Lidar-based physical singulation modeling method and apparatus
By combining lidar with a physical unit model model based on multi-source data, the problem of accuracy and automation in the monitoring of ancient wooden structures was solved. This method enables precise location and classification of defects, improving the objectivity and efficiency of monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING JIANYUANAN FIRE PROTECTION TECHNOLOGY CO LTD
- Filing Date
- 2026-04-14
- Publication Date
- 2026-07-07
AI Technical Summary
Existing technologies cannot accurately distinguish between normal deformation and abnormal damage in the monitoring of ancient wooden structures, and lack multi-source data fusion and automatic classification mechanisms, resulting in strong subjectivity and difficulty in quantification in monitoring.
A physical unit modeling method based on lidar is adopted. By collecting and preprocessing point cloud data, calculating micro-deformation and deformation residuals, and combining environmental humidity and radial shrinkage coefficient, a disease index is constructed to realize the location of disease grid cells and automatic classification of region types.
It enables accurate identification and scientific assessment of defects in the wooden structures of ancient buildings, improves the accuracy, efficiency and objectivity of monitoring, provides quantitative decision support, and automates the entire process from data collection to defect identification.
Smart Images
Figure CN122347745A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of laser scanning technology, and in particular to a method and apparatus for modeling physical unit models based on lidar. Background Technology
[0002] Under the long-term influence of the natural environment, ancient wooden structures are prone to hidden diseases such as surface decay and internal hollowing. Traditional monitoring methods mostly rely on manual visual inspection or tapping and listening, which are highly subjective and difficult to quantify. In recent years, three-dimensional laser scanning technology has been gradually applied to the protection of architectural heritage. However, existing methods usually only focus on geometric deformation and do not consider the coupling effect of the wood's swelling and shrinking characteristics with environmental factors, resulting in the inability to accurately distinguish between normal deformation and abnormal diseases.
[0003] Furthermore, data from a single sensor is insufficient to comprehensively reflect changes in material properties and lacks an automatic classification mechanism for different types of defects. Therefore, there is an urgent need for an intelligent monitoring method that integrates multi-source data and multi-timescale analysis to achieve accurate identification and scientific assessment of defects in the wooden structures of ancient buildings. Summary of the Invention
[0004] The purpose of this invention is to provide a physical unit modeling method and apparatus based on lidar, so as to solve at least one of the problems existing in the prior art.
[0005] To achieve the above objectives, the present invention adopts the following technical solution:
[0006] A physical unit modeling method based on lidar includes:
[0007] Collect and preprocess the raw point cloud data of the wooden pillars, and extract the point cloud statistical features of the current scanning window;
[0008] Based on point cloud data from adjacent scanning windows, the micro-deformation of the wooden column surface is calculated to identify suspicious active mesh cells;
[0009] At the end of each monitoring cycle, the ambient humidity data and radial shrinkage coefficient are fused together, and the deformation residual of each grid cell is calculated.
[0010] At the end of each long monitoring cycle, the disease index is determined based on the reflection intensity of the grid cells, the identification results of suspected active grid cells, and the deformation residuals of the grid cells, and the diseased grid cells are located.
[0011] At the end of each long monitoring cycle, disease areas are generated based on the geometric distribution of disease grid cells, and disease area types are classified.
[0012] Furthermore, within each scanning window, point clouds of the wooden column surface are acquired, and the acquired raw point clouds are preprocessed: First, radius filtering is used to remove isolated noise points. The search radius is set to 3mm. If there are fewer than 5 points within this radius, they are judged as noise and removed. Second, the point cloud data scanned by the three devices are stitched together based on a pre-deployed common target to form a complete column surface point cloud dataset.
[0013] Each point cloud is mapped to a grid cell according to its spatial coordinates. All point clouds falling into each grid cell are recorded. The average number of point clouds in all grid cells of the current scanning window is calculated and denoted as navg. When the number of point clouds in a certain grid cell is less than 0.3×navg, the grid cell is marked as a "missing data cell" and will not be included in the calculation in subsequent analysis.
