A method for identifying the connection position of underground pipes of different materials based on three-dimensional image reconstruction

Abstract

This invention discloses a method for identifying the connection location of underground dissimilar pipelines based on 3D image reconstruction, belonging to the field of location recognition technology. It addresses the problems of distorted spatial structure representation and insufficient accuracy in determining connection locations. By segmenting the underground pipeline, fusing ground-penetrating radar reflection data and electromagnetic detection results, abrupt gradient coefficients are calculated to mark potential material abrupt change segments. 3D scan data of the marked segments are retrieved, and combined with scan density and segment structure information, connection candidates are generated. Enhanced scanning is performed on the candidates, and the pit landing point is determined through spatial coarseness features. The existence of a connection interface is determined based on local point clouds, regional point clouds are extracted, and axis fitting is performed on the regional point clouds to obtain 3D misalignment data of the pipelines on both sides. The construction area is divided based on the soil obstacle density around the connection interface, and the connection center point is output based on misalignment constraints. This significantly improves construction accuracy and reduces the risk of blind excavation.

Description

Technical Field

[0001] This invention relates to the field of location recognition technology, and more specifically, to a method for identifying the location of underground dissimilar pipe connections based on three-dimensional image reconstruction. Background Technology

[0002] As the scale of urban underground space development continues to expand, various types of pipelines, such as water supply, gas, electricity, and communication, are distributed underground with highly intertwined, diverse materials, and complex structures. In particular, during the renovation of existing pipeline networks and the connection of new and old pipelines, the identification of the connection locations between pipelines of different materials has become an important technical aspect in the early stages of construction.

[0003] The existing technology has the following shortcomings: Currently, existing technologies mostly rely on single detection methods to locate underground pipelines, and lack a collaborative fusion processing mechanism between various detection data. They also lack the ability to perform phased closed-loop analysis based on three-dimensional structural information and multi-source perception results, as well as a refined discrimination mechanism for the spatial features of the connection interface. This results in problems such as discontinuous response, distorted spatial structure representation, and insufficient accuracy in determining the connection position in the connection area of ​​dissimilar materials pipelines, making it difficult to provide reliable excavation and positioning basis for construction. Therefore, a method for identifying the connection position of underground dissimilar materials pipelines based on three-dimensional image reconstruction is proposed.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for identifying the connection location of underground dissimilar pipelines based on three-dimensional image reconstruction. This method addresses the problems mentioned in the background art by employing a segment marking mechanism driven by multi-source detection data fusion and abrupt change features, a connection candidate extraction strategy combined with three-dimensional scanning density constraints, an adaptive determination method for pit landing points based on spatial roughness features, and a construction decision-making closed-loop processing mechanism oriented towards regional point clouds with axis fitting and misalignment constraint analysis.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for identifying the connection location of underground dissimilar pipes based on three-dimensional image reconstruction, comprising the following steps: Step S1: The underground heterogeneous pipeline is segmented, and the ground-penetrating radar reflection data and electromagnetic detection results of each segment are obtained. The abrupt change gradient coefficient is generated by combining the ground-penetrating radar reflection data and electromagnetic detection results, and the segments are marked according to the abrupt change gradient coefficient. Step S2: Access the 3D database to obtain the 3D scanning data of the marked section, construct the section structure information of the marked section based on the 3D scanning data, detect the scanning density of the marked section, and extract the docking candidate body in combination with the section structure information; Step S3: Perform encrypted scanning on the docking candidate and evaluate the spatial roughness features. Mark the pit landing point according to the spatial roughness features. After obtaining local point cloud data based on the pit landing point, determine whether to output the regional point cloud of the docking interface. Step S4: Fit the point cloud axis using the regional point cloud, detect the three-dimensional misalignment data of the point cloud axis, collect the soil obstacle density at the connection interface and divide the construction area, and output the connection center point by combining the construction area and the three-dimensional misalignment data.

[0007] In a preferred embodiment, in step S1, the area where the underground foreign material pipe is located is segmented, and the detection path is discretized according to a fixed spatial step size to obtain multiple continuous segments, each segment corresponding to a spatial interval. Underground heterogeneous material pipelines refer to underground transport structures formed by two or more materials and connected through interfaces. Ground-penetrating radar (GPR) equipment is used to obtain radar reflection data at the corresponding location. The GPR reflection data is the radar reflection intensity. Electromagnetic detection equipment is deployed to measure the response of underground conductive structures and obtain electromagnetic detection results, which are the electromagnetic induction response intensity.

[0008] In a preferred embodiment, in step S1, after obtaining the radar reflection intensity and electromagnetic induction response intensity of each segment, normalization processing is performed to obtain normalized radar reflection value and normalized electromagnetic response value. The absolute difference between the normalized radar reflection value and the absolute difference between the normalized electromagnetic response value are calculated between adjacent segments to obtain the change in radar reflection and the change in electromagnetic response. Based on the changes in radar reflection and electromagnetic response, a weighted summation method is used to construct the abrupt gradient coefficient. When the mutation gradient coefficient is greater than or equal to the preset mutation threshold, the corresponding segment is marked. When the mutation gradient coefficient is less than the preset mutation threshold, the corresponding segment will not be marked.

[0009] In a preferred embodiment, in step S2, a three-dimensional database is accessed, which is a structured data collection used to store three-dimensional scan data of underground space. Based on the spatial coordinate range corresponding to the marked section, the corresponding three-dimensional scanning data is extracted. The three-dimensional scanning data is represented by a point cloud set, which is a set of spatial coordinates of discrete sampling points in the underground space. The marked area is divided into different voxel units according to the preset voxel size, and each voxel unit corresponds to a spatial sub-region. The number of point clouds within each voxel is counted, the volume of the voxel is obtained by calculating the coordinate product of the preset voxel size, and the scan density is obtained by dividing the number of point clouds by the voxel volume.

