Live-line work three-dimensional modeling safety distance detection method, system, equipment and medium

Abstract

The invention relates to a live-line work three-dimensional modeling safety distance detection method, system and device and a medium. The method comprises the steps of obtaining a time sequence point cloud frame obtained through multi-angle scanning and a corresponding holder angle; projecting each time sequence point cloud frame to a unit spherical surface, updating grid occupation based on a unit spherical surface grid so as to calculate an accumulated view field coverage rate, and determining a target point cloud frame set according to the accumulated view field coverage rate; generating an initial pose transformation matrix for the target point cloud frame set based on the holder angle, and performing initial pose transformation processing on the target point cloud frame set according to the initial pose transformation matrix to obtain an initial aligned point cloud; performing point cloud registration processing on the initial alignment point cloud to obtain a registration transformation matrix, and transforming the initial alignment point cloud based on the registration transformation matrix to obtain a registration point cloud; and fusing the registered point clouds to generate a three-dimensional environment model, and calculating the nearest distance of the lead based on the three-dimensional environment model to obtain a safety distance detection result. The method has the effect of improving the detection reliability.

Description

Technical Field

[0001] This invention belongs to the technical field of three-dimensional environmental perception, and in particular relates to a method, system, equipment and medium for detecting safe distances in three-dimensional modeling of live-line work. Background Technology

[0002] Currently, live-line working scenarios are subject to objective constraints such as dense distribution of slender targets like wires, limited working space, and frequent obstructions and changes in perspective. During the operation, it is necessary to continuously obtain spatial information about the surrounding environment to support the judgment of safe distance.

[0003] Existing methods typically employ lidar to collect point clouds and perform point cloud registration and fusion to construct a 3D environment model. The distance to the guide wire is then calculated based on the 3D environment model for safety assessment. However, uneven coverage and void overlap can easily occur during the accumulation of point clouds from multiple angles, which reduces the model density and the stability of distance calculation. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, device and medium for detecting safe distances in 3D modeling of live-line work, so as to solve the technical problem that the safety distance detection results are unreliable due to the unstable field of view coverage during the point cloud acquisition and fusion process.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for detecting safe distances in three-dimensional modeling of live-line working, the method comprising: Obtain the temporal point cloud frames and corresponding gimbal angles obtained from multi-angle scanning; Each of the aforementioned temporal point cloud frames is projected onto a unit sphere, and the grid occupancy is updated based on the unit sphere grid to calculate the cumulative field of view coverage. The target point cloud frame set is determined based on the cumulative field of view coverage. An initial pose transformation matrix is ​​generated for the target point cloud frame set based on the gimbal angle. The target point cloud frame set is then subjected to initial pose transformation processing according to the initial pose transformation matrix to obtain an initial aligned point cloud. The initial aligned point cloud is subjected to point cloud registration processing to obtain a registration transformation matrix, and the initial aligned point cloud is transformed based on the registration transformation matrix to obtain a registered point cloud; The registered point cloud is fused to generate a three-dimensional environment model, and the shortest distance of the guide wire is calculated based on the three-dimensional environment model to obtain the safe distance detection result.

[0006] By adopting the above technical solutions, and by acquiring time-series point cloud frames obtained from multi-angle scanning and the corresponding pan-tilt angles, a multi-view spatial sampling data foundation can be provided for the working environment, thereby improving the data integrity and usability of subsequent 3D modeling. By projecting the time-series point cloud frames onto a unit sphere and calculating the cumulative field of view coverage based on the unit spherical grid to determine the target point cloud frame set, the effective coverage of point cloud acquisition can be quantitatively evaluated and high-value point cloud frames can be selected, thereby reducing redundant data and improving modeling efficiency and coverage uniformity. By generating an initial pose transformation matrix based on the pan-tilt angle and performing an initial pose transformation on the target point cloud frame set to obtain an initial aligned point cloud, the degree of initial misalignment between frames can be reduced by utilizing the prior scan posture, thereby reducing the registration search space and improving registration stability. By performing point cloud registration on the initial aligned point cloud to obtain a registration transformation matrix and fusing it to generate a 3D environment model to calculate the shortest distance of the conductor, a reliable environmental spatial expression can be formed under a unified coordinate system and the safe distance detection results can be accurately output, thereby improving the safety and real-time protection capabilities of live-line work.

[0007] In one example, the present invention can be further configured as follows: acquiring the temporal point cloud frame obtained by multi-angle scanning and the corresponding gimbal angle includes: According to the preset wide step multi-angle sequence, the two-axis electric gimbal is controlled to drive the non-repetitive scanning lidar to perform scanning and acquisition, thereby obtaining the time-series point cloud frame. When collecting each time-series point cloud frame, the gimbal angle corresponding to the current time-series point cloud frame is obtained and associated with it for storage, thus obtaining gimbal angle recording data. Based on the gimbal angle recording data, a correspondence between frame number and angle is established for the time-series point cloud frame.

[0008] By adopting the above technical solution, and by controlling the two-axis electric gimbal to drive the non-repeating scanning lidar to scan and collect data according to the preset wide step multi-angle sequence, and synchronously associating and storing the gimbal angle recording data, it is possible to obtain temporal point cloud frames with complementary viewpoints with fewer scanning frames and maintain a consistent correspondence between point cloud and attitude. This provides reliable input for coverage statistics and initial pose estimation and improves the reliability of multi-angle modeling.

[0009] In one example, the present invention can be further configured as follows: projecting each of the temporal point cloud frames onto a unit sphere, updating the grid occupancy based on the unit spherical grid to calculate the cumulative field of view coverage, and determining the target point cloud frame set based on the cumulative field of view coverage, includes: Divide the unit sphere into a unit spherical grid and create a grid occupancy table; Project the point cloud data in each of the time-series point cloud frames onto the unit spherical grid, and update the occupancy flag in the grid occupancy table; The cumulative field of view coverage is calculated by statistically analyzing the proportion of occupied grids in the grid occupancy table, and the target point cloud frame set is determined based on the cumulative field of view coverage.