[0014] Furthermore, for each grid cell (i,j), first check whether it has not been marked as a "missing data cell" in both the current scan window and the previous scan window. If both are valid, then perform the following calculations:
[0015] Let the set of point clouds that the current scanning window falls into the grid cell be Pt = {p1, p2, ..., pK}, where each point pk contains three-dimensional coordinates (xk, yk, zk); calculate the average value (xkp, ykp, zkp) of the three-dimensional coordinates of the point cloud set in the grid cell, and use it as the centroid coordinate of the point cloud of the grid cell in the current scanning window;
[0016] Similarly, calculate the centroid coordinates (xkpv, ykpv, zkpv) of the point cloud set that fell into the grid cell in the previous scan cycle, and calculate the displacement vector (△x, △y, △z) of the grid cell, and then calculate the displacement magnitude d(i,j) of the grid cell.
[0017] Where, △x=xkp-xkpv, △y=ykp-ykpv, △z=zkp-zkpv, d(i,j)=sqrt(△x2+△y2+△z2).
[0018] Furthermore, for the displacement modulus of all valid mesh elements in the current scanning window, calculate the average value μd and the standard deviation σd. If the displacement modulus of a valid mesh element is greater than (μd + 2 × σd), then the valid mesh element is determined to be a suspected active mesh element; otherwise, the valid mesh element is determined to be a normal active mesh element.
[0019] Furthermore, at the end of each monitoring cycle, the ambient humidity data and radial shrinkage coefficient are fused, and the deformation residual of each grid cell is calculated:
[0020] Obtain the average relative humidity RHavg during the current monitoring period, and combine it with the radial shrinkage coefficient αr of the wooden column to determine the theoretical change in the radius of the wooden column ΔR, ΔR = R0 × αr × (RHavg - RHr);
[0021] Where R0 is the initial radius of the wooden column, and RHr is the preset humidity;
[0022] For each grid cell, project the coordinates of all point clouds within that grid cell onto a plane perpendicular to the axis of the wooden column, and calculate the average distance from each point to the axis, which is taken as the actual radius Ra of that grid cell.
[0023] The deformation residual Rc of the grid cell is calculated based on the theoretical change ΔR of the wooden column radius and Ra, where Rc = Ra - (R0 + ΔR).
[0024] Furthermore, the average value of the reflection intensity of each grid cell in each scanning window during the long monitoring period is calculated as the representative reflection intensity of the grid cell, denoted as Ice; the average value μI and standard deviation σI of the representative reflection intensities of all grid cells are calculated.
[0025] Mesh elements with Ice less than (μI-1.5×σI) are marked as low-intensity anomalous elements, and the reflection anomalous feature FT of the mesh elements is set to min(1,(μI-1.5×σI-Ice) / (3×σI)).
[0026] Mesh elements with Ice greater than or equal to (μI-1.5×σI) are marked as normal intensity elements, and the reflection anomaly FT of the mesh elements is set to 0.
[0027] Furthermore, for each grid cell, its disease index BH is defined as:
[0028] BH=w1×(Na / Nt)+w2×min(1,|Rc / Rt|)+w3×FT;
[0029] Where w1, w2 and w3 are weighting coefficients, w1+w2+w3=1, Rt is the residual threshold, Na is the number of scan windows in which the grid cell is marked as a suspicious active grid cell in the current long monitoring period, and Nt is the total number of scan time windows included in the current long monitoring period.
[0030] When the disease index BH of a grid cell is greater than the preset disease index threshold Dth, the grid cell is determined to be a diseased cell.
[0031] Furthermore, the disease grid cells are mapped onto a two-dimensional cylindrical unfolded plane, with the horizontal axis representing the circumferential index and the vertical axis representing the layer number. The four-neighbor connectivity criterion is used to mark the regions. All disease grid cells are traversed, and the disease grid cells that are connected to each other are merged into the same disease region. The k-th disease region is denoted as Regk.