[0010] In a preferred embodiment, in step S2, segment structure information is constructed within each voxel unit. The segment structure information includes the point cloud spatial centroid. The spatial coordinates of all point clouds within the voxel unit are summed and divided by the total number of point clouds within the voxel unit to obtain the point cloud spatial centroid of the voxel unit. After obtaining the centroid of the point cloud space, an offset vector is constructed for each point cloud relative to the centroid of the point cloud space, and a covariance matrix is ​​constructed based on all offset vectors. The covariance matrix is ​​decomposed into three eigenvalues, and a structural change index is constructed based on these eigenvalues. Obtain the minimum and maximum values ​​of the scan density within the marked region, normalize the scan density, and obtain the density missing coefficient. The structural change index is multiplied by the density missing coefficient to obtain the connection consistency judgment score; When the connection consistency judgment score is greater than or equal to the preset score threshold, the corresponding voxel unit is marked as a candidate voxel, and a three-dimensional connected component analysis is performed on the candidate voxels to merge spatially adjacent candidate voxels to form connection candidate bodies.

[0011] In a preferred embodiment, in step S3, an encrypted scan is performed on the docking candidate, using the three-dimensional bounding box of the docking candidate as the boundary and setting the point spacing to perform the encrypted scan and then collect an encrypted point cloud; Using each point in the encrypted point cloud as the center, a neighborhood point set is constructed by selecting encrypted point clouds within a preset neighborhood radius. The local covariance matrix is ​​calculated based on the neighborhood point set to obtain the normal vector dispersion. The mean of each normal vector dispersion is taken to obtain the mean of the global normal vector dispersion. The encrypted point cloud is projected onto the axis coordinate system corresponding to the pipeline axis direction vector. The vertical distance from each point to the axis is calculated as the radial elevation value, and the standard deviation of the radial elevation value is calculated to obtain the surface elevation standard deviation. Spatial roughness characteristics are calculated after standardizing the mean of global normal vector dispersion and the standard deviation of surface elevation. When the spatial roughness feature is greater than the preset spatial roughness threshold, the center coordinates of the three-dimensional bounding box of the docking candidate will be marked as the landing point of the crater. Conversely, it is not marked as the landing point of the crater.

[0012] In a preferred embodiment, in step S3, three-dimensional scanning data within a preset point cloud retrieval radius are retrieved with the crater landing point as the center to form local point cloud data; Determine whether the coverage length of the local point cloud data in the direction of the pipeline axis is greater than the preset axis coverage length threshold, and count whether the number of valid points in the cross section perpendicular to the vector of the pipeline axis is greater than the preset cross section point count threshold; When the conditions in the previous section are met simultaneously, the connection interface calculation process will begin. Multiple cutting planes are constructed along the axial direction with the pipeline axis direction vector as the normal vector direction, and cross-sectional point sets are extracted. After performing circular fitting on each cross-sectional point set, the cross-sectional fitting residual is calculated. The cross-section with the largest fitting residual is determined as the connection interface, and the corresponding regional point cloud of the connection interface is output.

[0013] In a preferred embodiment, in step S4, the regional point cloud is divided into a front-connection point cloud and a rear-connection point cloud, with the connection interface as the boundary. A cutting plane is constructed along the pipe axis with a preset cross-sectional spacing, and a set of cross-sectional points is extracted. Circular fitting is performed on each set of cross-sectional points to obtain the center coordinates. The center coordinates are then arranged in axial order to form a sequence of center coordinates. After decentering the circle center sequence, a covariance matrix is ​​constructed. The eigenvector corresponding to the largest eigenvalue of the covariance matrix is ​​selected as the axis direction vector, and the mean of the circle center sequence coordinates is used as the axis reference point to obtain the front axis and the back axis. After extending the front and rear axis lines to the interface, the coordinates of the intersection points of the axes are obtained respectively. The Euclidean distance between the coordinates of the intersection points of the front and rear axis lines is calculated to obtain the three-dimensional misalignment data.

[0014] In a preferred embodiment, in step S4, points with radial distances exceeding a preset multiple of the pipe's outer diameter are extracted from the regional point cloud based on the front and rear axis lines as references and are used as obstacle points to form an obstacle point set. The ratio of the total number of obstacle point points to the spatial volume occupied by the regional point cloud is used as the soil obstacle density. The construction area is obtained by comparing the soil obstacle density with the preset obstacle density classification threshold; the coordinates of the midpoint of the connection interface are obtained by calculating the average coordinates of the intersection points of the front and rear axes. The angle bisector direction vector is obtained by summing the components of the direction vectors of the front and rear axes and normalizing them. Starting from the coordinates of the midpoint of the interface, the initial connection center point is obtained by translating the vector along the angle bisector direction by half the displacement of the axis. When the initial connection center point is located inside the boundary of the construction area, the initial connection center point will be output as the connection center point. When the initial connection center point is located outside the boundary of the construction area, the initial connection center point is translated and corrected in the direction pointing to the geometric center of the construction area until it enters the boundary of the construction area and is then output as the connection center point.

[0015] The technical effects and advantages of this invention are as follows: This invention achieves a closed-loop process from detection to construction decision-making through the synergistic coupling of multi-source sensing and 3D reconstruction. It involves segmenting underground pipelines, fusing ground-penetrating radar reflection data and electromagnetic detection results, and calculating abrupt gradient coefficients to mark potential material abrupt change segments. Next, it retrieves 3D scan data of the marked segments, combining scan density and segment structure information to generate connection candidates. Subsequently, it performs intensified scanning on the candidates, determining the pit landing point through spatial coarseness features and judging the existence of a connection interface based on local point clouds, extracting regional point clouds. Further, it performs axis fitting on the regional point clouds to obtain 3D misalignment data of the pipelines on both sides. Finally, it divides the construction area based on the soil obstacle density around the connection interface, outputting the connection center point based on misalignment constraints. This achieves stable identification of connection positions under conditions of weak non-metallic response and structural irregularities, significantly improving construction accuracy and reducing the risk of blind excavation. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the implementation of a method for identifying the connection location of underground dissimilar pipes based on three-dimensional image reconstruction, according to the present invention.

