A multi-period point cloud monitoring data processing method for a columnar member

Abstract

The present application belongs to the technical field of engineering structure point cloud monitoring, and discloses a kind of columnar component multi-period point cloud monitoring data processing method, including multi-period point cloud acquisition, coordinate unification, cross observation same name slice construction, slice joint quality evaluation, monitoring day center line optimization, period same name slice fusion, single cycle refinement gate and multi-period change calculation;According to the effective point number, sampling coverage, sampling uniformity and center line residual, slice joint quality weight is generated, according to slice quality, center line fitting quality, registration quality and high coverage rate, monitoring day fusion weight is generated, and two-stage weight fusion period same name slice is used;Standard cycle result is used for inter-period comparison, and single cycle refinement result is only used for this cycle analysis after quality gating;It can reduce the influence of local abnormal slice, low-quality repeated observation and attachment processing difference on monitoring result, improve the stability, comparability and traceability of multi-period position change and direction parameter change of columnar component.

Description

Technical Field

[0001] This invention belongs to the field of point cloud monitoring technology for engineering structures, and particularly relates to a method for processing multi-period point cloud monitoring data of columnar components. Background Technology

[0002] Steel columns of steel structure workshops, building columns, bridge piers, tower columns, and pipe gallery supports may be affected by foundation settlement, temperature changes, load adjustments, and changes in component connection status during long-term service, resulting in changes in horizontal position and directional parameters. In order to timely grasp the spatial status of columnar components, it is necessary to use mobile laser scanning equipment, SLAM scanning equipment, and ground laser scanning equipment to periodically collect point cloud data, and to uniformly process and compare point cloud data formed at different monitoring times.

[0003] Existing point cloud monitoring methods for columnar components typically extract multiple cross-sectional centers from a single point cloud and then fit the component's centerline using all cross-sectional centers. However, point clouds at different heights are easily affected by scanning distance, scanning direction, local occlusion, attachments, and connection structures, resulting in problems such as insufficient point count, unilateral concentration, and cross-sectional center offset. Existing methods are difficult to jointly evaluate the number of cross-sectional points, sampling coverage, sampling uniformity, and centerline residuals. Low-quality cross-sections may still have a significant impact on the centerline, top center, and bottom center.

[0004] Furthermore, point cloud density, registration quality, and effective height coverage vary among repeated observations within the same monitoring period. Existing methods typically select a single observation as the periodic result or directly average the final parameters of each observation, making it difficult to continuously transfer slice-level and observation-level quality to the periodic results. At the same time, performing attachment removal separately for each monitoring period can easily lead to different data processing standards, causing changes in position and orientation parameters during the period to be affected by non-structural factors, reducing the stability and comparability of long-term monitoring results. Summary of the Invention

[0005] This invention aims to solve the problems in existing multi-period point cloud monitoring of columnar components, such as inconsistent slice height ranges and sampling states in different observations, centerline fitting with equal weight for each section center, low-quality slices and low-quality repeated observations easily affecting the periodic representative results, and the direct participation of single-period attachment processing results in periodic comparisons changing the point cloud processing caliber for each period, resulting in non-structural factors interfering with changes in the position and orientation parameters of columnar components.

[0006] To address the aforementioned technical problems, this invention provides a method for processing multi-period point cloud monitoring data of columnar components, including multi-period point cloud acquisition, coordinate benchmark unification, column body extraction, cross-observation homologous slice construction, slice center extraction, slice joint quality evaluation, monitoring day centerline optimization, monitoring day quality evaluation, period homologous slice fusion, standard period result generation, single-period attachment refinement, refinement result quality gating, multi-period change calculation, and monitoring result output.

[0007] Acquire point cloud data of the same columnar component at multiple monitoring times, configure the monitoring object number, observation number, observation time, monitoring cycle, coordinate system identifier and length unit for each observation, and convert the point cloud of each observation to a unified internal length unit.

[0008] Within the monitoring area, a stable reference area, fixed control markers, and stable components that do not participate in the monitoring are selected. Based on the stable reference objects, the point clouds of each observation are transformed to the project reference coordinate system. The spatial transformation relationship and registration quality corresponding to each observation are saved. During the registration process, the columnar components to be monitored and their change-sensitive areas are excluded from the registration reference area.

[0009] Extract the main point cloud of columnar components from the point cloud with unified coordinates, establish a fixed local coordinate system for the project, and set fixed slices along the height direction of columnar components between the unified bottom height and the unified top height used in all monitoring cycles, so that slices with the same slice number in different observations correspond to the same height interval, and generate fixed slices with the same name across observations.

[0010] Outlier suppression and cross-section center calculation are performed on fixed slices with the same name in each observation. The effective number of slice points, slice center, lateral point cloud range, angular region occupancy status, point distribution status and slice effective status are saved. The initial centerline of the monitoring day is fitted using multiple effective slice centers.

[0011] A sampling quantity index is generated based on the number of valid points in the slice, a sampling coverage index is generated based on the occupancy status of the cross-sectional angle region, a sampling uniformity index is generated based on the difference in the number of points in each angle region, a centerline spatial consistency index is generated based on the horizontal residual of the slice center relative to the initial centerline of the monitoring day, and a joint slice quality weight is generated based on the above indices.

[0012] Based on the joint quality weight of the slices, a weighted centerline fitting is performed on the slice centers of each observation to obtain the optimized centerline for the monitoring day. The slice centerline residuals and joint quality weights of the slices are then recalculated based on the optimized centerline until the changes in the centerline parameters meet the convergence condition or reach the preset number of iterations.

[0013] The overall quality index for each monitoring day is generated based on the effective slice quality, centerline fitting quality, point cloud registration quality, and effective height coverage of each observation. The overall quality index for each monitoring day is then normalized within the same monitoring period to generate the monitoring day fusion weight.