[0010] By adopting the above technical solution, by dividing the unit spherical grid on the unit sphere and establishing a grid occupancy table, the point cloud data is projected onto the unit spherical grid and the occupancy mark is updated to count the proportion of occupied grids. This allows for the characterization of the cumulative contribution of multiple frame point clouds to the field of view and the calculation of the cumulative field of view coverage at a uniform scale. This enables the effective screening of the target point cloud frame set and improves the coverage balance and stability of point cloud modeling.

[0011] In one example, the present invention can be further configured as follows: the step of statistically analyzing the occupancy results based on the grid occupancy table and calculating the cumulative field of view coverage includes: The occupancy markers projected onto the unit spherical grid are used as the current occupancy grid set; The current occupied grid set is cumulatively updated with the historical occupied grid set of the previous time step to obtain the cumulative occupied grid set; The cumulative field of view coverage is calculated based on the proportion of the cumulative occupied grid set within the unit spherical grid.

[0012] By adopting the above technical solution, the cumulative occupied grid set is obtained by using the projected occupied marker as the current occupied grid set and updating it cumulatively with the historical occupied grid set of the previous moment. This allows for the continuous accumulation of effective field-of-view information of multiple frames of point clouds and avoids the duplicate counting of coverage statistics. As a result, the calculation result of the cumulative field-of-view coverage is more stable and reliable, and the accuracy of the target point cloud frame set determination is improved.

[0013] In one example, the present invention can be further configured as follows: the initial pose transformation processing of the target point cloud frame set according to the initial pose transformation matrix to obtain the initial aligned point cloud includes: The initial attitude relationship of each target point cloud frame relative to the reference point cloud frame is determined based on the gimbal angle recording data. Generate the initial pose transformation matrix corresponding to each of the target point cloud frames based on the initial pose relationship; Based on the initial pose transformation matrices, coordinate transformation and alignment are performed on each of the target point cloud frames to obtain the initial aligned point cloud.

[0014] By adopting the above technical solution, the initial attitude relationship between the target point cloud frame and the reference point cloud frame is determined based on the gimbal angle recording data, and an initial pose transformation matrix is ​​generated. The target point cloud frame is then subjected to coordinate transformation and alignment processing to obtain the initial aligned point cloud. This can eliminate large-scale attitude differences and improve the geometric consistency between frames before registration, thereby reducing the difficulty of point cloud registration and improving the convergence efficiency and stability of subsequent registration solutions.

[0015] In one example, the present invention can be further configured as follows: performing point cloud registration processing on the initial aligned point cloud to obtain a registration transformation matrix includes: The diagonal length of the scene bounding box is determined based on the initial aligned point cloud, and then a multi-level voxel scale is adaptively generated according to the diagonal length of the scene bounding box. The initial aligned point cloud is processed using voxel scales at different voxel scales to obtain a hierarchical point cloud. The hierarchical point cloud is coarsely registered based on normal distribution transformation to obtain the coarse registration transformation result, and the hierarchical point cloud is iteratively finely registered based on the coarse registration transformation result to obtain the registration transformation matrix.

[0016] By adopting the above technical solution, the diagonal length of the scene bounding box is determined based on the initial aligned point cloud, and multi-level voxel scales are adaptively generated. The point cloud downsampling granularity can be dynamically matched according to the scene scale while taking into account both computational load and detail preservation, thereby improving the adaptability and robustness of the registration process. By forming hierarchical point clouds at different voxel scales and performing coarse registration based on normal distribution transformation and iterative nearest-point fine registration, the accurate registration transformation matrix can be obtained by convergence step by step, thereby improving the alignment accuracy of multi-frame point clouds and improving the construction quality of the 3D environment model.

[0017] In one example, the present invention can be further configured as follows: fusing the registered point cloud to generate a three-dimensional environment model, and calculating the nearest distance of the guide wire based on the three-dimensional environment model to obtain a safe distance detection result, includes: The registered point cloud is subjected to outlier removal processing to obtain a denoised point cloud; The denoised point cloud is subjected to voxel merging and fusion processing to obtain a fused point cloud, and the three-dimensional environment model is generated based on the fused point cloud; In the three-dimensional environment model, the point cloud of the conductor region corresponding to the conductor is determined, and the conductor central axis parameters are extracted based on the point cloud of the conductor region; The spatial position of the robot end effector or tool is obtained, and the shortest distance from the robot end effector or tool to the guide wire is calculated based on the guide wire center axis parameters, thereby generating the safe distance detection result.

[0018] By adopting the above technical solutions, outlier removal from the registered point cloud yields a denoised point cloud, which suppresses the interference of scattered noise on the fusion modeling, thereby improving the stability and reliability of the 3D environment model. Voxel merging and fusing the denoised point cloud to generate a 3D environment model creates a continuous and dense environmental space representation, providing accurate support for conductor position analysis and distance calculation. Determining the conductor region point cloud and extracting the conductor's central axis parameters establishes the conductor's geometric representation and reduces the uncertainty in distance calculation, thus improving the accuracy of safe distance detection. Obtaining the robot's end effector or tool space position and calculating its closest distance to the conductor allows for real-time output of safe distance detection results, enhancing risk warning capabilities and safety protection levels during live-line work.