[0017] Figure 2 This is a schematic diagram illustrating the steps of a method for identifying the connection location of underground dissimilar pipes based on three-dimensional image reconstruction according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention achieves a closed-loop process from detection to construction decision-making through the synergistic coupling of multi-source sensing and 3D reconstruction. It involves segmenting underground pipelines, fusing ground-penetrating radar reflection data and electromagnetic detection results, and calculating abrupt gradient coefficients to mark potential material abrupt change segments. Next, it retrieves 3D scan data of the marked segments, combining scan density and segment structure information to generate connection candidates. Subsequently, it performs intensified scanning on the candidates, determining the pit landing point through spatial coarseness features, and judging the existence of a connection interface based on local point clouds, extracting regional point clouds. Further, it performs axis fitting on the regional point clouds to obtain 3D misalignment data of the pipelines on both sides. Finally, it divides the construction area based on the soil obstacle density around the connection interface, outputting the connection center point based on misalignment constraints, thus achieving stable identification of the connection location under conditions of weak non-metallic response and structural irregularities.

[0020] Example 1, as Figures 1 to 2 As shown, a method for identifying the location of underground dissimilar pipe connections based on 3D image reconstruction includes the following steps: Step S1: The underground heterogeneous pipeline is segmented, and the ground-penetrating radar reflection data and electromagnetic detection results of each segment are obtained. The abrupt change gradient coefficient is generated by combining the ground-penetrating radar reflection data and electromagnetic detection results, and the segments are marked according to the abrupt change gradient coefficient. Step S2: Access the 3D database to obtain the 3D scanning data of the marked section, construct the section structure information of the marked section based on the 3D scanning data, detect the scanning density of the marked section, and extract the docking candidate body in combination with the section structure information; Step S3: Perform encrypted scanning on the docking candidate and evaluate the spatial roughness features. Mark the pit landing point according to the spatial roughness features. After obtaining local point cloud data based on the pit landing point, determine whether to output the regional point cloud of the docking interface. Step S4: Fit the point cloud axis using the regional point cloud, detect the three-dimensional misalignment data of the point cloud axis, collect the soil obstacle density at the connection interface and divide the construction area, and output the connection center point by combining the construction area and the three-dimensional misalignment data.

[0021] The specific implementation is as follows: In step S1, the area where the underground dissimilar pipe is located is subjected to segmentation and multi-source detection data fusion processing to generate abrupt gradient coefficients for marking potential connection sections.

[0022] Underground dissimilar material pipelines refer to underground transport structures composed of two or more materials connected by interfaces. These materials include combinations of media with different dielectric constants and conductivity properties, such as metals, polyethylene, and concrete. The physical properties of underground dissimilar material pipelines exhibit discontinuous changes in space. Changes in dielectric constant are reflected in the electromagnetic wave propagation speed and reflection intensity, while changes in conductivity are reflected in the electromagnetic induction response intensity. Their structural characteristics typically manifest as a geometrically continuous composite with abrupt changes in physical properties.

[0023] Specifically, the underground heterogeneous pipes are continuously scanned, and the detection path is discretized according to a fixed spatial step size to obtain multiple continuous segments, each segment corresponding to a spatial interval.

[0024] Within each section, ground-penetrating radar (GPR) equipment is used to acquire radar reflection data at corresponding locations. This GPR reflection data represents the radar reflection intensity. At each sampling location, the GPR equipment emits electromagnetic pulse signals into the ground. These signals are reflected when they encounter interfaces between different media during propagation, and the echo signals are recorded by the receiving antenna. Time-domain sampling is performed on the echo signals to obtain the reflection signal showing the change in echo amplitude over time. Based on the propagation speed of the echo signal in the underground medium, the time axis is converted to a depth axis, yielding the reflection intensity distribution corresponding to the depth. Furthermore, within the estimated pipeline burial depth range, corresponding depth intervals are selected, and amplitude integration is performed on the depth intervals to obtain the radar reflection intensity at the sampling location. The radar reflection intensity characterizes the degree of change in the underground dielectric constant at the sampling location; a higher value indicates the presence of a significant interface or material difference.

[0025] Secondly, electromagnetic detection equipment is deployed to measure the response of underground conductive structures, obtaining electromagnetic detection results, which are the electromagnetic induction response intensity. Specifically, an alternating magnetic field is applied underground through a transmitting coil, inducing a current in the underground conductor, and the secondary magnetic field response signal is collected by the receiving coil of the electromagnetic detection equipment. The amplitude of the secondary magnetic field response signal is demodulated to obtain the electromagnetic induction response intensity at the corresponding location. The electromagnetic induction response intensity characterizes the conductivity distribution of the underground medium; a higher value indicates stronger conductivity or the presence of a metallic structure.

[0026] It should be noted that ground-penetrating radar equipment is an underground detection device based on the principle of electromagnetic wave propagation and reflection. It obtains information about the underground medium structure by emitting high-frequency electromagnetic pulses and receiving underground reflected signals. Electromagnetic detection equipment is an underground detection device based on the principle of electromagnetic induction. It induces current in underground conductors by emitting alternating magnetic fields and obtains secondary magnetic field response signals by receiving coils.

[0027] To ensure spatial consistency between ground-penetrating radar reflection data and electromagnetic detection results, the two devices are synchronously positioned during data acquisition, the spatial coordinates of each sampling point are recorded, and a time synchronization mechanism is used to ensure that the ground-penetrating radar reflection data and electromagnetic detection results correspond to the same spatial location and the same sampling time.

[0028] After obtaining the radar reflection intensity and electromagnetic induction response intensity of each segment, normalization is performed to eliminate the influence of different dimensions, resulting in normalized radar reflection values ​​and normalized electromagnetic response values, specifically: ; in, This is the normalized radar reflectance value. This is the normalized electromagnetic response value. Radar reflection intensity, The electromagnetic induction response intensity, and These represent the maximum and minimum values ​​of the radar reflection intensity, respectively. and These represent the maximum and minimum values ​​of the electromagnetic induction response intensity, respectively. This is the segment index value.