[0014] For multiple valid slices with the same slice number within the same monitoring period, a combined weight is generated based on the monitoring day fusion weight and the slice joint quality weight. The slice centers of each observation are fused using the combined weight to obtain the slice centers with the same name in each period. The number of valid observation support and fusion status of slices with the same name in each period are recorded.

[0015] The standard period representative centerline is fitted using the center of the periodic corresponding slice. Based on the standard period representative centerline, the bottom center and top center of the standard period are calculated at a uniform bottom height and a uniform top height. The standard period results are generated, including the period representative centerline parameters, top and bottom centers, position parameters, orientation parameters, number of effective observations, number of effective slices, height coverage, and fusion quality status.

[0016] In addition to the standard period results, a single-period attachment refinement branch is established. The sampling quantity of the column main body point cloud that has completed coordinate unification within the same monitoring period is balanced and the point cloud is merged to generate a period representative column point cloud. The column base components, column top connecting components, local additional point clusters and other non-main body points in the period representative column point cloud are processed. The same-name slice, slice center, slice quality, center line and top and bottom center are recalculated and fixed to generate a single-period refinement result.

[0017] Quality gating is applied to the single-cycle refinement results based on the point retention rate, effective slice retention rate, effective height coverage, cross-sectional sampling coverage change, top position change, and centerline stability before and after refinement. If the quality gating conditions are met, the single-cycle refinement results are applied to the single-cycle analysis. If the quality gating conditions are not met, the results are reverted to the standard cycle results, and the single-cycle refinement results are set to be prohibited from participating in multi-cycle comparisons.

[0018] Select the reference period and comparison period, call only the standard period results of the two periods, calculate the changes in the top position, bottom position, axis component, axis increment vector during the period, the difference in the horizontal offset amplitude of the period itself, and the changes in the direction parameters, and generate multi-period change results of the columnar component.

[0019] Output the slice quality and joint slice quality weight for each observation, the optimized centerline for the monitoring day, the overall quality index for the monitoring day, the fusion results of slices with the same name for the period, the standard period results, the single-period refinement quality gating status, the multi-period change results, coordinate system identifier, unit, algorithm version, processing parameters, abnormal status and rollback reasons, forming a traceable multi-period point cloud monitoring data processing result.

[0020] The beneficial effects of this invention are as follows: (1) Under the conditions of unified coordinate reference, unified height range and unified slice number, construct cross-observation fixed same-name slices, and generate slice joint quality weights based on effective point number, sampling coverage, sampling uniformity and centerline residuals, so that low-density, unilaterally concentrated and significantly deviated slices reduce the influence on centerline fitting, and solve the problem of inconsistent positions of different observation slices and interference of low-quality sections on centerline calculation.

[0021] (2) Based on the slice quality, centerline fitting quality, registration quality and effective height coverage, the monitoring day fusion weight is generated, and multiple repeated observations are fused with slices of the same name within the period as objects, so that slice-level quality information and monitoring day-level quality information are continuously transmitted to the standard period representative centerline, solving the problem that selecting a single observation and directly averaging each final parameter is difficult to reflect local occlusion and observation quality differences.

[0022] (3) Set up two isolated data processing branches for standard period results and single-period refinement results, and perform quality gating on the refinement results through point retention rate, effective slice retention rate, high coverage rate and centerline stability, so that single-period attachment processing can improve the quality of the main analysis, while prohibiting refinement results with different processing calibers from entering the period comparison, ensuring the consistency, comparability and traceability of multi-period change results. Attached Figure Description

[0023] The accompanying drawings are provided to further understand the present invention and form part of the specification. They are used together with the embodiments of the present invention to explain the invention and do not constitute a limitation thereof.

[0024] Figure 1 This is a flowchart of the multi-period point cloud monitoring data processing method for columnar components according to the present invention; Figure 2 This is a schematic diagram illustrating the processing of standard periodic results and single-period refinement results in this invention. Detailed Implementation

[0025] Example 1, see Figures 1-2 The present invention provides a method for processing multi-period point cloud monitoring data of columnar components, including multi-period point cloud acquisition steps, coordinate benchmark unification steps, column body extraction steps, cross-observation homonymous slice construction steps, slice center extraction steps, slice joint quality evaluation steps, monitoring day centerline optimization steps, monitoring day quality evaluation steps, period homonymous slice fusion steps, standard period result generation steps, single period attachment refinement steps, refinement result quality gating steps, multi-period change calculation steps, and monitoring result output steps.

[0026] This embodiment is applied to a long-term deformation monitoring project of steel columns in a large steel structure factory building. The factory building has a span of 36m and contains 36 box-shaped steel columns with a cross-sectional dimension of 500mm×500mm. The monitoring height of the columns is 11.8m. Stiffening plates and connecting bolts are installed at the column base, and beam-column connection plates are installed at the column top. Cable brackets, fire pipe clamps, and temporary measurement marks are attached to the surface of some steel columns. There are obstacles on site, such as equipment obstruction, passage of transport vehicles, and limited scanning station positions.

[0027] The project uses handheld SLAM laser scanning equipment for periodic data collection and ground-based laser scanning equipment to verify some steel columns. A monitoring cycle is formed every two months. Each monitoring cycle consists of three consecutive days of independent observation. Each observation restarts the scanning equipment, replans the scanning route, and re-performs scene registration, so that the three observations within the cycle can reflect differences in scanning direction, equipment drift, and local occlusion.

[0028] Step S1: Obtain point cloud data of the same steel column formed at multiple observation times, record the monitoring object number, observation number, acquisition time, monitoring cycle, scanning equipment number, internal length unit and coordinate system identifier of each point cloud, and convert all point clouds into point cloud data with meters as the internal length unit.

