates the first The measurement data point and the first The first shared local structure tensor of the measurement data points 1 eigenvector; The data latitude represents the measurement data.

[0065] It should be noted that the symmetric anisotropic distance measurement method designed in this invention aims to solve the problem of accurately identifying anisotropic data clusters caused by physical factors such as instrument axis errors or atmospheric lateral refraction. Its core lies in the introduction of a shared local structure tensor. This concept. By and The local structure tensors of each component are averaged and fused. It no longer favors any particular perspective, but objectively reflects the geometry of the transition region on the ellipsoidal trajectory of these two points. Shared principal axis obtained by eigenvalue decomposition This becomes an unbiased reference coordinate system capable of accurately describing the current extension direction of the trajectory. Based on this shared benchmark that accurately reflects the local data shape, a shared local principal axis deviation factor is established. The calculation also possesses mathematical symmetry. The line connecting two points was measured quantitatively. The direction of the data trajectory it occupies is related to the natural extension direction of the data trajectory (i.e., sharing the main extension axis). Are they consistent? If the direction of the line connecting the two points is exactly the direction in which the data trajectory extends, The value is If the line connecting two points is perpendicular to the direction of the trajectory's extension, The value is Due to calculation At that time, the same shared tensor was used. Vectors that are opposite in direction but collinear The sum of the squares of its projections on each principal axis and They are exactly the same, therefore they have This ensures the symmetry of the distance.

[0066] Finally, by fusing and weighting the original Euclidean distance and the principal axis deviation factor, a morphology-adaptive symmetric anisotropic distance is obtained. Specifically, the result of adding the constant 1 to the shared local principal axis deviation factor is used as the first distance optimization weight, and the result of multiplying the first distance optimization weight of any two target measurement data points with the Euclidean distance between the two target measurement data points is used as the morphology-adaptive symmetric anisotropic distance between the two target measurement data points.

[0067] In one embodiment, the first The measurement data point and the first The expression for calculating the shape-adaptive symmetric anisotropic distance between measurement data points is as follows:

[0068]

[0069] in, Indicates the first The measurement data point and the first Shape-adaptive symmetrical anisotropic distance between measurement data points; Indicates the first The measurement data point and the first Shared local principal axis deviation factor for all measurement data points; Indicates the first The measurement data point and the first Euclidean distance between measurement data points.

[0070] It should be noted that by applying the shared local principal axis deviation factor to the calculation of symmetric anisotropic distances, a direction-sensitive distance penalty mechanism that can intelligently identify and adapt is implemented. When the algorithm calculates the distance from the measured data points... Explore its neighbors At that time, if Subsequent points on the same trajectory due to the continuous change in lateral refraction effect will have their connecting line aligned with the direction of the data cluster's extension. It will approach Shape-adaptive symmetric anisotropic distance It is hardly magnified. This allows the algorithm to follow the trajectory formed by the physical process, thus recognizing the entire elongated ellipsoidal data cluster as a whole. Conversely, if If it is a gross error located outside the trajectory due to other accidental factors, then... arrive The connecting lines will inevitably deviate from the main axis of the trajectory. It will approach Its effective distance This will be significantly amplified. This distance penalty effectively prevents the algorithm from deviating from the normal trajectory to connect an irrelevant gross error.

[0071] Thus, the method of obtaining a morphologically adaptive symmetric anisotropic distance metric by performing principal axis deviation analysis on the shared local structure tensor is completed.

[0072] Step S5: By jointly judging the shape adaptive distance and the neighborhood radius, the gross error identification result is obtained and abnormal measurement data is removed.

[0073] After obtaining the adaptive neighborhood radius and morphologically adaptive symmetric anisotropic distance metric of the target measurement data points, a minimum data point threshold can be set, and DBSCAN clustering can be performed on the measurement data points in the original measurement dataset based on the adaptive neighborhood radius of the measurement data points and the morphologically adaptive symmetric anisotropic distance between the measurement data points. It should be noted that in this embodiment of the invention, the minimum data point threshold is set to 15. This threshold setting can be adjusted according to specific circumstances and is not required. During the clustering process, when the algorithm needs to determine the measurement data points... Are at the measurement data points? When within the neighborhood, acquire the target measurement data point. Adaptive neighborhood radius and measurement data points and A symmetric anisotropic distance metric that adapts to the shape of the measured data points and The morphologically adaptive symmetric anisotropic distance metric between the measurement data points is less than or equal to the distance between the measurement data points. Only when the adaptive neighborhood radius is used is the measurement data point determined. Located at the measurement data point Within its neighborhood.

[0074] Complete the DBSCAN clustering process and obtain the core points, boundary points and noise points marked during the clustering process; use the measurement data points marked as noise points as gross data points and delete the gross data points from the original dataset to obtain the optimized measurement dataset.

[0075] This completes the process of obtaining gross error identification results and removing abnormal measurement data by jointly judging the morphological adaptive distance and the neighborhood radius.