[0029] The normalized radar reflection value and the normalized electromagnetic response value have a numerical range of [0,1]. The larger the value, the stronger the signal response of the corresponding segment.

[0030] Furthermore, the absolute difference between the normalized radar reflection value and the absolute difference between the normalized electromagnetic response value are calculated between adjacent segments to obtain the radar reflection change and electromagnetic response change. The radar reflection change reflects the magnitude of the change in electromagnetic reflection characteristics between adjacent segments; the larger the value, the more drastic the change in the medium interface. The electromagnetic response change reflects the degree of change in conductivity characteristics; the larger the value, the more obvious the change in material properties.

[0031] Based on the changes in radar reflection and electromagnetic response, a weighted summation method is used to construct the abrupt gradient coefficient, the specific expression of which is as follows: ; in, The mutation gradient coefficient, This represents the change in radar reflection. The change in electromagnetic response and These are preset weighting coefficients used to adjust the contribution ratio of radar reflection and electromagnetic response in sudden change detection.

[0032] The mutation gradient coefficient is used to comprehensively characterize the intensity of changes in physical properties between segments. The larger the value, the more likely there is a material mutation or structural change in that segment.

[0033] Finally, the mutation gradient coefficients corresponding to each segment are compared with the preset mutation thresholds. When the mutation gradient coefficient is greater than or equal to the preset mutation threshold, the corresponding segment is marked to indicate potential candidate areas for dissimilar material connection locations. When the mutation gradient coefficient is less than the preset mutation threshold, the corresponding segment will not be marked.

[0034] It should be noted that the preset mutation threshold is used to determine whether there are significant physical property changes in a segment, and its setting is based on the statistical distribution of the mutation gradient coefficient. Specifically, the mutation gradient coefficient sequence of all segments is statistically analyzed, its mean and standard deviation are calculated, and the preset mutation threshold is defined as the sum of the mean and twice the standard deviation.

[0035] Through the above processing, the abrupt change sections along the underground pipeline can be quantitatively marked, providing input basis for the subsequent extraction of connection candidates.

[0036] In step S2, for the marked section, three-dimensional data retrieval, structural modeling, and connection candidate extraction are performed.

[0037] Specifically, the process involves accessing a 3D database, which is a structured collection of data used to store 3D scanning data of underground spaces. Based on the spatial coordinate range corresponding to the marked section, the corresponding 3D scanning data is extracted. The 3D scanning data is represented as a point cloud set, which is a set of spatial coordinates of discrete sampling points in the underground space. It originates from laser scanning equipment and reflects the spatial geometric distribution characteristics of the underground structure. The more point clouds there are, the more thoroughly the marked section has been scanned.

[0038] After acquiring the point cloud set, voxelization is performed on the marked segments. Specifically, the marked segments are divided into different voxel units according to the preset voxel size, and each voxel unit corresponds to a spatial sub-region.

[0039] The number of point clouds is counted within each voxel unit. The volume of the voxel unit is obtained by calculating the product of the coordinates of the preset voxel size. The scanning density is obtained by dividing the number of point clouds by the volume of the voxel unit. The scanning density represents the sampling density of point clouds within a unit volume. The larger the value, the more complete the structural information of the voxel region. The smaller the value, the more sparse the data, reflecting the weak response characteristics of non-metallic or heterogeneous material regions.

[0040] Furthermore, segmental structure information is constructed within each voxel unit, including the spatial centroid of the point cloud. The spatial coordinates of all points within the voxel unit are summed and divided by the total number of points within the voxel unit to obtain the spatial centroid of the voxel unit. The spatial centroid of the point cloud represents the spatial center position of the structure within the voxel unit.

[0041] After obtaining the centroid of the point cloud space, an offset vector is constructed for each point cloud relative to the centroid. Based on all offset vectors, a covariance matrix is ​​constructed: ; in, Let covariance matrix be the variance matrix. The total number of point clouds within a voxel unit. Let be the spatial coordinates of the i-th point cloud within the voxel unit. Let k be the spatial region corresponding to the k-th voxel unit. For the spatial centroid of the point cloud, This is the transpose of a vector.

[0042] Eigenvalue decomposition of the covariance matrix yields three eigenvalues. (Sorted from largest to smallest), the three eigenvalues ​​originate from the inherent properties of the covariance matrix, representing the degree of dispersion of the point cloud along the three orthogonal principal directions. Among them, It represents the maximum extent of the point cloud along the main direction. The larger the value, the more obvious the directional extension structure in the region. It indicates the extent of expansion in the secondary principal direction, and is used to reflect the distribution of the structure in the secondary principal direction; It represents the degree of dispersion in the smallest direction. The smaller the value, the more concentrated the point cloud is in that direction, and the more likely it is to form a planar or interface structure.

[0043] It should be noted that eigenvalue decomposition is a mathematical method for spectral decomposition of the covariance matrix. It decomposes the covariance matrix into a combination of eigenvectors and eigenvalues. The eigenvalues ​​reflect the variance of the data in the direction of the corresponding eigenvector. In 3D point cloud processing, the three eigenvalues ​​correspond to the degree of dispersion of the point cloud in three orthogonal directions.

[0044] By observing the relationship between the magnitudes of the three characteristic values, local structural morphology can be quantitatively distinguished, thereby identifying continuous pipe sections, interface regions, or regions of structural abrupt changes.

[0045] Constructing a structural change index based on eigenvalues: ; in, This is a structural change index. and All of them are eigenvalues.

[0046] The structural change index is a dimensionless parameter that characterizes the degree of structural anisotropy. The larger the value, the more obvious the geometric structure change in the area, and the more likely there are pipe bends, interfaces, or dissimilar material connections.

[0047] To avoid inconsistencies caused by directly superimposing parameters with different dimensions, the scan density is normalized. The minimum and maximum scan density values ​​within the marked region are obtained, and the density missing coefficient is defined as: ; in, Density missing coefficient, For scan density, and These represent the minimum and maximum values ​​of the scanning density within the marked segment, respectively.