[0029] Step S2: Select a stable reference area from the fixed control markers on the factory floor, adjacent stable walls, and equipment foundations that do not participate in deformation monitoring. Convert each point cloud to the project reference coordinate system. During the registration process, shield the monitored steel column and deformation-sensitive components connected to the steel column to avoid absorbing the actual displacement of the steel column into the registration transformation.

[0030] Step S3: Establish a monitoring bounding box based on the design location of the steel column, extract candidate point clouds of the target steel column area from the point cloud of the unified coordinate system, and then extract the main point cloud of the column based on the main direction of the column, the range of cross-sectional dimensions, vertical continuity and horizontal distance from the design axis. Establish a fixed local coordinate system for the project. The local Z direction of the local coordinate system is along the design height direction of the steel column, and the local X direction and local Y direction are located in the cross-sectional plane of the steel column.

[0031] Step S4: Determine a uniform bottom height and a uniform top height within the common effective height range for all monitoring cycles. In this project, the uniform bottom height is set at 0.60m from the ground level, and the uniform top height is set at 11.20m from the ground level. 53 slices with the same name are divided according to a fixed slice thickness of 0.20m, so that the same slice number in any monitoring day and any monitoring cycle corresponds to the same height range.

[0032] Step S5: Perform outlier suppression and cross-section center calculation on the point cloud within each slice, and save the original number of points, the number of valid points, the slice center, the lateral point cloud range, the angular region occupancy status, the point distribution status, the slice validity status, the observation number, and the monitoring cycle number.

[0033] Step S6: Fit the initial centerline of the monitoring day using the effective slice centers from each observation, and generate joint slice quality weights based on the number of effective points, the coverage of the angular region, the uniformity of the number of regional points, and the horizontal residual of the slice center relative to the initial centerline.

[0034] Step S7: Refit the daily optimized centerline based on the slice joint quality weights, and iteratively update the slice centerline residuals and slice joint quality weights until the horizontal position of the top of the centerline formed by two adjacent fittings does not exceed 0.20 mm and the centerline direction parameter does not exceed 0.005°, or until five iterations are completed.

[0035] Step S8: Generate the overall quality index of the monitoring day based on the effective slice joint quality, centerline fitting quality, point cloud registration quality and effective height coverage, and normalize the overall quality index of each monitoring day into the monitoring day fusion weight within the same monitoring period.

[0036] Step S9: Merge slice centers with the same slice number within the same monitoring period. Each time an observation participates in the fusion of a slice with the same name, it is simultaneously controlled by the monitoring day fusion weight and the slice joint quality weight. If an observation lacks a valid slice at the corresponding height, it will not participate in the fusion of the corresponding slice with the same name.

[0037] Step S10: Calculate the representative center line of the standard period using the center of the same slice after period fusion, and calculate the bottom center and top center of the standard period at the same bottom height and the same top height respectively based on the representative center line of the standard period, to generate the standard period result for comparison of different monitoring periods.

[0038] Step S11: In addition to the standard cycle results, establish a single-cycle attachment refinement branch, perform sampling quantity balancing and point cloud merging on the column main body point cloud with unified coordinates within the same monitoring cycle to generate a cycle representative column point cloud, and then perform non-main body point processing on the column foot connectors, column top connectors, cable brackets, pipe clamps and local isolated point clusters in the cycle representative column point cloud to generate a single-cycle refined point cloud.

[0039] Step S12: Recalculate the fixed corresponding slices, slice center, slice quality, centerline, and top and bottom centers based on the single-cycle refined point cloud. Perform quality gating based on the point retention rate, effective slice retention rate, effective height coverage, cross-sectional coverage change, and centerline stability. Generate single-cycle refined results when the quality gating conditions are met; otherwise, revert to the standard cycle results.

[0040] Step S13: Select the reference period and comparison period, and only call the standard period results of the two periods to calculate the top position change, bottom position change, axis component change, axis increment vector, period self-horizontal offset amplitude difference and direction parameter change, without reading the single period refinement results.

[0041] Step S14 outputs the slice center, slice quality index, slice joint quality weight, monitoring day optimization centerline, monitoring day overall quality index, monitoring day fusion weight, periodic slice center, standard period representative centerline, single period refinement quality gating status, multi-period change results, coordinate system identifier, length unit, algorithm version, parameter version, rollback reason, and manual review status.

[0042] Through the above processing, this method establishes a continuous transfer relationship between local slice quality, single observation quality and monitoring cycle quality, and sets the standard cycle comparison results and single cycle attachment refinement results as mutually isolated data branches, thus solving the problem of non-structural influences on the long-term changes of columnar components caused by local occlusion, sampling bias, low-quality repeated observations and differences in attachment treatment in each period.

[0043] Example 2: This example is based on all the above examples. The steps of acquiring multi-period point clouds, unifying coordinate references, extracting the main body of the cylinder, and constructing cross-observation homonymous slices are specifically performed as follows.

[0044] Before each scan begins, an observation task is created in the project data management system. The observation task must record at least the steel column number, monitoring cycle number, observation number, scanning equipment number, scanning personnel number, start time, and end time. Three observations within the same monitoring cycle are saved as independent data objects, and point clouds collected over three consecutive days are not directly merged into a single raw file.

[0045] The SLAM scanning equipment starts from the fixed entrance of the factory building and moves along the scanning route, passing through the stable wall, fixed control signs, the area around the target steel column, and the foundation of the stable equipment in sequence. After the scanning is completed, it returns to the fixed entrance to form a closed route. The ground laser scanning equipment is set up on both sides of the target steel column and is converted to the project reference coordinate system through the fixed control signs.

[0046] The coordinate reference adopts a phased processing method. First, the initial rigid transformation is completed according to the fixed control marks. Then, local fine registration is performed using stable walls and stable equipment foundations. During the fine registration process, the point cloud of the target steel column and components within 1.5m of the outer surface of the target steel column is excluded. This project requires that the root mean square error of the registration residual in the stable reference area does not exceed 4mm.