[0019] In a second aspect, the present invention provides a three-dimensional modeling safety distance detection system for live-line working, the system comprising: The point cloud acquisition module is used to acquire time-series point cloud frames obtained from multi-angle scanning and the corresponding gimbal angles; The coverage statistics module is used to project each of the time-series point cloud frames onto a unit sphere, update the grid occupancy based on the unit sphere grid to calculate the cumulative field of view coverage, and determine the target point cloud frame set based on the cumulative field of view coverage. The initial alignment module is used to generate an initial pose transformation matrix for the target point cloud frame set based on the gimbal angle, and to perform initial pose transformation processing on the target point cloud frame set according to the initial pose transformation matrix to obtain an initial aligned point cloud. The point cloud registration module is used to perform point cloud registration processing on the initial aligned point cloud to obtain a registration transformation matrix, and transform the initial aligned point cloud based on the registration transformation matrix to obtain a registered point cloud. The model ranging module is used to fuse the registered point cloud to generate a three-dimensional environment model, and calculate the shortest distance of the guide wire based on the three-dimensional environment model to obtain the safe distance detection result.

[0020] By adopting the above technical solutions, and by acquiring time-series point cloud frames obtained from multi-angle scanning and the corresponding pan-tilt angles, a multi-view spatial sampling data foundation can be provided for the working environment, thereby improving the data integrity and usability of subsequent 3D modeling. By projecting the time-series point cloud frames onto a unit sphere and calculating the cumulative field of view coverage based on the unit spherical grid to determine the target point cloud frame set, the effective coverage of point cloud acquisition can be quantitatively evaluated and high-value point cloud frames can be selected, thereby reducing redundant data and improving modeling efficiency and coverage uniformity. By generating an initial pose transformation matrix based on the pan-tilt angle and performing an initial pose transformation on the target point cloud frame set to obtain an initial aligned point cloud, the degree of initial misalignment between frames can be reduced by utilizing the prior scan posture, thereby reducing the registration search space and improving registration stability. By performing point cloud registration on the initial aligned point cloud to obtain a registration transformation matrix and fusing it to generate a 3D environment model to calculate the shortest distance of the conductor, a reliable environmental spatial expression can be formed under a unified coordinate system and the safe distance detection results can be accurately output, thereby improving the safety and real-time protection capabilities of live-line work.

[0021] In a third aspect, the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the described method for detecting safe distances in three-dimensional modeling of live-line work.

[0022] In a fourth aspect, the present invention provides a storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the live-line work three-dimensional modeling safety distance detection method. Attached Figure Description

[0023] The accompanying drawings, which form part of this specification, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart of a method for detecting safe distances in 3D modeling of live-line working, as described in an embodiment of the present invention. Figure 2 A schematic diagram of the hardware components of a safety distance detection system for 3D modeling of live-line working. Figure 3 Comparison of point clouds for guide lines (left: 64-line lidar; right: non-repeating scanning lidar). Figure 4 This is a structural block diagram of the live-line working 3D modeling safety distance detection system according to an embodiment of the present invention; Figure 5 This is a structural block diagram of an electronic device according to an embodiment of the present invention.

[0024] Explanation of reference numerals in the attached diagram: 1. Non-repeating scanning lidar; 2. Two-axis motorized pan-tilt unit; 3. Control and processing unit; 4. Insulated arm mounting bracket; 5. Insulated work bucket. Detailed Implementation

[0025] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0026] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this invention is for describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0027] Example 1 like Figure 1 As shown, this invention discloses a method for detecting safe distances in three-dimensional modeling of live-line work, which specifically includes the following steps: S10: Obtain the time-series point cloud frames and corresponding gimbal angles obtained from multi-angle scanning.

[0028] Specifically, such as Figure 2 The schematic diagram of the hardware composition of the 3D modeling safety distance detection system for live-line work shown illustrates that a non-repetitive scanning LiDAR 1 is mounted on a two-axis electric pan-tilt unit 2 and fixed to the corresponding position of an insulated work hopper 5 via an insulated arm mounting bracket 4. This ensures that the non-repetitive scanning LiDAR 1 maintains a relatively stable installation posture with respect to the work space during multi-angle scanning. Simultaneously, the point cloud data output by the non-repetitive scanning LiDAR 1 is received and buffered by the control processing unit 3. By performing multi-angle spatial scanning on the live-line work area, a point cloud sampling sequence arranged in chronological order is formed. At each sampling moment, the collected point cloud data is marked with a frame number and a timestamp is added to obtain a time-series point cloud frame. At the same time, the corresponding pan-tilt unit angle is read synchronously during point cloud acquisition and a correlation is established with the time-series point cloud frame, thereby obtaining point cloud frames and angle information that can be used for subsequent field-of-view coverage statistics and initial pose transformation processing.

[0029] S20: Project each temporal point cloud frame onto a unit sphere, update the grid occupancy based on the unit sphere grid to calculate the cumulative field of view coverage, and determine the target point cloud frame set based on the cumulative field of view coverage.

[0030] Specifically, the point cloud coordinates in the temporal point cloud frame are spherically mapped to normalize the point cloud direction information to a unit spherical space, and a spherical grid is constructed on the unit sphere to represent the scanning field of view coverage. During the continuous reception of temporal point cloud frames, the occupancy status of the spherical grid is updated according to the projection result of each frame of point cloud to form a cumulative occupancy result. Then, the cumulative field of view coverage is calculated based on the cumulative occupancy result, and the target point cloud frame set that meets the coverage requirements is determined accordingly.

[0031] S30: Generate an initial pose transformation matrix for the target point cloud frame set based on the gimbal angle, and perform initial pose transformation processing on the target point cloud frame set according to the initial pose transformation matrix to obtain the initial aligned point cloud.

[0032] Specifically, the multi-angle scanning posture represented by the gimbal angle is used as the prior pose information. The angle changes of each frame in the target point cloud frame set relative to the reference posture are converted into rigid body pose transformation relationship to generate the corresponding initial pose transformation matrix. The initial pose transformation matrix is ​​then applied to the point cloud coordinates in the target point cloud frame set to complete the initial spatial alignment, thereby obtaining the initial aligned point cloud for subsequent point cloud registration processing.