[0048] The density missing coefficient is a dimensionless parameter with a value range of [0,1]. It is used to characterize the sparsity of data. The larger the value, the lower the scanning density of the marked area and the more obvious the data missing.

[0049] Multiplying the structural change index by the density missing coefficient yields the connection consistency judgment score, which is a dimensionless parameter used to simultaneously constrain structural abruptness and data sparsity. The connection consistency judgment score only achieves a large value when both the structural change index and the density missing coefficient are significant. The larger the value, the more significant the voxel unit is, indicating that it simultaneously exhibits significant structural changes and has scanning holes.

[0050] The connection consistency judgment score of each voxel unit is compared with the preset scoring threshold. When the connection consistency judgment score is greater than or equal to the preset scoring threshold, the corresponding voxel unit is marked as a candidate voxel, and three-dimensional connected component analysis is performed on the candidate voxels to merge spatially adjacent candidate voxels to form connection candidate bodies.

[0051] It should be noted that 3D connected component analysis is a spatial clustering method for voxel units that meet specific conditions in a 3D discrete space. It groups interconnected voxels into the same connected region by judging their adjacency relationships (including face adjacency, edge adjacency, or vertex adjacency). A preset scoring threshold is used to determine whether a voxel unit belongs to a connection candidate, and its setting is based on the statistical characteristics of connection consistency scoring. Specifically, the scoring sequence of all voxel units is statistically analyzed, and its quantile value is calculated; for example, the p-th quantile is taken as the preset scoring threshold.

[0052] The above processing enables spatial extraction of potential connection areas within the marked segment.

[0053] In step S3, encrypted scanning is performed on each docking candidate. Encrypted scanning refers to taking the spatial range of the docking candidate as the scanning boundary and using a sampling interval higher than the three-dimensional scanning data density to re-collect three-dimensional point clouds of the pipeline body and surrounding strata within the spatial range, in order to obtain fine-grained geometric information of the candidate surface.

[0054] The spatial range of the docking candidate is represented by a three-dimensional bounding box. The three-dimensional bounding box is determined by the coordinate extreme values ​​of the docking candidate. The minimum and maximum values ​​in the three coordinate axes are used to form six boundary surfaces. The cuboid region enclosed by the six boundary surfaces is used as the spatial boundary of the encrypted scan. Based on the length of the three-dimensional bounding box of the docking candidate in the axial direction and the width in the radial direction, the spatial scale to be covered by the encrypted scan is determined, and the point spacing of the encrypted scan is set to half of the point spacing corresponding to the scan density in step S2. After completing the encrypted scan, the point cloud obtained from the encrypted scan is denoted as the encrypted point cloud. Each encrypted point cloud is designated as its center, and the encrypted point clouds within a preset neighborhood radius corresponding to the center are designated as neighboring points. These neighboring points are then combined into a neighboring point set. A local covariance matrix is ​​constructed based on this neighboring point set. ,in, The number of neighboring points within the neighborhood radius. Let i be the three-dimensional coordinate vector of the i-th neighboring point. It is the coordinate mean vector of the nearest point set; After performing eigenvalue decomposition on the local covariance matrix, three eigenvalues ​​are obtained. The smallest eigenvalue is divided by the sum of the three eigenvalues, and the ratio is taken as the normal vector dispersion at that point. ,in, Let be the three eigenvalues ​​of the covariance matrix, and d be the dispersion of the normal vector. The greater the dispersion of the normal vector, the more irregular the local surface where the center of the encrypted point cloud is located; the mean of the dispersion of the normal vector of the encrypted point cloud is taken to obtain the mean of the global normal vector dispersion.

[0055] The encrypted point cloud is projected onto the axis coordinate system corresponding to the direction vector of the pipeline axis. The axis coordinate system takes the reference point of the point cloud axis as the origin, the direction vector of the axis as the vertical axis, and the plane perpendicular to the axis direction as the cross section. The vertical distance from the encrypted point cloud to the axis is calculated, and the vertical distance is used as the radial elevation value of the encrypted point cloud. The radial elevation value reflects the degree to which the point deviates from the pipeline axis in the radial direction.

[0056] Among them, the pipeline axis direction vector is the unit direction vector of the spatial extension direction of the pipeline centerline within the section where the connection candidate is located; The standard deviation of the radial elevation values ​​for all points is calculated to obtain the standard deviation of the surface elevation. After standardizing the mean of the global normal vector dispersion and the standard deviation of the surface elevation, the global dispersion coefficient and the surface elevation coefficient are obtained respectively. Spatial roughness characteristics are calculated by combining the global dispersion coefficient and the surface elevation coefficient: ,in, and To preset the rough weighting coefficients, The global dispersion coefficient, This is the surface elevation coefficient. It is a spatial coarseness feature; Spatial roughness reflects the overall degree to which the surface of the docking candidate deviates from a smooth surface in the local neighborhood. The larger the value, the higher the degree of geometric irregularity of the docking candidate surface. It should be noted that the preset neighborhood radius can be set according to the multiple relationship of the spacing between encrypted scanning points. For example, 5 times the spacing between encrypted scanning points can be selected as the preset neighborhood radius. The preset roughness weight coefficient can be set according to the pipe material type. The standardization processing method includes, but is not limited to, standard linear transformation based on interval scaling, Z-Score standardization method based on statistics, or normalization method based on nonlinear mapping function. The application method of standardization processing will not be elaborated here.

[0057] Compare the spatial coarseness features with a preset spatial coarseness threshold: When the spatial roughness feature is greater than or equal to the preset spatial roughness threshold, it indicates that there are geometric irregularities on the surface of the docking candidate. Then, the center coordinates of the three-dimensional bounding box corresponding to the docking candidate are marked as the pit landing point. Conversely, this indicates the surface geometry of the docking candidate and does not mark the landing point of the probe.