[0047] Each coordinate transformation saves the original coordinate system identifier, target coordinate system identifier, rotation parameters, translation parameters, registration reference area, and registration residual. After the point cloud enters the project reference coordinate system, a coordinate transformation completion marker is written. Subsequent modules will not repeat the coordinate transformation when they detect the coordinate transformation completion marker.

[0048] The point cloud extraction of the main body of the column first establishes candidate bounding boxes according to the horizontal range of 0.80m and the height range of 0.30m to 11.50m around the design axis of the steel column. Then, according to the cross-sectional dimensions of the column, points that are more than 0.45m away from the design axis are deleted, and suspended point clusters that are not continuous with the main body of the steel column are deleted according to the vertical continuity.

[0049] In the standard periodic branch, the column base stiffening plate and the column top connecting plate are not subject to forced complete deletion. Instead, the uniform bottom height and uniform top height are set in positions where the connection structure has less impact, so that each period participates in the comparison according to the same height range and the same treatment caliber.

[0050] Cross-observation fixed-name slices are established according to Formula 1:

[0051] In Formula 1: This represents the lower boundary height of the i-th slice with the same name; This represents the height of the upper boundary of the i-th slice with the same name; This indicates a uniform bottom height used throughout the entire monitoring cycle; This indicates the uniform top height used throughout the entire monitoring cycle; Indicates a fixed slice thickness; This represents the fixed representative height of the i-th slice with the same name.

[0052] The unified bottom height and unified top height are written into the monitoring object configuration file when the first monitoring cycle is established. Subsequent monitoring cycles shall not change them individually based on the current point cloud coverage. If the top area point cloud is missing on a certain monitoring day, the corresponding top slice will be marked as missing, and the slice numbering system for this observation will not be shortened.

[0053] This project divides the cross-sectional plane into 12 angular regions centered on the initial design axis. The included angle of each angular region is 30°. The direction of the fixed angular region is determined according to the local coordinate system of the project and does not rotate with the main direction of each point cloud, so that the sampling coverage index in different observations has a unified spatial meaning.

[0054] Regarding parameter adjustments: When the column height is less than 6m, the fixed slice thickness can be set to 0.10m to 0.15m. When the column height exceeds 20m and the point cloud density is low, the fixed slice thickness can be set to 0.25m to 0.40m. Once the slice thickness is used for the first standard period result, it remains unchanged in subsequent multi-period comparisons.

[0055] Through the above processing, the point clouds of each observation are incorporated into the same project coordinate system based on the stable reference area, and divided according to the same bottom height, top height and slice number, so that slices with the same number have a consistent spatial meaning in all observations, thus solving the problem of slice misalignment caused by each observation independently layering according to its own point cloud range.

[0056] Example 3: This example is based on all the above examples. The slice center extraction step, the initial centerline fitting step, and the slice joint quality evaluation step are specifically performed in the following manner.

[0057] For the i-th slice of the d-th observation, extract the point cloud with height within the fixed slice range, mark the points that are more than the sum of the cross-sectional design radius and the allowable additional distance from the lateral median of the slice point cloud as lateral outliers, and then delete the isolated points based on the number of local neighborhood points.

[0058] The center of the box-shaped steel column section is calculated using a robust constraint method with four-sided point clouds. Local positive X-side, local negative X-side, local positive Y-side, and local negative Y-side point clouds are extracted respectively. When there are effective samples on both sides of the same pair, the center in the corresponding direction is determined by the midpoint of the robust positions on both sides. When one side is missing, the temporary center is estimated by the design section width and the robust position on the other side, and the coverage quality and uniformity quality of the corresponding slice are reduced.

[0059] The center of each slice is recorded as follows:

[0060] in, and This represents the horizontal center coordinates of the i-th slice in the d-th observation. A fixed slice with the same name is used to represent the height, instead of the average height of the actual point cloud within the slice.

[0061] On the monitoring day, the initial center line is subjected to robust linear fitting in the local X and local Y directions. The initial center line is only used to determine the degree of deviation of the slice center from the spatial trend of the main body of the column, and is not directly used as the final monitoring result.

[0062] For each slice, calculate the sampling quantity index, angular region coverage index, angular region uniformity index, and centerline spatial consistency index, and construct the joint quality weight of the slices according to Formula 2:

[0063] In Formula 2: This indicates the number of slices sampled. Indicates the number of valid points in the slice; Indicates the coverage index of the angular area; Indicates the number of angular regions with valid point cloud; Indicates the total number of preset angle regions; An index representing the uniformity of the number of points in an angular region; This represents the number of valid points within the k-th angular region; This represents the horizontal residual of the slice center relative to the initial centerline of the monitoring day; and These represent the predicted center coordinates of the initial centerline at height z_i on the monitoring day; This represents the scale of the slice center residual for the d-th observation; This indicates the consistency index of the centerline space; This represents the joint quality weight of the slices after normalization; These represent the influencing parameters corresponding to the number of samples, angular region coverage, regional uniformity, and spatial consistency of the centerline, respectively. Indicates the lower bound of slice weights; This indicates a positive number that prevents the denominator from being zero.

[0064] In this project, slices with fewer than 80 valid points are not deleted directly; instead, the sampling quantity index is limited to a lower range. Slices with fewer than 4 angular regions are marked as low-coverage slices. Slices with a center horizontal residual exceeding three times the monitoring day residual scale are marked as abnormal slices.

[0065] When there are many sampling points, but they are mainly concentrated on the side of the steel column facing the scanning equipment, the sampling quantity index may remain high, but the angular area coverage index and uniformity index will decrease, making it impossible for the corresponding slice to obtain high weight solely based on the number of sampling points.

[0066] When the local attachment point clusters are parallel to the column surface and there are many points, the slice center may shift locally. At this time, the spatial consistency index of the centerline decreases, and the influence of the attachment point clusters is further suppressed through residual evaluation.