[0033] S40: Perform point cloud registration processing on the initial aligned point cloud to obtain the registration transformation matrix, and transform the initial aligned point cloud based on the registration transformation matrix to obtain the registered point cloud.

[0034] Specifically, based on the initial aligned point cloud, inter-frame spatial consistency constraints are further constructed. By iteratively matching and solving the error convergence problem of the spatial correspondence between point clouds, a registration transformation matrix that can characterize the relative motion relationship between point cloud frames is obtained. The registration transformation matrix is ​​then used to perform coordinate transformation on the initial aligned point cloud to eliminate residual alignment deviation, thereby obtaining a spatially consistent registered point cloud.

[0035] S50: The registered point cloud is fused to generate a three-dimensional environment model, and the shortest distance of the traverse is calculated based on the three-dimensional environment model to obtain the safe distance detection result.

[0036] Specifically, the registered point clouds are fused and superimposed in a unified coordinate system to form a continuous and dense spatial point cloud representation. The fusion result generates a three-dimensional environmental model to describe the spatial structure of the live-line working area. In the three-dimensional environmental model, distance calculation is performed on the spatial location of the conductor to obtain the shortest distance between the robot end effector or tool and the conductor, thereby forming a safe distance detection result for live-line working safety determination.

[0037] In one embodiment, step S10, namely acquiring the temporal point cloud frame obtained from multi-angle scanning and the corresponding gimbal angle, includes: S11: Based on the preset wide-step multi-angle sequence, control the two-axis electric pan-tilt unit to drive the non-repeating scanning lidar to perform scanning and acquisition, and obtain time-series point cloud frames.

[0038] Specifically, such as Figure 2 The schematic diagram of the hardware composition of the 3D modeling safety distance detection system for live-line work shown illustrates that the non-repetitive scanning LiDAR 1 is driven by a two-axis motorized gimbal 2 to rotate sequentially in the pitch and azimuth directions according to a preset wide-step multi-angle sequence to cover different perspectives of the target environment. The wide-step setting in pitch and azimuth is used to form complementary field of view coverage within a few frames and shorten the modeling time. At each preset angle pose, the non-repetitive scanning LiDAR 1 is triggered to perform integral acquisition to obtain a corresponding frame of point cloud data. The continuously acquired frame point clouds are then combined into a temporal point cloud frame sequence according to the acquisition order for subsequent processing.

[0039] S12: When collecting each time-series point cloud frame, obtain the gimbal angle corresponding to the current time-series point cloud frame and store it in association to obtain gimbal angle recording data.

[0040] Specifically, during the start and end of each frame of point cloud integration acquisition, the angle feedback values ​​of the two-axis motorized gimbal 2 in the pitch and azimuth directions are read synchronously, and the angle feedback values ​​are bound to the currently acquired time-series point cloud frame as the same recording unit to complete the association storage of angle and point cloud. At the same time, a frame number and timestamp are added to each record to form traceable gimbal angle recording data, thereby ensuring that the attitude information of the corresponding frame can be accurately referenced when the initial pose transformation matrix is ​​generated later.

[0041] S13: Based on the gimbal angle recording data, establish the correspondence between frame number and angle for time-series point cloud frames.

[0042] Specifically, a unique frame number is assigned to each time-series point cloud frame based on the gimbal angle recording data, and the frame number, along with the corresponding pitch and azimuth angles, is written into the frame angle correspondence table. This allows the gimbal angle of the target point cloud frame to be quickly retrieved through the frame number during subsequent processing and used for field coverage statistics and initial attitude relationship calculation.

[0043] In one embodiment, step S20 involves projecting each temporal point cloud frame onto a unit sphere, updating the grid occupancy based on the unit spherical grid to calculate the cumulative field of view coverage, and determining the target point cloud frame set based on the cumulative field of view coverage, including: S21: Divide the unit spherical surface into a unit spherical grid and create a grid occupancy table.

[0044] Specifically, the pitch and azimuth directions are discretized in the unit spherical space according to a preset angular resolution to form a unit spherical grid, and an occupancy mark is assigned to each grid cell to construct a grid occupancy table. The grid occupancy table is used to record whether the solid angles of each direction of the unit spherical surface have been hit by the point cloud, thereby providing a unified data structure and update entry for the statistics of cumulative field coverage.

[0045] S22: Project the point cloud data in each time-series point cloud frame onto the unit spherical grid and update the occupancy flag in the grid occupancy table.

[0046] Specifically, the three-dimensional coordinates of the point cloud points in each time-series point cloud frame are converted into spherical coordinate parameters to obtain the corresponding orientation angle information. Based on the orientation angle information, the point cloud points are mapped to the corresponding grid cells in the unit spherical grid. On the grid cells that are hit by the mapping, the occupancy mark is updated to the hit state to indicate that the solid angle in that direction has been illuminated by the laser, thereby realizing the frame-by-frame writing and cumulative update of the point cloud field of view coverage contribution for each frame.

[0047] S23: Calculate the cumulative field of view coverage by statistically analyzing the proportion of occupied grids in the grid occupancy table, and determine the target point cloud frame set based on the cumulative field of view coverage.

[0048] Specifically, based on the grid occupancy table, the number of grid cells in the hit state is counted to obtain the number of occupied grid cells. The ratio of the number of occupied grid cells to the total number of unit spherical grid cells is then calculated to obtain the cumulative field of view coverage C, where C satisfies C = (solid angle illuminated by laser / total solid angle) × 100%, and can be equivalently expressed as... ,in This indicates whether a unit spherical grid is hit by a point cloud. MN represents the total number of grids. The grid resolution is ≤0.5°×0.5°. When the cumulative field of view coverage reaches the preset coverage requirement, the acquired temporal point cloud frames are determined as the target point cloud frame set.