[0058] The center coordinates of the three-dimensional bounding box are obtained by taking the average of the minimum and maximum values ​​of the three-dimensional bounding box in the X, Y, and Z coordinate axes respectively, and the center coordinates of the three-dimensional bounding box are used as the three-dimensional coordinates of the crater landing point. Access the 3D database, take the crater landing point as the center, extend the spatial range determined by the preset point cloud retrieval radius to obtain the local retrieval range, retrieve the 3D scan data corresponding to the local retrieval range as local point cloud data.

[0059] Determine whether the local point cloud data meets the following output conditions, which include the following two items: First, the coverage length of the local point cloud data in the direction of the pipeline axis is not less than the preset axis coverage length threshold. Secondly, taking the crater landing point as the center and the direction vector perpendicular to the pipeline axis as a cross section, extract the points in the local point cloud data that fall within the preset thickness range of the cross section. Points that fall within the thickness range of the cross section and whose radial distance does not exceed the outer diameter of the pipeline are recorded as valid points. The number of valid points is counted to obtain the number of valid points. The number of valid points is not less than the preset cross section point count threshold.

[0060] When the local point cloud data simultaneously meets both of the above output conditions, the connection interface calculation process is entered and the local point cloud data containing the connection interface is output as the regional point cloud. It should be noted that the preset cross-section spacing can be set according to the spacing of the encrypted scanning points to ensure that each cross-section point set contains a sufficient number of points to participate in the circular fitting; the preset cross-section thickness can be set according to the spacing of the encrypted scanning points, taking twice the spacing of the encrypted scanning points; the preset axis coverage length threshold can be set according to the minimum circle center sequence span required for the least squares fitting of the spatial straight line in step S4; the preset cross-section point number threshold is determined according to the minimum effective point number requirement required for the interface plane fitting; and the preset spatial roughness threshold is set according to the historical encrypted scanning data statistics of similar pipe connection areas and continuous pipe sections.

[0061] Using the pipe axis direction vector as the normal vector direction, cutting planes perpendicular to the axis are constructed sequentially along the axis direction at preset cross-sectional intervals. Points falling within the preset thickness range of each cutting plane in the local point cloud data are extracted to obtain the point set of each cross-section. For each cross-sectional point set, a circular fitting is performed. Circular fitting refers to finding an optimal circle within the plane containing the cross-sectional point set that minimizes the sum of the distances from each point in the cross-sectional point set to the circumference of the circle. The coordinates of the center and radius of the optimal circle are used as the fitting circle parameters for that cross-section. The distance from each point in the cross-sectional point set to the circumference of the fitting circle is calculated. The straight line length between the point and the nearest point on the circumference of the fitting circle is the distance from that point to the fitting circle. The average distance from all points in the cross-sectional point set to the fitting circle is taken as the fitting residual for that cross-section. The calculation formula is as follows: e ; Where m is the number of points in the cross-section point set. Let j be the coordinates of the j-th point in the cross-sectional plane. Let r be the coordinates of the center of the fitted circle, r be the radius of the fitted circle, and e be the fitting residual of the cross section; The fitting residuals of each section are arranged along the axial direction. The section with the largest fitting residual is taken as the connection interface. The local point cloud data of the section containing the connection interface and the preset range on both sides along the axial direction are taken as the regional point cloud of the connection interface.

[0062] It should be noted that the preset interface extension range can be set according to the estimated outer diameter of the pipe to ensure that the area point cloud completely covers the pipe body on both sides of the connection interface in the axial direction.

[0063] In step S4, the area point cloud is divided into the pre-connection point cloud and the post-connection point cloud along the pipeline axis, with the connection interface as the boundary. A cross section perpendicular to the axis direction is constructed for the point cloud of the front section and the point cloud of the rear section of the docking with a preset cross section spacing. A cutting plane perpendicular to the axis direction is constructed for the point cloud of the front section and the point cloud of the rear section of the docking with a preset cross section spacing. Points falling within a preset thickness range of each cutting plane are extracted to obtain the point set of each cross section. The point set is then fitted with a circle to obtain the center coordinates of each cross section. Among them, the axis direction vector of the segment corresponding to the current connection candidate body in the segment structure information output in step S2 is used as the axis direction; Connect the center coordinates of each cross-section sequentially to form a circle center sequence. Perform spatial line least squares fitting on this circle center sequence. Spatial line least squares fitting refers to finding a spatial line that minimizes the sum of the squared distances from each circle center coordinate in the circle center sequence to that line. Specifically, this is done by decentering the mean of the circle center sequence to construct a covariance matrix. The eigenvector corresponding to the largest eigenvalue of the covariance matrix is ​​taken as the line direction vector, and the mean of the circle center sequence coordinates is taken as the reference point of the line. ,in, It is a unit direction vector. Let be the covariance matrix of the circle center sequence after decentering. Specifically, it is a three-dimensional covariance matrix constructed by subtracting the mean of the circle center coordinates from the coordinates of each circle center in the sequence. d is the eigenvector corresponding to the largest eigenvalue of the covariance matrix. This represents taking the unit vector that maximizes the objective function value; The eigenvector corresponding to the largest eigenvalue of the covariance matrix is ​​taken as the direction vector of the line, and the mean of the coordinates of all the center points in the circle center sequence is taken as the reference point of the line. The direction vector of the line determines the orientation of the line in space, and the reference point of the line determines the position of the line in space. Together, they uniquely determine a spatial line. The above fitting process is performed on the point cloud of the front section and the point cloud of the rear section respectively to obtain the direction vector of the front axis, the reference point of the front axis, the direction vector of the rear axis and the reference point of the rear axis. The front axis is determined by the direction vector of the front axis and the reference point of the front axis, and the rear axis is determined by the direction vector of the rear axis and the reference point of the rear axis. The front axis and the rear axis together constitute the point cloud axis. Extend the front axis and the rear axis to the plane where the docking interface is located. The three-dimensional coordinates of the intersection points of the two axes and the docking interface are used as the intersection coordinates of the front axis and the rear axis, respectively. The Euclidean distance between the coordinates of the intersection of the front and rear axes is calculated to obtain three-dimensional misalignment data, which reflects the degree of spatial offset of the axis centers of the two pipes at the joint interface. The larger the value, the more obvious the lateral offset of the axes of the two pipes at the joint. Based on the front and rear axis lines, extract points in the regional point cloud whose radial distance exceeds a preset multiple of the pipe's outer diameter and record them as obstacle points. The combination of obstacle points forms an obstacle point set. The soil obstacle density is obtained by dividing the total number of points in the obstacle point set by the enclosing volume of the actual space occupied by the regional point cloud. Compare the soil obstacle density with a preset obstacle density classification threshold: When the soil obstacle density is less than the preset obstacle density classification threshold, it indicates that the obstacle at the connection interface is sparse. The horizontal projection range of the three-dimensional bounding box of the connection candidate on the ground surface is used as the boundary and is taken as the construction area. When the soil obstacle density is greater than or equal to the preset obstacle density classification threshold, it indicates that the obstacles at the connection interface are dense. The area occupied by the spatial convex hull of the obstacle point set is removed from the stratum space covered by the three-dimensional bounding box of the connection candidate. The area after expanding the remaining available space outward by the horizontal projection of the outer boundary of the ground surface with the preset avoidance margin is taken as the boundary and is used as the construction area. It should be explained that the preset obstacle density classification threshold is obtained based on historical survey data of the geological strata type in the construction area; the preset avoidance margin can be set according to the type of obstacle. Calculate the average coordinates of the intersection points of the front and rear axes to obtain the coordinates of the midpoint of the connection interface; add the corresponding components of the direction vectors of the front and rear axes to obtain the sum of the two vectors, and then divide the sum of the vectors by its own magnitude to obtain the angle bisector direction vector. Starting from the coordinates of the midpoint of the connection interface, the initial connection center point coordinates are obtained by translating half of the three-dimensional misalignment data along the direction of the angle bisector: ,in, The coordinates of the intersection point of the front axis are... The coordinates of the intersection point of the latter section of the axis. For three-dimensional misaligned data, The direction vector of the angle bisector. The coordinates of the initial connection center point; When the initial connection center point is located inside the boundary of the construction area, the coordinates of the initial connection center point will be directly output as the connection center point. When the initial connection center point is located outside the boundary of the construction area, the system moves from the initial connection center point in the direction pointing to the geometric center of the construction area until it enters the boundary of the construction area. Then, the coordinates of this position are output as the connection center point.