[0067] Regarding parameter adjustments: When the scanning station can cover the four sides of the column, improve the impact parameters corresponding to the coverage and uniformity of the angle area. When the mobile SLAM scan mainly covers the two sides of the column, reduce the impact parameters of the coverage of the angle area and improve the impact parameters of the spatial consistency of the centerline. Once a parameter version is used for the same long-term monitoring project, the change time and reason for the change must be recorded.

[0068] Through the above processing, the slice quality simultaneously reflects the number of effective points, sampling orientation, sampling balance, and deviation from the overall spatial trend of the column, solving the problems that the point count index alone cannot identify unilateral dense sampling, and the residual index alone cannot reflect the point cloud coverage quality.

[0069] Example 4: Based on all the above examples, the calculation steps for the optimized centerline on the monitoring day and the overall quality evaluation steps on the monitoring day are specifically performed as follows.

[0070] The daily optimized centerline is calculated by performing weighted fitting on the local X and Y directions based on the joint quality weights of the slices, according to Formula 3:

[0071] In Formula 3: and Let represent the slope parameters of the optimized centerline in the local X and Y directions, respectively, for the d-th observation. and These represent the intercept parameters of the optimized centerline in the local X and Y directions, respectively, for the d-th observation. This represents the optimized centerline of the d-th observation.

[0072] After completing one weighted fitting, the horizontal residuals of each slice, the spatial consistency index of the centerline, and the joint quality weight of the slices are recalculated based on the optimized centerline. Weighted fitting is then performed again, and the centerline parameters, the change in the top centerline, and the number of abnormal slices are retained during each iteration.

[0073] When the horizontal distance between two adjacent iterations of the top center is less than 0.20 mm and the change in the centerline direction angle is less than 0.005°, the iteration is stopped. If the iteration fails to converge after five iterations, the daily fitting stability is marked as a low-stability state.

[0074] The effective slice quality statistics are generated based on the joint quality weight of all effective slices and the proportion of low-quality slices. The centerline fitting quality is generated based on the weighted residual scale and the iterative convergence state. The registration quality is generated based on the registration residual of the stable reference area. The effective height coverage is generated based on the ratio of the actual height range supported by effective slices to the unified target height range.

[0075] Calculate the overall quality index of the monitoring day and the fusion weight of the monitoring days within the cycle according to Formula 4:

[0076] In Formula 4: This represents the overall quality index for the monitoring day of the d-th observation; This represents the statistical value of the effective slice quality; Indicates the quality of the centerline fit; Indicates the registration quality of the point cloud; Indicates effective height coverage; These represent the influencing parameters corresponding to the four types of daily quality factors monitored; This represents the normalized daily fusion weight of the d-th observation within the k-th monitoring period; This represents the set of valid observations included in the k-th monitoring period.

[0077] This project defines observations that have a stable reference area registration residual of no more than 4 mm, an effective height coverage of no less than 85%, and an effective number of slices of no less than 70% of the total number of slices as valid observations that can participate in periodic fusion.

[0078] When the registration quality of a certain observation is low but the sampling status of some slices is good, the overall quality index of the monitoring day decreases, which reduces the impact of this observation on the entire cycle. At the same time, higher quality slices can still participate in the fusion of slices with the same name with limited weight, and all effective local information is not completely deleted due to a single overall quality factor.

[0079] If only one observation in the same monitoring period meets the valid observation conditions, a normal period fusion result will not be generated. Instead, the current period will be marked as an insufficient observation support state, and a supplementary observation task will be generated. A single observation will not be directly disguised as a fusion result of multiple observations.

[0080] Regarding parameter adjustments: When there are many stable control points and the ground laser scanning accuracy is high, improve the parameters that affect registration quality. When mobile SLAM equipment is collecting data in narrow scenes, improve the parameters that affect effective height coverage and centerline fitting quality to ensure that the observation weights correspond to the actual acquisition conditions.

[0081] Through the above processing, the daily monitoring quality not only reflects the centerline fitting residual, but also takes into account the slice quality, coordinate registration and effective height coverage, so that observations with poor local point cloud conditions, large registration residuals and insufficient coverage height are less affected in periodic fusion.

[0082] Example 5: This example is based on all the above examples. The steps of merging slices with the same period name, calculating the center line of the standard period, and calculating the top and bottom centers of the standard period are specifically performed as follows.

[0083] The k-th monitoring period contains three independent observations. For a fixed slice number i, the effective status, slice center, slice joint quality weight, and monitoring day fusion weight of the corresponding slice in the three observations are retrieved respectively.

[0084] If an observation is occluded at the i-th slice and is marked as an invalid slice, this observation will not participate in the fusion of the i-th slice with the same name, but other valid slices in this observation can still participate in the fusion at the corresponding height.

[0085] Calculate the center of the slice with the same name for the same period and the representative center line of the standard period according to Formula 5:

[0086] Formula 5: This represents the combined weight when the d-th observation participates in the fusion of the i-th slice with the same name in the k-th monitoring cycle; This represents the periodic fusion center of the i-th slice with the same name in the k-th monitoring period; This represents the set of observations that have valid data at the i-th slice during the k-th monitoring period; The standard period for the kth monitoring period represents the centerline; This indicates the bottom center of the standard period in the k-th monitoring cycle; This represents the top center of the standard cycle in the kth monitoring period.

[0087] The number of valid observations is recorded simultaneously at the center of the slice with the same name in the period. When the number of observations is three, it is marked as a complete support slice; when the number of observations is two, it is marked as a partial support slice; and when the number of observations is one, it is marked as a weak support slice.

[0088] Weakly supported slices can participate in the fitting of the standard period centerline, but their final weight is modified by the number of supports. In this project, the final weight of weakly supported slices is limited to within 50% of the average weight of full supported slices.