[0032] Furthermore, for each affected area, its average deformation residual Rp and average reflection anomaly characteristic FTa are calculated to classify the affected area type:
[0033] If Rp is less than the first preset residual r1 and FTa is greater than the first preset anomaly threshold y1, the disease area type is classified as surface decay; if Rp is greater than the second preset residual r2 and FTa is less than the second preset anomaly threshold y2, the disease area type is classified as internal hollowing; if the above conditions are not met, the disease area type is classified as composite anomaly.
[0034] According to another aspect of this application, a physical unit modeling device based on lidar is provided, comprising:
[0035] The preprocessing unit is used to collect and preprocess the raw point cloud data of the wooden pillars and extract the point cloud statistical features of the current scanning window;
[0036] The identification unit is used to calculate the micro-deformation of the wooden column surface based on the point cloud data of adjacent scanning windows in order to identify suspicious active mesh cells;
[0037] The deformation analysis unit is used to fuse environmental humidity data and radial shrinkage coefficient at the end of each monitoring cycle and calculate the deformation residual of each grid cell.
[0038] Disease location unit is used to determine the disease index and locate the disease grid unit at the end of each long monitoring cycle based on the reflection intensity of the grid unit, the identification results of the suspected active grid unit, and the deformation residual of the grid unit.
[0039] The type division unit is used to generate disease areas and classify disease area types based on the geometric distribution of disease grid cells at the end of each long monitoring cycle.
[0040] The beneficial effects of this invention are as follows: This invention proposes a method for monitoring defects in ancient wooden structures based on lidar and multi-source data fusion. By establishing a multi-timescale monitoring strategy and a refined cylindrical mesh system, it achieves simultaneous tracking of micro-deformation, environmental response, and reflection characteristics of the wood surface. The method innovatively introduces the radial shrinkage coefficient to calculate theoretical deformation, thereby separating the influence of environmental humidity and highlighting abnormal residuals caused by defects. Simultaneously, it integrates reflection intensity characteristics and displacement activity to construct a defect index, achieving accurate location of surface decay and internal hollowing. Based on this, it generates defect areas based on connected component analysis and automatically classifies defect types, providing intuitive and quantitative decision support for the protection of ancient buildings. Compared with traditional manual inspection, this method significantly improves the accuracy, efficiency, and objectivity of monitoring, achieving full automation from data acquisition to defect identification, and has significant engineering application value. Attached Figure Description
[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.
[0042] Figure 1 This is a flowchart illustrating the physical unit modeling method based on lidar in this embodiment.
[0043] Figure 2 This is a flowchart illustrating the method for locating diseased grid cells in this embodiment.
[0044] Figure 3 This is a flowchart illustrating the method for classifying disease area types in this embodiment.
[0045] Figure 4 This is a schematic diagram of the physical unit modeling device based on lidar in this embodiment. Detailed Implementation
[0046] To more clearly illustrate the present invention, the following description, in conjunction with preferred embodiments and accompanying drawings, further explains the invention. Similar components in the drawings are indicated by the same reference numerals. Those skilled in the art should understand that the specific description below is illustrative rather than restrictive and should not be construed as limiting the scope of protection of the present invention.
[0047] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0048] Specifically, this embodiment is applied to the monitoring of surface decay and internal hollowing defects in ancient wooden structures.
[0049] Please see Figure 1 As shown, it is a flowchart of the physical unit modeling method based on lidar in this embodiment. The data relied upon by this method are all collected by a sensor system deployed around the wooden pillar.
[0050] Specifically, three ground-based lidar units are deployed at a 120-degree angle at a distance of 50cm from the surface of the wooden column to acquire three-dimensional point cloud data of the column. Temperature and humidity sensors are deployed at typical locations at the base and shaft of the column to collect data on the surface temperature of the wood and the relative humidity of the environment. Furthermore, the historical repair records of the wooden column are obtained, which record the wood species and radial shrinkage coefficient. The radial shrinkage coefficient is obtained by querying a wood science database or by measuring samples from the same batch of wood according to the national standard GB / T1932-2009 "Method for Determination of Wood Shrinkage", and the unit is mm / (mm·%RH). This embodiment does not specifically limit the specific model of the sensors or the data acquisition method; those skilled in the art can freely set them according to the site conditions.