[0064] The connection center point is used to mark the docking position of the two pipes after the axis deviation is corrected within the construction area, providing a spatial positioning basis for the pipe alignment and connection operations.

[0065] Finally, it should be noted that in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0066] Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0067] In this document, the singular forms “a,” “an,” and “the” may also include the plural forms unless the context clearly indicates otherwise. It should also be understood that terms such as “comprising / including” or “having” specify the presence of the stated features, integrals, steps, operations, components, parts, or combinations thereof, but do not preclude the possibility of the presence or addition of one or more other features, integrals, steps, operations, components, parts, or combinations thereof. Meanwhile, the term “and / or” as used in this specification includes any and all combinations of the associated listed items.

[0068] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

[0069] The above description of the disclosed embodiments will enable those skilled in the art to make or use various modifications to these embodiments. It will be readily apparent to those skilled in the art that the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying the connection location of underground dissimilar pipes based on three-dimensional image reconstruction, characterized in that: Includes the following steps: Step S1: The underground heterogeneous pipeline is segmented, and the ground-penetrating radar reflection data and electromagnetic detection results of each segment are obtained. The abrupt change gradient coefficient is generated by combining the ground-penetrating radar reflection data and electromagnetic detection results, and the segments are marked according to the abrupt change gradient coefficient. Step S2: Access the 3D database to obtain the 3D scanning data of the marked section, construct the section structure information of the marked section based on the 3D scanning data, detect the scanning density of the marked section, and extract the docking candidate body in combination with the section structure information; Step S3: Perform encrypted scanning on the docking candidate and evaluate the spatial roughness features. Mark the pit landing point according to the spatial roughness features. After obtaining local point cloud data based on the pit landing point, determine whether to output the regional point cloud of the docking interface. Step S4: Fit the point cloud axis using the regional point cloud, detect the three-dimensional misalignment data of the point cloud axis, collect the soil obstacle density at the connection interface and divide the construction area, and output the connection center point by combining the construction area and the three-dimensional misalignment data.

2. The method for identifying the location of underground dissimilar pipe connections based on three-dimensional image reconstruction according to claim 1, characterized in that: In step S1, the area where the underground foreign material pipe is located is segmented, and the detection path is discretized according to a fixed spatial step size to obtain multiple continuous segments, each segment corresponding to a spatial interval; Underground heterogeneous material pipelines refer to underground transport structures formed by two or more materials and connected through interfaces. Ground-penetrating radar (GPR) equipment is used to obtain radar reflection data at the corresponding location. The GPR reflection data is the radar reflection intensity. Electromagnetic detection equipment is deployed to measure the response of underground conductive structures and obtain electromagnetic detection results, which are the electromagnetic induction response intensity.

3. The method for identifying the connection location of underground dissimilar pipes based on three-dimensional image reconstruction according to claim 2, characterized in that: In step S1, after obtaining the radar reflection intensity and electromagnetic induction response intensity of each section, normalization processing is performed to obtain normalized radar reflection value and normalized electromagnetic response value. The absolute difference between the normalized radar reflection value and the absolute difference between the normalized electromagnetic response value are calculated between adjacent segments to obtain the change in radar reflection and the change in electromagnetic response. Based on the changes in radar reflection and electromagnetic response, a weighted summation method is used to construct the abrupt gradient coefficient. When the mutation gradient coefficient is greater than or equal to the preset mutation threshold, the corresponding segment is marked. When the mutation gradient coefficient is less than the preset mutation threshold, the corresponding segment will not be marked.