[0089] The standard period represents the center line, which is weighted and fitted using the final weights of the period-named slices. The unified top center and the unified bottom center are both calculated using the standard period represents the center line, without directly using the average coordinates of the point clouds in the top and bottom regions.

[0090] The standard period itself includes the local X-axis component relative to the bottom, the local Y-axis component, the horizontal offset amplitude, the uniform comparison height, and the cylinder orientation angle, and saves the number of valid observations, the number of valid period slices, the number of weakly supported slices, the height coverage, and the fusion quality status.

[0091] In the first monitoring cycle of this project, the overall quality indicators for the three monitoring days were 0.91, 0.76 and 0.58, respectively, and the normalized fusion weights for the monitoring days were 0.405, 0.338 and 0.257, respectively. In the third observation, multiple upper slices were missing due to equipment obstruction. The missing slices did not participate in the corresponding height fusion, while the slices that were not missing still participated in the cycle results according to the combined weights.

[0092] Compared with directly calculating the arithmetic mean of the final top coordinates of three observations, the fusion of corresponding slices can preserve the local quality differences of each observation at different heights, avoid discarding all information when the overall quality of an observation is low but some height point clouds are good, and also avoid local abnormal slices from entering the periodic results with the final parameter averaging.

[0093] Through the above processing, slice-level quality information and daily monitoring quality information are jointly transmitted to the standard cycle representative centerline, so that the cycle results are based on multiple effective observations at the same height position, solving the problem that directly averaging the centerline parameters of each observation cannot reflect local occlusion and slice quality differences.

[0094] Example 6: Based on all the above examples, the single-cycle attachment refining step and the refining result quality gating step are specifically performed as follows.

[0095] For the k-th monitoring period, the point cloud of the main body of the column observed in each period is transformed to a unified local coordinate system of the project, and the sampling quantity between observations is balanced according to the fixed voxel size to avoid the observation with a large number of points occupying too high a proportion in the point cloud of the representative column of the period.

[0096] After balancing the number of samples, the observation point clouds from each observation are merged. The standard period is used to represent the center line to establish the main body envelope region of the column. Isolated point clusters that are far from the main body envelope region, lateral protruding point clusters that only appear at local heights, point clouds of column base stiffening plates, point clouds of column top connecting plates, point clouds of cable supports, and point clouds of pipe clamps are identified.

[0097] Non-major point processing is performed only within the single-cycle refinement branch; the standard cycle results and the cycle-named slices used for multi-cycle comparisons are not changed by this step.

[0098] After completing the non-main point processing, the single-cycle refinement center line and the single-cycle refinement top and bottom centers are recalculated according to the same uniform height range, fixed slice number, slice center calculation method and slice quality evaluation method as the standard cycle branch.

[0099] Calculate the single-cycle refinement quality gate parameters according to Formula 6:

[0100] In Formula Six: This represents the point retention rate after refinement in the k-th monitoring period; Indicates the number of column points retained after refinement; The period before the refining process represents the number of column points; Indicates the effective slice retention rate; This indicates the number of effective slices after refinement. This indicates the number of effective cycle slices before the refining process; Indicates the effective height coverage after refinement; This indicates that the refined point cloud has a height range that is effectively supported by slicing. Indicates a unified target altitude range; This indicates the horizontal change between the standard top position and the refined top position before and after the refining process.

[0101] This project sets the following basic gating conditions for refining the results: point retention rate of no less than 70%, effective slice retention rate of no less than 90%, effective height coverage of no less than 85%, top position variation of no more than 8mm, and centerline direction variation of no more than 0.08°.

[0102] The median value of the cross-sectional angle region coverage index after refinement should not be lower than 90% of the median value before refinement, to prevent the non-subject point identification process from deleting a large number of point clouds of the true side of the column.

[0103] When the refinement result meets all gate control conditions, it is written to the single-cycle refinement result library and the result is marked as single-cycle analysis. When the refinement result does not meet any key gate control conditions, the reason for refinement failure is written, and the single-cycle analysis page automatically calls the standard cycle result.

[0104] The data attributes of the single-cycle refinement results are set as follows: When the multi-period change calculation module reads the results, it first checks the above data attributes. If the data attribute is false, it is prohibited to enter the reference period and comparison period selection list.

[0105] There is a temporary cable support on the surface of the column in a certain period. Refinement can reduce the influence of the support point cloud on the center line of a single period. However, the support has been removed in the next monitoring period. If the two periods are directly compared using different refinement results, there will be a difference in the processing caliber. Therefore, multi-period comparison always reads the standard period result that has not been affected by the single-period attachment treatment.

[0106] Regarding parameter adjustments: When the proportion of attachments in the column point cloud is high, increase the point retention rate limit; when the column base and top structures are complex, increase the safety distance between the uniform height boundary and the connection structure; after adjusting the gating threshold, save the parameter version and prohibit direct mixing and comparison of results from the old and new parameter versions.

[0107] Through the above processing, the attachment refinement can improve the quality of the column body analysis in a single monitoring cycle. At the same time, by using usage marking and quality gating, it can prevent the erroneous rejection of refinement and the inclusion of differences in processing between cycles into long-term change results, thus resolving the contradiction between improving the accuracy of a single cycle and the consistency of the comparison caliber across multiple cycles.

[0108] Example 7: This example is based on all the above examples. It uses the data processing process of the box-shaped steel column with the number C17 in the aforementioned steel structure workshop for two monitoring cycles to illustrate the multi-cycle change calculation and result output.

[0109] Reference period A includes three independent observations. The number of effective slices on the three monitoring days are 52, 49 and 44, respectively. The registration residuals of the stable reference area are 2.1 mm, 3.0 mm and 3.8 mm, respectively, and the effective height coverage is 98.1%, 92.5% and 84.9%, respectively.