[0049] In one embodiment, step S23, namely, calculating the cumulative field of view coverage based on the grid occupancy table statistics of the occupancy results, includes: S231: Use the occupancy markers projected onto the unit spherical grid as the current occupancy grid set.

[0050] Specifically, after completing the unit spherical projection of the current temporal point cloud frame, the occupancy markers of all raster cells hit by the projection in this frame are extracted into the current occupied raster set, and the current occupied raster set is cached in the form of raster index or occupancy bitmap to represent the new contribution of the current frame to the field of view coverage, thereby providing an input set for the cumulative update of historical occupancy results.

[0051] S232: Update the current occupied grid set with the historical occupied grid set of the previous time step to obtain the cumulative occupied grid set.

[0052] Specifically, the current occupied raster set is updated by performing a set union update with the historical occupied raster set saved at the previous moment to obtain the cumulative occupied raster set, and the occupancy mark of the corresponding raster unit is synchronously refreshed in the raster occupancy table to maintain the consistency of the cumulative state, so that the cumulative field of view coverage can reflect the overall coverage of all temporal point cloud frames from the start of scanning to the current moment.

[0053] S233: Calculate the cumulative field of view coverage based on the occupancy ratio of the cumulative occupied grid set in a unit spherical grid.

[0054] Specifically, the number of elements in the cumulative occupied grid set is counted to obtain the cumulative occupied grid number, and the cumulative field of view coverage C is calculated by the ratio of the cumulative occupied grid number to the total number of unit spherical grids to obtain the increasing result of the coverage over time. C can be expressed as C = (cumulative occupied grid number / Ngrid) × 100%, and the coverage result is written into the coverage record sequence to determine the truncation position and stopping condition of the target point cloud frame set.

[0055] In one embodiment, step S30, namely, performing initial pose transformation processing on the target point cloud frame set according to the initial pose transformation matrix to obtain the initial aligned point cloud, includes: S31: Determine the initial attitude relationship between each target point cloud frame and the reference point cloud frame based on the gimbal angle recording data.

[0056] Specifically, the pitch and azimuth angles corresponding to the reference point cloud frame are selected from the gimbal angle recording data as the reference attitude, and the gimbal angles and reference attitudes of each target point cloud frame are differentially calculated to obtain the initial attitude relationship of each frame relative to the reference point cloud frame. The initial attitude relationship is used to characterize the rigid body rotation change caused by multi-angle scanning, thereby providing angle prior constraints for the generation of the initial pose transformation matrix.

[0057] S32: Generate the initial pose transformation matrix corresponding to each target point cloud frame based on the initial pose relationship.

[0058] Specifically, the pitch and azimuth angles in the initial attitude relationship are converted into rotation matrices and combined with the translation term of the origin of the coordinate system to form a homogeneous transformation expression, so as to generate an initial pose transformation matrix corresponding to each target point cloud frame. The initial pose transformation matrix is ​​used to map each frame of point cloud from the acquisition attitude coordinate system to a reference coordinate system consistent with the reference point cloud frame, thereby providing an initial alignment state that is closer to the real solution for subsequent point cloud registration processing.

[0059] S33: Based on each initial pose transformation matrix, perform coordinate transformation and alignment processing on each target point cloud frame to obtain the initial aligned point cloud.

[0060] Specifically, homogeneous coordinate expansion is performed on the point cloud coordinates in each target point cloud frame, and matrix multiplication is performed with the corresponding initial pose transformation matrix to complete the coordinate transformation. The transformed point cloud coordinates are then uniformly projected onto the reference coordinate system of the base point cloud frame and superimposed between frames to obtain an initial aligned point cloud that has eliminated large-scale pose differences and can be used for fine registration solution.

[0061] In one embodiment, step S40, namely, performing point cloud registration processing on the initial aligned point cloud to obtain a registration transformation matrix, includes: S41: Determine the diagonal length of the scene bounding box based on the initial aligned point cloud, and then adaptively generate multi-level voxel scales according to the diagonal length of the scene bounding box.

[0062] Specifically, in the initial aligned point cloud, the minimum and maximum coordinates of the point cloud in the three axes are counted to construct the scene bounding box, and the diagonal length of the bounding box, diag, is calculated as the scene scale description quantity. Based on diag, the voxel reference scale λ is calculated and set to λ = 0.02 × diag. At the same time, according to the registration requirements of multi-level voxels from coarse to fine, a multi-level voxel scale sequence is generated for subsequent hierarchical point cloud construction. When diag is in a small scene range, a lower limit is set to lock λ to avoid excessive computation due to excessively small voxels, and when diag is in a large scene range, an upper limit is set to lock λ to avoid loss of detail due to excessively large voxels.

[0063] S42: Perform voxel processing on the initial aligned point cloud at different voxel scales to obtain a hierarchical point cloud.

[0064] Specifically, voxelization is performed sequentially on the initial aligned point cloud for the multi-level voxel scale sequence. At each voxel scale, representative points of the point cloud falling into the same voxel grid are selected or aggregated to complete downsampling, thereby forming a hierarchical point cloud representation from coarse to fine. A unified coordinate system and index relationship are maintained between the hierarchical point clouds to support the cascaded solution of subsequent coarse and fine registration.

[0065] S43: Perform coarse registration based on normal distribution transformation on the hierarchical point cloud to obtain the coarse registration transformation result, and perform iterative nearest point fine registration on the hierarchical point cloud according to the coarse registration transformation result to obtain the registration transformation matrix.