[0051] This method employs a multi-timescale monitoring strategy, defining the following basic time units:
[0052] Scanning window: The duration of a single LiDAR scan, which is set to 5 minutes in this embodiment;
[0053] Short monitoring cycle: The time interval between two adjacent scanning windows, which is set to 10 minutes in this embodiment;
[0054] Monitoring cycle: The cycle used to statistically analyze environmental changes and surface deformation; in this embodiment, it is set to 24 hours.
[0055] Long monitoring period: The period used to assess the development trend of the disease, which is set to 30 days in this embodiment;
[0056] During system initialization, a unified spatial grid system is established based on the point cloud data of the wooden pillars: the surface of the wooden pillars is divided into layers at 2cm intervals along the height direction, and each layer is divided into grids at 2cm arc lengths along the circumference direction to form a uniform cylindrical grid. Each grid unit has a unique coordinate identifier (layer number, circumferential sequence number), which serves as the basic spatial unit for all subsequent analyses.
[0057] The method includes:
[0058] Step S1: Collect and preprocess the raw point cloud data of the wooden pillars, and extract the point cloud statistical features of the current scanning window.
[0059] Specifically, within each scanning window, point clouds of the wooden column surface are acquired, and the acquired raw point clouds are preprocessed: First, radius filtering is used to remove isolated noise points. The search radius is set to 3mm. If there are fewer than 5 points within this radius, they are judged as noise and removed. Second, the scanned point cloud data are stitched together based on pre-deployed common targets to form a complete column surface point cloud dataset.
[0060] Each point cloud is mapped to a grid cell according to its spatial coordinates. All point clouds falling into each grid cell are recorded. The average number of point clouds in all grid cells of the current scanning window is calculated and denoted as navg. When the number of point clouds in a certain grid cell is less than 0.3×navg, the grid cell is marked as a "missing data cell" and will not be included in the calculation in subsequent analysis.
[0061] Specifically, within each scanning window, the acquired and preprocessed point cloud data is mapped to a pre-established unified spatial grid system. This grid system is based on the surface of the wooden column, divided into layers every 2 cm along the height direction and one grid every 2 cm arc length along the circumference direction. Each grid cell has a unique coordinate identifier (layer number, circumferential sequence number). During mapping, the layer number and circumferential sequence number of each point are calculated based on its three-dimensional spatial coordinates: first, the layer to which the point belongs is determined based on the height coordinate (z value), and then its position in the circumferential direction is determined by the azimuth angle of the point (calculated from the x and y coordinates), thereby obtaining the corresponding grid cell identifier. All point clouds within the current scanning window are traversed, each point is assigned to the corresponding grid cell, and the number of point clouds falling into each grid cell is counted, denoted as n(i,j), where i is the layer number and j is the circumferential sequence number.
[0062] Specifically, this step involves collecting and preprocessing point cloud data for each scanning window, removing isolated noise points through radius filtering, and stitching together point clouds from multiple devices based on a common target to ensure the purity and integrity of the original data. At the same time, the point cloud is mapped to a pre-established grid system and missing data cells are marked, providing a reliable data foundation for subsequent analysis and effectively avoiding calculation biases caused by local data sparsity.
[0063] Please continue reading. Figure 1 As shown, the physical unit modeling method based on lidar also includes:
[0064] Step S2: Based on the point cloud data of adjacent scanning windows, calculate the micro-deformation of the wooden column surface to identify suspicious active mesh cells.
[0065] Specifically, for each grid cell (i,j), first check whether it has not been marked as a "missing data cell" in both the current scan window and the previous scan window. If both are valid, then perform the following calculations:
[0066] Let the set of point clouds that the current scanning window falls into the grid cell be Pt = {p1, p2, ..., pK}, where each point pk contains three-dimensional coordinates (xk, yk, zk); calculate the average value (xkp, ykp, zkp) of the three-dimensional coordinates of the point cloud set in the grid cell, and use it as the centroid coordinate of the point cloud of the grid cell in the current scanning window;
[0067] Similarly, calculate the centroid coordinates (xkpv, ykpv, zkpv) of the point cloud set that fell into the grid cell in the previous scan cycle, and calculate the displacement vector (△x, △y, △z) of the grid cell, and then calculate the displacement magnitude d(i,j) of the grid cell.