[0076] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for acquiring total station measurement data, characterized in that, The total station measurement data acquisition method includes: Step S1: Obtain the original measurement dataset with time-series labels by performing three-dimensional coordinate calculation and processing on the original total station measurement data; Step S2: Obtain the adaptive neighborhood radius of each point by comprehensively analyzing the local neighborhood distances of the original measurement data points; Step S3: By fusing and constructing the local structure tensors of the measurement points, the shared local structure tensors between point pairs are obtained; Step S4: Obtain a morphologically adaptive symmetric anisotropic distance metric by performing principal axis deviation analysis on the shared local structure tensor; Step S5: By jointly judging the shape adaptive distance and the neighborhood radius, obtain the gross error identification results and remove abnormal measurement data; The process of obtaining the adaptive neighborhood radius of each point by comprehensively analyzing the local neighborhood distances of the original measurement data points includes: setting the number of nearest neighbor data points and obtaining the set of nearest neighbor data points for any target measurement data point among all measurement data points; for any target measurement data point, taking the Euclidean distance between the target measurement data point and the farthest data point in its set of nearest neighbor data points as the local maximum neighborhood arm length of the target measurement data point; taking the average Euclidean distance between the target measurement data point and all data points in its set of nearest neighbor data points as the local average neighborhood arm length of the target measurement data point; and taking the square root of the sum of the squares of the local maximum neighborhood arm length and the local average neighborhood arm length of the target measurement data point as the adaptive neighborhood radius of the target measurement data point. The step of fusing and constructing the local structure tensor of measurement points to obtain the shared local structure tensor between point pairs includes: for any target measurement data point's nearest neighbor data set, taking the vector pointing from the geometric center of the nearest neighbor data set to any target neighbor measurement data point in the nearest neighbor data set as the position vector of the target measurement data point relative to the target neighbor measurement data point; taking the cross product of the position vector of the target measurement data point relative to the target neighbor measurement data point and itself as the first vector cross product of the target measurement data point relative to the target neighbor measurement data point; for any target measurement data point's nearest neighbor data set, taking the sum of all first vector cross products in the nearest neighbor data set as the local structure tensor of the target measurement data point; for any two target measurement data points' local structure tensors, taking the average of the two target measurement data points' local structure tensors as the shared local structure tensor of the two target measurement data points; The method of obtaining a morphologically adaptive symmetric anisotropic distance metric by performing principal axis direction deviation analysis on the shared local structure tensor includes: obtaining a set of shared principal axis directions by performing eigenvector decomposition on the shared local structure tensor; obtaining a shared local principal axis deviation factor by performing projection analysis on the connection direction of the measurement point pair and the set of principal axis directions; and obtaining a morphologically adaptive symmetric anisotropic distance by fusing and weighting the original Euclidean distance and the principal axis deviation factor. The step of obtaining gross error identification results and removing abnormal measurement data by jointly judging the morphological adaptive distance and neighborhood radius includes: setting a minimum data point number threshold; performing DBSCAN clustering on the measurement data points in the original measurement dataset based on the adaptive neighborhood radius of the measurement data points and the morphological adaptive symmetric anisotropic distance between the measurement data points; completing the clustering process and obtaining the core points, boundary points and noise points marked during the clustering process; and deleting the measurement data points marked as noise points from the original dataset as gross error data points to obtain a measurement dataset with abnormal measurement data removed.

2. The total station measurement data acquisition method according to claim 1, characterized in that, The process involves performing three-dimensional coordinate calculation and processing on the raw total station measurement data to obtain a raw measurement dataset with time-series labels, including: By setting multiple monitoring points in the structure or area to be measured, and using a total station automated measurement system, periodic three-dimensional coordinate measurements are performed on each monitoring point within a preset measurement cycle. The three-dimensional spatial coordinate values ​​of the monitoring point and its corresponding time label are obtained at each measurement time. The measurement results of all the monitoring points are organized in chronological order to form an original measurement dataset containing the three-dimensional coordinates and time labels of each monitoring point at multiple measurement times. The original measurement dataset consists of multiple measurement data points, and each measurement data point contains its three-dimensional coordinate values ​​and acquisition time information.

3. The total station measurement data acquisition method according to claim 1, characterized in that, The process of obtaining a set of shared principal axis directions by performing eigenvector decomposition on the shared local structure tensor includes: Eigenvector decomposition is performed on the shared local structure tensor to obtain eigenvalue-eigenvector pairs corresponding to the number of dimensions of the shared local structure tensor. All eigenvalue-eigenvector pairs are sorted in descending order according to the size of the eigenvalues, and the eigenvector corresponding to the first ranked eigenvalue is taken as the principal axis direction of the shared local structure tensor. The eigenvectors corresponding to the remaining eigenvalues ​​are taken as the secondary axis directions. The set formed by the principal axis direction and the secondary axis direction is taken as the set of shared principal axis directions.

4. The total station measurement data acquisition method according to claim 1, characterized in that, The method of obtaining the shared local principal axis deviation factor by projecting the set of measurement point pairs connecting the line direction and the principal axis direction includes: For any two target measurement data points, the direction vector between the two target measurement data points is obtained through their coordinates; the dot product of the direction vector between the two target measurement data points and any feature vector in the set of shared principal axis directions of the two target measurement data points is taken as the first vector dot product between the two target measurement data points. The result of summing all the first vector dot products between the two target measurement data points is used as the denominator, and the result of summing all the first vector dot products between the two target measurement data points except for the principal axis direction is used as the numerator. The result of the corresponding fraction is used as the shared local principal axis deviation factor of the two target measurement data points.

5. The total station measurement data acquisition method according to claim 1, characterized in that, The process of obtaining a shape-adaptive symmetric anisotropic distance by fusing and weighting the original Euclidean distance with the principal axis deviation factor includes: The result of adding the constant 1 to the shared local principal axis deviation factor is used as the first distance optimization weight. The result of multiplying the first distance optimization weight of any two target measurement data points with the Euclidean distance between the two target measurement data points is used as the morphologically adaptive symmetric anisotropic distance between the two target measurement data points.

6. A total station measurement data acquisition system, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a total station measurement data acquisition method according to any one of claims 1-5.

Images