[0066] Specifically, a coarse registration solution based on normal distribution transformation is first performed on the hierarchical point cloud corresponding to a coarser voxel scale to obtain the coarse registration transformation result. During the coarse registration process, a maximum number of iterations and a convergence threshold are set to promote rapid and stable convergence of the coarse registration result. Then, the coarse registration transformation result is used as the initial pose for fine registration, and iterative nearest-point fine registration is performed on the hierarchical point cloud corresponding to a finer voxel scale to further reduce inter-frame residuals. Fine registration uses a point-to-plane metric and sets a maximum number of iterations and an adjacent error change threshold as stopping conditions. Simultaneously, outlier corresponding points with excessive distances are removed, and the proportion of inlier points is constrained to improve registration robustness. Finally, a registration transformation matrix representing the relative rigid body transformation relationship between the initially aligned point cloud frames is obtained, and the registration error satisfies… And it decreases as the voxel scale decreases.

[0067] In one embodiment, step S50, namely fusing the registered point cloud to generate a three-dimensional environment model, and calculating the nearest distance of the guide wire based on the three-dimensional environment model to obtain the safe distance detection result, includes: S51: Perform outlier removal on the registered point cloud to obtain a denoised point cloud.

[0068] Specifically, for each point in the registered point cloud, the average distance to the set of neighboring points is calculated, and the global mean and standard deviation are statistically analyzed. If the average distance exceeds the range of the weighted standard deviation of the global mean, the point is identified as an outlier and deleted. This reduces the interference of background scattered points and flying points on subsequent fusion and central axis extraction while maintaining the continuity of slender targets such as wires, thus obtaining a noise-suppressed denoised point cloud.

[0069] S52: Perform voxel merging and fusion processing on the denoised point cloud to obtain a fused point cloud, and generate a 3D environment model based on the fused point cloud.

[0070] Specifically, the denoised point cloud is fused into voxels in a unified coordinate system, and point cloud points falling into the same fused voxel are aggregated to reduce redundant points and improve the consistency of point cloud density distribution, thereby obtaining a fused point cloud, such as... Figure 3 The comparison of point clouds of conductors shown shows that, compared with the point cloud performance of common 64-line lidar, non-repeating scanning lidar, after wide-step multi-angle complementary scanning and fusion, can form a higher point cloud coverage density and reduce the probability of missed detection in slender target areas such as conductors. This makes the fused point cloud present a more continuous, denser, and less void-filled distribution in space. The fused point cloud is stored and output as the point cloud representation of the three-dimensional environment model for use in the spatial structure representation and safety distance calculation requirements of live working areas. At the same time, it can be written in binary point cloud format to reduce the amount of data and improve the efficiency of on-site processing and transmission.

[0071] S53: Determine the point cloud of the traverse region corresponding to the traverse in the 3D environment model, and extract the traverse central axis parameters based on the traverse region point cloud.

[0072] Specifically, such as Figure 3 The comparison of conductor point clouds shown demonstrates that in scenarios with small conductor diameters and background structures such as towers or buildings, high-density conductor point clouds generated by non-repeating scanning LiDAR are more readily extracted from the 3D environment model. Therefore, in the 3D environment model, the conductor region point cloud is determined based on the elongated and continuous spatial geometric features of the conductor point cloud. Central axis fitting is then performed on the conductor region point cloud to obtain the conductor central axis parameters, where the conductor central axis is obtained using parametric equations. p0 represents the coordinates of the reference point on the central axis of the conductor, v represents the direction vector of the central axis of the conductor, t represents the scalar parameter along the direction vector, and the conductor radius r is estimated based on the cross-sectional distribution of the point cloud of the conductor region as a distance correction parameter for the surface of the conductor entity, thus outputting the conductor central axis parameter and conductor radius parameter that can be used for the nearest distance calculation.

[0073] S54: Obtain the spatial position of the robot end effector or tool, calculate the shortest distance from the robot end effector or tool to the guide wire based on the guide wire center axis parameters, and generate a safe distance detection result.

[0074] Specifically, the spatial position x of the robot end effector or tool in the coordinate system of the 3D environment model is obtained, and the center axis parameter of the guide wire is substituted into the nearest distance calculation formula. The shortest distance from the robot end effector or tool to the outer surface of the conductor is obtained, where L(t) is a point on the central axis of the conductor and satisfies L(t) = p0 + tv, p0 is the coordinate of the central axis reference point, v is the central axis direction vector, t is the central axis parameter, and r is the conductor radius. Based on the shortest distance d, a safe distance detection result containing distance values ​​and judgment labels is generated.

[0075] Example 2 like Figure 4 As shown, based on the same inventive concept as the above embodiments, the present invention also provides a three-dimensional modeling safety distance detection system for live-line working, comprising: The point cloud acquisition module is used to acquire time-series point cloud frames obtained from multi-angle scanning and the corresponding gimbal angles; The coverage statistics module is used to project each temporal point cloud frame onto a unit sphere, update the grid occupancy based on the unit sphere grid to calculate the cumulative field of view coverage, and determine the target point cloud frame set based on the cumulative field of view coverage. The initial alignment module is used to generate an initial pose transformation matrix for the target point cloud frame set based on the gimbal angle, and to perform initial pose transformation processing on the target point cloud frame set according to the initial pose transformation matrix to obtain the initial aligned point cloud. The point cloud registration module is used to perform point cloud registration processing on the initial aligned point cloud, obtain the registration transformation matrix, and transform the initial aligned point cloud based on the registration transformation matrix to obtain the registered point cloud; The model ranging module is used to fuse the registered point cloud to generate a three-dimensional environment model, and calculate the shortest distance of the traverse based on the three-dimensional environment model to obtain the safe distance detection result.