[0068] Where △x = xkp - xkpv, △y = ykp - ykpv, △z = zkp - zkpv, d(i,j) = sqrt(△x 2 +△y 2 +△z 2 );
[0069] For the displacement modulus of all valid mesh elements in the current scan window, calculate the average value μd and the standard deviation σd. If the displacement modulus of a valid mesh element is greater than (μd + 2 × σd), then the valid mesh element is determined to be a suspected active mesh element; otherwise, the valid mesh element is determined to be a normal active mesh element.
[0070] Specifically, this step calculates micro-deformation and identifies suspicious active grid cells by comparing the centroid displacement of the point cloud of the same grid cell in adjacent scanning windows. This allows for real-time capture of subtle changes on the wood surface, early detection of potential disease activity areas, and provides a quantitative basis for dynamic monitoring and early warning.
[0071] Please continue reading. Figure 1 As shown, the physical unit modeling method based on lidar also includes:
[0072] Step S3: At the end of each monitoring cycle, the ambient humidity data and radial shrinkage coefficient are fused, and the deformation residual of each grid cell is calculated.
[0073] Specifically, the average relative humidity RHavg during the current monitoring period is obtained, and combined with the radial shrinkage coefficient αr of the wooden column, the theoretical change in the radius of the wooden column ΔR is determined, where ΔR = R0 × αr × (RHavg - RHr).
[0074] Where R0 is the initial radius of the wooden column, and RHr is the preset humidity;
[0075] For each grid cell, project the coordinates of all point clouds within that grid cell onto a plane perpendicular to the axis of the wooden column, and calculate the average distance from each point to the axis, which is taken as the actual radius Ra of that grid cell.
[0076] The deformation residual Rc of the grid cell is calculated based on the theoretical change ΔR of the wooden column radius and Ra, where Rc = Ra - (R0 + ΔR).
[0077] Specifically, in this embodiment, the preset humidity is 60%RH.
[0078] Specifically, this step integrates environmental humidity data with the radial shrinkage coefficient of wood to calculate the deformation residual of each grid cell after removing the effects of normal moisture expansion and contraction. This highlights the inelastic deformation caused by disease, effectively separates environmental factors from disease effects, and significantly improves the accuracy of disease identification.
[0079] Please continue reading. Figure 1 As shown, the physical unit modeling method based on lidar also includes:
[0080] Step S4: At the end of each long monitoring cycle, the disease index is determined based on the reflection intensity of the grid cells, the identification results of suspected active grid cells, and the deformation residual of the grid cells, and the diseased grid cells are located.
[0081] Please see Figure 2 As shown, the method for locating the diseased grid cells includes:
[0082] Step S41: At the end of each long monitoring cycle, determine the reflection anomaly characteristics of the grid cells based on the reflection intensity of each scanning window grid cell.
[0083] Specifically, the average reflection intensity of each grid cell in each scanning window during the long monitoring period is calculated as the representative reflection intensity of the grid cell, denoted as Ice; the average value μI and standard deviation σI of the representative reflection intensities of all grid cells are calculated.
[0084] Mesh elements with Ice less than (μI-1.5×σI) are marked as low-intensity anomalous elements, and the reflection anomalous feature FT of the mesh elements is set to min(1,(μI-1.5×σI-Ice) / (3×σI)).
[0085] Mesh elements with Ice greater than or equal to (μI-1.5×σI) are marked as normal intensity elements, and the reflection anomaly FT of the mesh elements is set to 0.
[0086] Specifically, at the end of each long monitoring cycle, this step calculates reflection anomaly characteristics based on the statistical characteristics of reflection intensity of each scanning window. By marking low-intensity anomaly units, it can sensitively reflect the material changes on the wood surface caused by diseases such as decay, providing key spectral information for disease type identification.