[0110] Comparison period B also includes three independent observations, with the number of effective slices on the three monitoring days being 51, 50 and 46, respectively. The registration residuals of the stable reference area are 2.4 mm, 2.7 mm and 3.5 mm, respectively, and the effective height coverage is 96.2%, 94.3% and 88.7%, respectively.

[0111] Reference period A and comparison period B are used to generate standard period representative center lines according to the same-name slice fusion method. The bottom center and top center of the standard period are calculated at the same bottom height and the same top height.

[0112] Calculate the changes in position and orientation parameters between different monitoring periods using Formula 7:

[0113] In Formula 7: and These represent the local X-direction component and the local Y-direction component of the top relative to the bottom in the k-th monitoring period, respectively; This represents the horizontal offset amplitude of the k-th monitoring period. These represent the changes in the axis components of the comparison period relative to the reference period; This represents the horizontal magnitude of the axis increment vector during the cycle. This represents the difference in the horizontal offset amplitude between two monitoring periods; This represents the difference in the azimuth angle between two monitoring cycles; This indicates the unified comparison height used in both monitoring cycles.

[0114] The local X-axis component of the top relative to the bottom of reference period A is 6.8 mm, the local Y-axis component is 3.4 mm, and the horizontal offset amplitude is 7.60 mm; the local X-axis component of comparison period B is 10.9 mm, the local Y-axis component is 5.1 mm, and the horizontal offset amplitude is 12.04 mm.

[0115] This yields a local X-axis component change of 4.1 mm, a local Y-axis component change of 1.7 mm, a horizontal amplitude of the periodic axis increment vector of 4.44 mm, and a difference in the horizontal offset amplitude of the period itself of 4.44 mm. Although the two values ​​are close in this data, the system still saves them separately and does not merge them into the same index because the values ​​are close.

[0116] In comparison cycle B, there is a temporary fire pipe clamp near the column height of 6.8m. After deleting the corresponding non-main points in the single-cycle refinement branch, the refinement point retention rate is 92.6%, the effective slice retention rate is 98.1%, the effective height coverage rate is 96.2%, and the refinement top position changes by 1.9mm compared to the standard cycle top position, which meets the quality gate control conditions. Therefore, the single-cycle analysis page displays the refinement center line.

[0117] The multi-period comparison module still reads the standard period representative centerline of comparison period B, but does not read the refined centerline. The monitoring report marks "standard period result" in the multi-period change table and "refined result is only for the main analysis of this period" in the single-period analysis appendix.

[0118] The system further outputs the local X-direction change at the top, the local Y-direction change at the top, the local X-direction change at the bottom, and the local Y-direction change at the bottom, respectively, to distinguish between the overall translation of the column and the change in the directional parameters of the top relative to the bottom.

[0119] When the changes during the week exceed the manual review threshold set by the project, the system does not directly output a structural safety conclusion. Instead, it generates point cloud registration review, low-quality slice review, on-site component status inspection and supplementary scanning tasks, and provides all slice weights, monitoring day fusion weights, standard cycle centerline parameters and abnormal slice locations to the engineers.

[0120] The monitoring results report includes the project reference coordinate system, project local coordinate system, unified bottom height, unified top height, fixed slice thickness, number of observations within the cycle, overall quality indicators for the monitoring day, number of valid slices with the same name in the cycle, standard cycle position parameters, single-cycle refined gating status, multi-cycle variation parameters, algorithm version, and processing parameter version.

[0121] Through the above processing, this method can, under the conditions of repeated scanning and long-term periodic monitoring of actual steel structure workshops, progressively transmit the differences in local slice sampling, the registration quality of monitoring days, and the differences in observation within the cycle to the standard cycle results. At the same time, it restricts the refinement of single-cycle attachments to an independent analysis branch, so that the changes in the horizontal position and directional parameters of columnar components have a unified calculation caliber, clear quality sources, and complete traceability information.

Claims

1. A method for processing multi-period point cloud monitoring data of columnar components, characterized in that, Includes the following steps: Step S1: Obtain point cloud data of the same columnar component formed at multiple monitoring times, configure the monitoring object number, observation number, observation time and monitoring cycle for each observation point cloud, unify the length unit of each observation point cloud, and generate a multi-period columnar component point cloud dataset. Step S2: Receive the multi-period columnar component point cloud dataset, convert the observation point cloud of each observation to the project reference coordinate system according to the stable reference object in the monitoring area, extract the main point cloud of the columnar component from the point cloud after coordinate transformation, establish a fixed project local coordinate system, and divide cross-observation same-name slices with fixed slice numbers and fixed height intervals between the unified bottom height and the unified top height to generate an observation slice dataset. Step S3: Receive the observation slice dataset, perform outlier suppression and cross-sectional center calculation on each cross-observation slice with the same name, generate slice centers, fit the initial centerline of the monitoring day using multiple effective slice centers corresponding to the same observation, and generate joint slice quality weights based on the number of effective points in the slice, cross-sectional sampling coverage status, cross-sectional sampling uniformity status, and the horizontal residual of the slice center relative to the initial centerline of the monitoring day. Step S4: Receive the slice center and the slice joint quality weight, perform weighted centerline fitting on multiple slice centers of the same observation according to the slice joint quality weight, generate the monitoring day optimized centerline, and update the horizontal residual and the slice joint quality weight according to the monitoring day optimized centerline until the centerline parameter change meets the convergence condition or reaches the preset number of iterations; Step S5: Receive the optimized centerline of the monitoring day and the joint quality weight of the slice; generate the overall quality index of the monitoring day based on the effective slice quality, centerline fitting quality, point cloud registration quality and effective height coverage of each observation; and perform normalization on the overall quality index of the monitoring day within the corresponding monitoring period to generate the monitoring day fusion weight. Step S6: Receive the monitoring day fusion weight, the slice joint quality weight, and the slice center. For valid slices with the same slice number within the same monitoring period, generate a combined weight based on the monitoring day fusion weight and the slice joint quality weight. Fuse the slice centers corresponding to each valid slice based on the combined weight to generate slice centers with the same name for the period. Fit a standard period representative centerline using multiple slice centers with the same name for the period. Generate a standard period result based on the standard period representative centerline. Step S7: Receive the main point cloud of the columnar component within the same monitoring period and the standard period result; perform sampling quantity balancing and point cloud merging on the main point cloud of the columnar component within the same monitoring period to generate a period representative column point cloud; perform non-main point processing on the period representative column point cloud and recalculate the center line and top and bottom centers to generate a single-period refinement result; generate a refinement quality gating result based on the point cloud retention status, effective slice status, height coverage status, and center line change status before and after refinement processing; and set the single-period refinement result to be prohibited from participating in multi-period comparisons. Step S8: Receive the standard cycle results corresponding to multiple monitoring cycles, select the reference cycle and comparison cycle, generate the positional and directional parameter changes during the cycle based on the standard cycle representative centerline, the top center of the standard cycle, and the bottom center of the standard cycle, and output the slice quality, observation quality, cycle fusion status, refinement quality gating status, and multi-cycle change results.