4. The method for identifying the location of underground dissimilar pipe connections based on three-dimensional image reconstruction according to claim 1, characterized in that: In step S2, the three-dimensional database is accessed. The three-dimensional database is a structured data collection used to store three-dimensional scan data of underground space. Based on the spatial coordinate range corresponding to the marked section, the corresponding three-dimensional scanning data is extracted. The three-dimensional scanning data is represented by a point cloud set, which is a set of spatial coordinates of discrete sampling points in the underground space. The marked area is divided into different voxel units according to the preset voxel size, and each voxel unit corresponds to a spatial sub-region. The number of point clouds within each voxel is counted, the volume of the voxel is obtained by calculating the coordinate product of the preset voxel size, and the scan density is obtained by dividing the number of point clouds by the voxel volume.

5. The method for identifying the location of underground dissimilar pipe connections based on three-dimensional image reconstruction according to claim 4, characterized in that: In step S2, segment structure information is constructed within each voxel unit. The segment structure information includes the point cloud spatial centroid. The spatial coordinates of all point clouds within the voxel unit are summed and divided by the total number of point clouds within the voxel unit to obtain the point cloud spatial centroid of the voxel unit. After obtaining the centroid of the point cloud space, an offset vector is constructed for each point cloud relative to the centroid of the point cloud space, and a covariance matrix is ​​constructed based on all offset vectors. The covariance matrix is ​​decomposed into three eigenvalues, and a structural change index is constructed based on these eigenvalues. Obtain the minimum and maximum values ​​of the scan density within the marked region, normalize the scan density, and obtain the density missing coefficient. The structural change index is multiplied by the density missing coefficient to obtain the connection consistency judgment score; When the connection consistency judgment score is greater than or equal to the preset score threshold, the corresponding voxel unit is marked as a candidate voxel, and a three-dimensional connected component analysis is performed on the candidate voxels to merge spatially adjacent candidate voxels to form connection candidate bodies.

6. The method for identifying the location of underground dissimilar pipe connections based on three-dimensional image reconstruction according to claim 1, characterized in that: In step S3, an encrypted scan is performed on the docking candidate. The encrypted scan is performed with the three-dimensional bounding box of the docking candidate as the boundary and the point spacing is set, and then an encrypted point cloud is collected. Using each point in the encrypted point cloud as the center, a neighborhood point set is constructed by selecting encrypted point clouds within a preset neighborhood radius. The local covariance matrix is ​​calculated based on the neighborhood point set to obtain the normal vector dispersion. The mean of each normal vector dispersion is taken to obtain the mean of the global normal vector dispersion. The encrypted point cloud is projected onto the axis coordinate system corresponding to the pipeline axis direction vector. The vertical distance from each point to the axis is calculated as the radial elevation value, and the standard deviation of the radial elevation value is calculated to obtain the surface elevation standard deviation. Spatial roughness characteristics are calculated after standardizing the mean of global normal vector dispersion and the standard deviation of surface elevation. When the spatial roughness feature is greater than the preset spatial roughness threshold, the center coordinates of the three-dimensional bounding box of the docking candidate will be marked as the landing point of the crater. Conversely, it is not marked as the landing point of the crater.

7. The method for identifying the location of underground dissimilar pipe connections based on three-dimensional image reconstruction according to claim 6, characterized in that: In step S3, local point cloud data is formed by retrieving three-dimensional scanning data within the radius of the preset point cloud retrieval point, with the crater landing point as the center. Determine whether the coverage length of the local point cloud data in the direction of the pipeline axis is greater than the preset axis coverage length threshold, and count whether the number of valid points in the cross section perpendicular to the vector of the pipeline axis is greater than the preset cross section point count threshold; When the conditions in the previous section are met simultaneously, the connection interface calculation process will begin. Multiple cutting planes are constructed along the axial direction with the pipeline axis direction vector as the normal vector direction, and cross-sectional point sets are extracted. After performing circular fitting on each cross-sectional point set, the cross-sectional fitting residual is calculated. The cross-section with the largest fitting residual is determined as the connection interface, and the corresponding regional point cloud of the connection interface is output.

8. The method for identifying the connection location of underground dissimilar pipes based on three-dimensional image reconstruction according to claim 1, characterized in that: In step S4, the regional point cloud is divided into the front section point cloud and the rear section point cloud, with the connection interface as the boundary. A cutting plane is constructed along the pipe axis with a preset cross-sectional spacing, and a set of cross-sectional points is extracted. Circular fitting is performed on each set of cross-sectional points to obtain the center coordinates. The center coordinates are then arranged in axial order to form a sequence of center coordinates. After decentering the circle center sequence, a covariance matrix is ​​constructed. The eigenvector corresponding to the largest eigenvalue of the covariance matrix is ​​selected as the axis direction vector, and the mean of the circle center sequence coordinates is used as the axis reference point to obtain the front axis and the back axis. After extending the front and rear axis lines to the interface, the coordinates of the intersection points of the axes are obtained respectively. The Euclidean distance between the coordinates of the intersection points of the front and rear axis lines is calculated to obtain the three-dimensional misalignment data.

9. The method for identifying the location of underground dissimilar pipe connections based on three-dimensional image reconstruction according to claim 8, characterized in that: In step S4, based on the front and rear axis lines, points whose radial distance exceeds a preset multiple of the pipe outer diameter are extracted from the regional point cloud as obstacle points and form an obstacle point set. The ratio of the total number of obstacle point sets to the spatial volume occupied by the regional point cloud is used as the soil obstacle density. The construction area is obtained by comparing the soil obstacle density with the preset obstacle density classification threshold; the coordinates of the midpoint of the connection interface are obtained by calculating the average coordinates of the intersection points of the front and rear axes. The angle bisector direction vector is obtained by summing the components of the direction vectors of the front and rear axes and normalizing them. Starting from the coordinates of the midpoint of the interface, the initial connection center point is obtained by translating the vector along the angle bisector direction by half the displacement of the axis. When the initial connection center point is located inside the boundary of the construction area, the initial connection center point will be output as the connection center point. When the initial connection center point is located outside the boundary of the construction area, the initial connection center point is translated and corrected in the direction pointing to the geometric center of the construction area until it enters the boundary of the construction area and is then output as the connection center point.

Images