[0087] Please continue reading. Figure 2 As shown, the method for locating the diseased grid cells further includes:
[0088] Step S42: Determine the disease index based on the reflection anomaly characteristics of the grid cells, the identification results of suspicious active grid cells, and the deformation residuals of the grid cells, and locate the diseased grid cells.
[0089] Specifically, for each grid cell, its disease index BH is defined as:
[0090] BH=w1×(Na / Nt)+w2×min(1,|Rc / Rt|)+w3×FT;
[0091] Where w1, w2 and w3 are weighting coefficients, w1+w2+w3=1, Rt is the residual threshold, Na is the number of scan windows in which the grid cell is marked as a suspicious active grid cell in the current long monitoring period, and Nt is the total number of scan time windows included in the current long monitoring period.
[0092] When the disease index BH of a grid cell is greater than the preset disease index threshold Dth, the grid cell is determined to be a diseased cell.
[0093] Specifically, in this embodiment, w1 is 0.3, w2 is 0.4, w3 is 0.3, the preset disease index threshold is 0.5, and the residual threshold is 2mm.
[0094] Specifically, this step integrates abnormal reflection characteristics, suspicious activity frequency, and deformation residuals to construct a disease index. By weighted fusion of multi-dimensional indicators and comparison with preset thresholds, quantitative positioning of disease grid units is achieved, overcoming the limitation of easy misjudgment by a single indicator and significantly enhancing the robustness of disease detection.
[0095] Please continue reading. Figure 1As shown, the physical unit modeling method based on lidar also includes:
[0096] Step S5: At the end of each long monitoring cycle, based on the geometric distribution of the disease grid cells, generate disease areas and classify disease area types.
[0097] Please see Figure 3 As shown, the method for classifying diseased areas includes:
[0098] Step S51: At the end of each long monitoring cycle, generate the disease area based on the geometric distribution of the disease grid cells.
[0099] Specifically, the disease grid cells are mapped onto a two-dimensional cylindrical unfolded plane, with the horizontal axis representing the circumferential index and the vertical axis representing the layer number. The four-neighbor connectivity criterion is used to mark the regions. All disease grid cells are traversed, and the disease grid cells that are connected to each other are merged into the same disease region. The k-th disease region is denoted as Regk.
[0100] Specifically, this step maps the disease grid cells onto a two-dimensional cylindrical unfolded plane and uses the four-neighbor connectivity criterion for region labeling. This enables the discrete disease cells to be clustered into continuous disease regions with practical significance, visually displaying the spatial distribution of the disease and facilitating subsequent assessment and remediation planning.
[0101] Please continue reading. Figure 3 As shown, the method for classifying diseased areas includes:
[0102] Step S52: Extract the regional characteristics of the diseased area and classify the diseased area types.
[0103] Specifically, for each affected area, its average deformation residual Rp and average reflection anomaly characteristic FTa are calculated to classify the affected area type:
[0104] If Rp is less than the first preset residual r1 and FTa is greater than the first preset anomaly threshold y1, the disease area type is classified as surface decay; if Rp is greater than the second preset residual r2 and FTa is less than the second preset anomaly threshold y2, the disease area type is classified as internal hollowing; if the above conditions are not met, the disease area type is classified as composite anomaly.
[0105] Specifically, in this embodiment, the first preset residual is -0.5mm, the second preset residual is 0.5mm, the first preset abnormal threshold is 0.3, and the second preset abnormal threshold is 0.2.
[0106] Specifically, this step extracts the average deformation residual and average reflection anomaly characteristics of the diseased area, and automatically classifies the surface decay, internal hollowing, or combined anomaly types according to preset thresholds. This achieves intelligent identification of the nature of the disease, provides a scientific basis for formulating targeted protection measures, and greatly improves the automation level of the monitoring system.
[0107] Please see Figure 4 As shown, the physical unit modeling device based on lidar includes:
[0108] The preprocessing unit is used to collect and preprocess the raw point cloud data of the wooden pillars and extract the point cloud statistical features of the current scanning window;
[0109] The identification unit is used to calculate the micro-deformation of the wooden column surface based on the point cloud data of adjacent scanning windows in order to identify suspicious active mesh cells;
[0110] The deformation analysis unit is used to fuse environmental humidity data and radial shrinkage coefficient at the end of each monitoring cycle and calculate the deformation residual of each grid cell.