2. The method for processing multi-period point cloud monitoring data of columnar components according to claim 1, characterized in that, In step S2, the stable reference object includes a stable reference area, fixed control markers, and stable components that do not participate in monitoring. During coordinate transformation, the monitored columnar component and the change-sensitive area connected to the monitored columnar component are excluded from the registration reference range. The local Z direction of the fixed project local coordinate system is consistent with the main extension direction of the columnar component, and the local X direction and local Y direction are located in the cross-sectional plane of the columnar component. The same uniform bottom height, uniform top height, fixed slice thickness, slice number, and slice height range are used for all monitoring cycles.

3. The method for processing multi-period point cloud monitoring data of columnar components according to claim 2, characterized in that, In step S3, the cross-sectional plane of the columnar component is divided into multiple fixed-angle regions. A sampling quantity index is generated based on the ratio of the number of effective points in the slice to the statistical value of the number of effective slice points in the same observation. A sampling coverage index is generated based on the ratio of the number of angle regions with effective point clouds to the total number of fixed-angle regions. A sampling uniformity index is generated based on the distribution difference of the number of effective points in each fixed-angle region. A centerline spatial consistency index is generated based on the horizontal distance between the slice center and the predicted center at the same height as the initial centerline of the monitoring day. Finally, the slice joint quality weight is generated based on the sampling quantity index, the sampling coverage index, the sampling uniformity index, and the centerline spatial consistency index.

4. The method for processing multi-period point cloud monitoring data of columnar components according to claim 3, characterized in that, In step S4, weighted linear fitting is performed on the local X-direction coordinates and local Y-direction coordinates of multiple slice centers according to the slice joint quality weights to generate the monitoring day optimized centerline. After completing one weighted linear fitting, the horizontal residual, centerline spatial consistency index and slice joint quality weights of each slice center are recalculated according to the monitoring day optimized centerline, and weighted linear fitting is performed again. The iteration stops when the horizontal change of the top center and the change of the centerline direction parameter formed by two adjacent fittings both meet the convergence condition.

5. The method for processing multi-period point cloud monitoring data of columnar components according to claim 4, characterized in that, In steps S5 and S6, the effective slice quality is generated based on the joint quality statistics of the effective slices; the centerline fitting quality is generated based on the weighted residual of the optimized centerline on the monitoring day and the iterative convergence state; the point cloud registration quality is generated based on the registration residual of the stable reference object; and the effective height coverage rate is generated based on the ratio between the effective slice coverage height and the unified target height. The combined weight is the result of the combination of the fusion weight of the monitoring day and the corresponding slice joint quality weight. If an observation lacks an effective slice at the corresponding slice number, it will not participate in the fusion of the center of the slice with the same name in the corresponding period.

6. The method for processing multi-period point cloud monitoring data of columnar components according to claim 5, characterized in that, The standard period results include the standard period representative centerline, uniform bottom height, uniform top height, standard period bottom center, standard period top center, local X-direction component of the top relative to the bottom, local Y-direction component of the top relative to the bottom, period's own horizontal offset amplitude, period's own orientation angle, number of effective observations, number of effective period slices, effective height coverage, and fusion quality status; the standard period bottom center and the standard period top center are calculated based on the coordinates of the standard period representative centerline at the uniform bottom height and the uniform top height, respectively.

7. The method for processing multi-period point cloud monitoring data of columnar components according to claim 6, characterized in that, In step S7, the non-subject point processing includes identifying and processing the point cloud of column base components, the point cloud of column top connecting components, local additional point clusters, and isolated point clusters; the refined quality gating result is generated based on the point retention rate, effective slice retention rate, effective height coverage rate, cross-sectional sampling coverage change, top position change before and after refinement, and centerline direction parameter change; when the refined quality gating result meets the preset quality conditions, the single-cycle refined result is applied to the single-cycle analysis; when the refined quality gating result does not meet the preset quality conditions, the single-cycle refined result is discarded and the process reverts to the standard cycle result.

8. The method for processing multi-period point cloud monitoring data of columnar components according to claim 7, characterized in that, In step S8, the top position change, bottom position change, local X-direction axis component change, local Y-direction axis component change, cycle axis increment vector, cycle self-horizontal offset amplitude difference, and cycle self-directional angle difference are calculated only based on the standard cycle results corresponding to the reference cycle and the comparison cycle. The output also includes coordinate system identifier, length unit, fixed slice thickness, slice joint quality weight, overall quality index of monitoring day, fusion weight of monitoring day, number of valid observations supported by slices with the same name in the period, reason for reverting to refinement results, algorithm version, and processing parameter version.

Images