[0076] Optionally, the point cloud acquisition module includes: The scanning and acquisition submodule is used to control the two-axis electric pan-tilt head to drive the non-repeating scanning lidar to perform scanning and acquisition according to the preset wide step multi-angle sequence, so as to obtain time-series point cloud frames. The angle recording submodule is used to obtain the gimbal angle corresponding to the current time-series point cloud frame when collecting each time-series point cloud frame and perform associated storage to obtain gimbal angle recording data. The frame angle mapping submodule is used to establish the correspondence between frame number and angle for time-series point cloud frames based on the gimbal angle recording data.

[0077] Optional, the coverage statistics module includes: The grid partitioning submodule is used to divide a unit spherical surface into unit spherical grids and create a grid occupancy table. The spherical projection submodule is used to project the point cloud data in each time-series point cloud frame onto a unit spherical grid and update the occupancy flag in the grid occupancy table. The coverage calculation submodule is used to calculate the cumulative field of view coverage by statistically analyzing the proportion of occupied rasters in the raster occupancy table, and to determine the target point cloud frame set based on the cumulative field of view coverage.

[0078] Optional, the coverage computation submodule includes: The currently occupied cell is used to set the occupied markers projected onto the unit spherical grid as the currently occupied grid set; The occupancy accumulation unit is used to cumulatively update the current occupancy grid set with the historical occupancy grid set of the previous time step to obtain the cumulative occupancy grid set; The coverage update unit is used to calculate the cumulative field of view coverage based on the proportion of the cumulative occupied raster set in the unit spherical raster.

[0079] Optionally, the initial alignment module includes: The attitude relationship submodule is used to determine the initial attitude relationship between each target point cloud frame and the reference point cloud frame based on the gimbal angle recording data. The matrix generation submodule is used to generate the initial pose transformation matrix corresponding to each target point cloud frame based on the initial pose relationship. The coordinate alignment submodule is used to perform coordinate transformation and alignment processing on each target point cloud frame based on each initial pose transformation matrix to obtain the initial aligned point cloud.

[0080] Optionally, the point cloud registration module includes: The scale generation submodule is used to determine the diagonal length of the scene bounding box based on the initial aligned point cloud, and then adaptively generate multi-level voxel scales according to the diagonal length of the scene bounding box. The voxel layering submodule is used to perform voxel processing on the initial aligned point cloud at different voxel scales to obtain a layered point cloud. The coarse and fine registration submodule is used to perform coarse registration based on normal distribution transformation on the hierarchical point cloud to obtain the coarse registration transformation result, and to perform iterative nearest point fine registration on the hierarchical point cloud based on the coarse registration transformation result to obtain the registration transformation matrix.

[0081] Optionally, the model ranging module includes: The outlier removal submodule is used to remove outliers from the registered point cloud to obtain a denoised point cloud. The fusion modeling submodule is used to perform voxel merging and fusion processing on the denoised point cloud to obtain a fused point cloud, and generate a 3D environment model based on the fused point cloud. The conductor axis extraction submodule is used to determine the point cloud of the conductor region corresponding to the conductor in the 3D environment model, and extract the conductor center axis parameters based on the conductor region point cloud; The distance calculation submodule is used to obtain the spatial position of the robot end effector or tool, and calculate the shortest distance from the robot end effector or tool to the guide wire based on the guide wire center axis parameters, and generate a safe distance detection result.

[0082] Example 3 like Figure 5 As shown, the present invention also provides an electronic device 100 for implementing a method for detecting safe distances in three-dimensional modeling of live-line work; The electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on at least one processor 102, and at least one communication bus 104.

[0083] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the live-line work three-dimensional modeling safety distance detection method of Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101.

[0084] The memory 101 may primarily include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the electronic device 100 (such as audio data), etc. In addition, the memory 101 may include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other non-volatile solid-state storage device.

[0085] At least one processor 102 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 102 may be a microprocessor or any conventional processor. Processor 102 is the control center of electronic device 100, connecting various parts of electronic device 100 via various interfaces and lines.

[0086] The memory 101 in the electronic device 100 stores multiple instructions to implement a method for detecting safe distances in 3D modeling for live-line work, and the processor 102 can execute multiple instructions to achieve the following: Obtain the temporal point cloud frames and corresponding gimbal angles obtained from multi-angle scanning; Each of the aforementioned temporal point cloud frames is projected onto a unit sphere, and the grid occupancy is updated based on the unit sphere grid to calculate the cumulative field of view coverage. The target point cloud frame set is determined based on the cumulative field of view coverage. An initial pose transformation matrix is ​​generated for the target point cloud frame set based on the gimbal angle. The target point cloud frame set is then subjected to initial pose transformation processing according to the initial pose transformation matrix to obtain an initial aligned point cloud. The initial aligned point cloud is subjected to point cloud registration processing to obtain a registration transformation matrix, and the initial aligned point cloud is transformed based on the registration transformation matrix to obtain a registered point cloud; The registered point cloud is fused to generate a three-dimensional environment model, and the shortest distance of the guide wire is calculated based on the three-dimensional environment model to obtain the safe distance detection result.

[0087] Example 4 If the modules / units integrated in the electronic device 100 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or system capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, and read-only memory (ROM).

[0088] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0089] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0090] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including an instruction set implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0091] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0092] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for detecting safe distances in three-dimensional modeling of live-line working, characterized in that, The method includes: Obtain the temporal point cloud frames and corresponding gimbal angles obtained from multi-angle scanning; Each of the aforementioned temporal point cloud frames is projected onto a unit sphere, and the grid occupancy is updated based on the unit sphere grid to calculate the cumulative field of view coverage. The target point cloud frame set is determined based on the cumulative field of view coverage. An initial pose transformation matrix is ​​generated for the target point cloud frame set based on the gimbal angle. The target point cloud frame set is then subjected to initial pose transformation processing according to the initial pose transformation matrix to obtain an initial aligned point cloud. The initial aligned point cloud is subjected to point cloud registration processing to obtain a registration transformation matrix, and the initial aligned point cloud is transformed based on the registration transformation matrix to obtain a registered point cloud; The registered point cloud is fused to generate a three-dimensional environment model, and the shortest distance of the guide wire is calculated based on the three-dimensional environment model to obtain the safe distance detection result.