[0111] Disease location unit is used to determine the disease index and locate the disease grid unit at the end of each long monitoring cycle based on the reflection intensity of the grid unit, the identification results of the suspected active grid unit, and the deformation residual of the grid unit.
[0112] The type division unit is used to generate disease areas and classify disease area types based on the geometric distribution of disease grid cells at the end of each long monitoring cycle.
[0113] The physical unit modeling device based on lidar provided in this application can execute the physical unit modeling method based on lidar provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects of the execution method.
[0114] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. For those skilled in the art, other variations or modifications can be made based on the above description. It is impossible to exhaustively list all the implementation methods here. All obvious variations or modifications derived from the technical solutions of the present invention are still within the protection scope of the present invention.
Claims
1. A method for modeling a physical unit based on lidar, characterized in that, include: Collect and preprocess the raw point cloud data of the wooden pillars, and extract the point cloud statistical features of the current scanning window; Based on point cloud data from adjacent scanning windows, the micro-deformation of the wooden column surface is calculated to identify suspicious active mesh cells; At the end of each monitoring cycle, the ambient humidity data and radial shrinkage coefficient are fused together, and the deformation residual of each grid cell is calculated. At the end of each long monitoring cycle, the disease index is determined based on the reflection intensity of the grid cells, the identification results of suspected active grid cells, and the deformation residuals of the grid cells, and the diseased grid cells are located. At the end of each long monitoring cycle, disease areas are generated based on the geometric distribution of disease grid cells, and disease area types are classified.
2. The physical unit modeling method based on lidar according to claim 1, characterized in that, Within each scanning window, the point cloud of the wooden column surface is acquired, and the acquired raw point cloud is preprocessed: First, radius filtering is used to remove isolated noise points. The search radius is set to 3mm. If there are fewer than 5 points within this radius, they are judged as noise and removed. Secondly, the point cloud data scanned by the three devices are stitched together based on a pre-deployed common target to form a complete cylindrical point cloud dataset. Each point cloud is mapped to each grid cell according to its spatial coordinates. All point clouds falling into each grid cell are recorded. The average number of point clouds in all grid cells of the current scanning window is calculated and denoted as navg. When the number of point clouds in a certain grid cell is less than 0.3×navg, the grid cell is marked as a "missing data cell" and will not be included in the calculation in subsequent analysis.
3. The physical unit modeling method based on lidar according to claim 2, characterized in that, For each grid cell (i,j), first check whether it has not been marked as a "missing data cell" in both the current scan window and the previous scan window. If both are valid, then perform the following calculations: Let the set of point clouds that the current scanning window falls into the grid cell be Pt = {p1, p2, ..., pK}, where each point pk contains three-dimensional coordinates (xk, yk, zk); calculate the average value (xkp, ykp, zkp) of the three-dimensional coordinates of the point cloud set in the grid cell, and use it as the centroid coordinate of the point cloud of the grid cell in the current scanning window; Similarly, calculate the centroid coordinates (xkpv, ykpv, zkpv) of the point cloud set that fell into the grid cell in the previous scan cycle, and calculate the displacement vector (△x, △y, △z) of the grid cell, and then calculate the displacement magnitude d(i,j) of the grid cell. Where △x = xkp - xkpv, △y = ykp - ykpv, △z = zkp - zkpv, d(i,j) = sqrt(△x 2 +△y 2 +△z 2 ).
4. The physical unit modeling method based on lidar according to claim 3, characterized in that, For the displacement modulus of all valid mesh elements in the current scan window, calculate the average value μd and the standard deviation σd. If the displacement modulus of a valid mesh element is greater than (μd + 2 × σd), then the valid mesh element is determined to be a suspected active mesh element; otherwise, the valid mesh element is determined to be a normal active mesh element.
5. The physical unit modeling method based on lidar according to claim 4, characterized in that, At the end of each monitoring cycle, the ambient humidity data and radial shrinkage coefficient are fused, and the deformation residual of each grid cell is calculated: Obtain the average relative humidity RHavg during the current monitoring period, and combine it with the radial shrinkage coefficient αr of the wooden column to determine the theoretical change in the radius of the wooden column ΔR, ΔR = R0 × αr × (RHavg - RHr); Where R0 is the initial radius of the wooden column, and RHr is the preset humidity; For each grid cell, project the coordinates of all point clouds within that grid cell onto a plane perpendicular to the axis of the wooden column, and calculate the average distance from each point to the axis, which is taken as the actual radius Ra of that grid cell. The deformation residual Rc of the grid cell is calculated based on the theoretical change ΔR of the wooden column radius and Ra, where Rc = Ra - (R0 + ΔR).
6. The physical unit modeling method based on lidar according to claim 5, characterized in that, The average reflection intensity of each grid cell in each scanning window during the long monitoring period is calculated as the representative reflection intensity of the grid cell, denoted as Ice; the average value μI and standard deviation σI of the representative reflection intensities of all grid cells are calculated. Mesh elements with Ice less than (μI-1.5×σI) are marked as low-intensity anomalous elements, and the reflection anomalous feature FT of the mesh elements is set to min(1,(μI-1.5×σI-Ice) / (3×σI)). Mesh elements with Ice greater than or equal to (μI-1.5×σI) are marked as normal intensity elements, and the reflection anomaly FT of the mesh elements is set to 0.
7. The physical unit modeling method based on lidar according to claim 6, characterized in that, For each grid cell, its disease index BH is defined as: BH=w1×(Na / Nt)+w2×min(1,|Rc / Rt|)+w3×FT; Where w1, w2 and w3 are weighting coefficients, w1+w2+w3=1, Rt is the residual threshold, Na is the number of scan windows in which the grid cell is marked as a suspicious active grid cell in the current long monitoring period, and Nt is the total number of scan time windows included in the current long monitoring period. When the disease index BH of a grid cell is greater than the preset disease index threshold Dth, the grid cell is determined to be a diseased cell.
8. The physical unit modeling method based on lidar according to claim 7, characterized in that, The disease grid cells are mapped onto a two-dimensional cylindrical unfolded plane, with the horizontal axis representing the circumferential index and the vertical axis representing the layer number. The four-neighbor connectivity criterion is used to mark the regions. All disease grid cells are traversed, and the disease grid cells that are connected to each other are merged into the same disease region. The k-th disease region is denoted as Regk.
9. The physical unit modeling method based on lidar according to claim 8, characterized in that, For each affected area, calculate its average deformation residual Rp and average reflection anomaly characteristic FTa, and then classify the affected area type: If Rp is less than the first preset residual r1 and FTa is greater than the first preset anomaly threshold y1, the disease area type is classified as surface decay; if Rp is greater than the second preset residual r2 and FTa is less than the second preset anomaly threshold y2, the disease area type is classified as internal hollowing; if the above conditions are not met, the disease area type is classified as composite anomaly.
10. A physical unit modeling device based on lidar, applied to the physical unit modeling method based on lidar as described in any one of claims 1-9, characterized in that, include: The preprocessing unit is used to collect and preprocess the raw point cloud data of the wooden pillars and extract the point cloud statistical features of the current scanning window. The identification unit is used to calculate the micro-deformation of the wooden column surface based on the point cloud data of adjacent scanning windows in order to identify suspicious active mesh cells; The deformation analysis unit is used to fuse environmental humidity data and radial shrinkage coefficient at the end of each monitoring cycle and calculate the deformation residual of each grid cell. Disease location unit is used to determine the disease index and locate the disease grid unit at the end of each long monitoring cycle based on the reflection intensity of the grid unit, the identification results of the suspected active grid unit, and the deformation residual of the grid unit. The type division unit is used to generate disease areas and classify disease area types based on the geometric distribution of disease grid cells at the end of each long monitoring cycle.