2. The method for detecting safe distances in three-dimensional modeling of live-line working according to claim 1, characterized in that, The acquisition of the temporal point cloud frames obtained from multi-angle scanning and the corresponding gimbal angles includes: According to the preset wide step multi-angle sequence, the two-axis electric gimbal is controlled to drive the non-repetitive scanning lidar to perform scanning and acquisition, thereby obtaining the time-series point cloud frame. When collecting each time-series point cloud frame, the gimbal angle corresponding to the current time-series point cloud frame is obtained and associated with it for storage, thus obtaining gimbal angle recording data. Based on the gimbal angle recording data, a correspondence between frame number and angle is established for the time-series point cloud frame.

3. The method for detecting safe distances in three-dimensional modeling of live-line working according to claim 1, characterized in that, The step of projecting each of the temporal point cloud frames onto a unit sphere, updating the grid occupancy based on the unit spherical grid to calculate the cumulative field of view coverage, and determining the target point cloud frame set based on the cumulative field of view coverage includes: Divide the unit sphere into a unit spherical grid and create a grid occupancy table; Project the point cloud data in each of the time-series point cloud frames onto the unit spherical grid, and update the occupancy flag in the grid occupancy table; The cumulative field of view coverage is calculated by statistically analyzing the proportion of occupied grids in the grid occupancy table, and the target point cloud frame set is determined based on the cumulative field of view coverage.

4. The method for detecting safe distances in three-dimensional modeling of live-line working according to claim 3, characterized in that, The step of calculating the cumulative field of view coverage based on the grid occupancy table includes: The occupancy markers projected onto the unit spherical grid are used as the current occupancy grid set; The current occupied grid set is cumulatively updated with the historical occupied grid set of the previous time step to obtain the cumulative occupied grid set; The cumulative field of view coverage is calculated based on the proportion of the cumulative occupied grid set within the unit spherical grid.

5. The method for detecting safe distances in three-dimensional modeling of live-line working according to claim 2, characterized in that, The step of performing initial pose transformation processing on the target point cloud frame set according to the initial pose transformation matrix to obtain the initial aligned point cloud includes: The initial attitude relationship of each target point cloud frame relative to the reference point cloud frame is determined based on the gimbal angle recording data. Generate the initial pose transformation matrix corresponding to each of the target point cloud frames based on the initial pose relationship; Based on the initial pose transformation matrices, coordinate transformation and alignment are performed on each of the target point cloud frames to obtain the initial aligned point cloud.

6. The method for detecting safe distances in three-dimensional modeling of live-line working according to claim 1, characterized in that, The step of performing point cloud registration processing on the initial aligned point cloud to obtain the registration transformation matrix includes: The diagonal length of the scene bounding box is determined based on the initial aligned point cloud, and then a multi-level voxel scale is adaptively generated according to the diagonal length of the scene bounding box. The initial aligned point cloud is processed using voxel scales at different voxel scales to obtain a hierarchical point cloud. The hierarchical point cloud is coarsely registered based on normal distribution transformation to obtain the coarse registration transformation result, and the hierarchical point cloud is iteratively finely registered based on the coarse registration transformation result to obtain the registration transformation matrix.

7. The method for detecting safe distances in three-dimensional modeling of live-line working according to claim 1, characterized in that, The process of fusing the registered point cloud to generate a three-dimensional environment model, and calculating the nearest distance to the guide wire based on the three-dimensional environment model to obtain the safe distance detection result, includes: The registered point cloud is subjected to outlier removal processing to obtain a denoised point cloud; The denoised point cloud is subjected to voxel merging and fusion processing to obtain a fused point cloud, and the three-dimensional environment model is generated based on the fused point cloud; In the three-dimensional environment model, the point cloud of the conductor region corresponding to the conductor is determined, and the conductor central axis parameters are extracted based on the point cloud of the conductor region; The spatial position of the robot end effector or tool is obtained, and the shortest distance from the robot end effector or tool to the guide wire is calculated based on the guide wire center axis parameters, thereby generating the safe distance detection result.

8. A three-dimensional modeling safety distance detection system for live-line working, characterized in that, The system includes: The point cloud acquisition module is used to acquire time-series point cloud frames obtained from multi-angle scanning and the corresponding gimbal angles; The coverage statistics module is used to project each of the time-series point cloud frames onto a unit sphere, update the grid occupancy based on the unit sphere grid to calculate the cumulative field of view coverage, and determine the target point cloud frame set based on the cumulative field of view coverage. The initial alignment module is used to generate an initial pose transformation matrix for the target point cloud frame set based on the gimbal angle, and to perform initial pose transformation processing on the target point cloud frame set according to the initial pose transformation matrix to obtain an initial aligned point cloud. The point cloud registration module is used to perform point cloud registration processing on the initial aligned point cloud to obtain a registration transformation matrix, and transform the initial aligned point cloud based on the registration transformation matrix to obtain a registered point cloud. The model ranging module is used to fuse the registered point cloud to generate a three-dimensional environment model, and calculate the shortest distance of the guide wire based on the three-dimensional environment model to obtain the safe distance detection result.

9. An electronic device, characterized in that, It includes a processor and a memory, the processor being used to execute a computer program stored in the memory to implement the steps of the live-line work three-dimensional modeling safety distance detection method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction, which, when executed by a processor, implements the steps of the live-line working three-dimensional modeling safety distance detection method as described in any one of claims 1 to 7.

Images