A method for registering laser scanning point clouds based on image analysis

Abstract

This invention discloses a laser scanning point cloud registration method based on image analysis, comprising the following steps: acquiring point cloud data, scene image data, spatial target sphere position data, and environmental parameter data to generate a multi-source spatial observation dataset; performing spatial topology correlation analysis to generate a spatial reference propagation relationship set; performing propagation parameter calculation to obtain propagation direction, propagation weight, and propagation perturbation parameters, generating a compensated spatial reference propagation field; performing local structure correlation analysis to generate a point cloud local structure set; mapping the point cloud local structure set to the compensated spatial reference propagation field to generate stable and drifting point cloud structures; generating a point cloud registration propagation path; and performing spatial alignment processing to generate the laser scanning point cloud registration result. This invention employs spatial propagation correlation and propagation path optimization methods to achieve continuous and stable registration of laser scanning point clouds, possessing the advantages of high registration stability and high spatial alignment accuracy.

Description

Technical Field

[0001] This invention relates to the field of laser scanning point cloud registration, and more particularly to a laser scanning point cloud registration method based on image analysis. Background Technology

[0002] With the development of laser scanning technology and 3D reconstruction technology, laser scanning point cloud registration methods have been widely used in fields such as architectural surveying, industrial modeling, spatial scene reconstruction and autonomous driving environment perception. Existing laser scanning point cloud registration methods usually achieve spatial registration between different point clouds through point cloud feature matching, spatial position alignment or rigid body transformation, and combine scene image data to improve the recognition ability of point cloud structure correspondence, thereby improving point cloud registration accuracy and spatial reconstruction effect.

[0003] Most existing technologies only perform registration processing based on local point cloud features or single spatial correspondences, lacking overall analysis of the continuous state of spatial propagation and the stable state of structure. When there are structural drift regions, local spatial disturbances, or image structure breaks in the scene, it is easy to cause interruption of point cloud propagation path, instability of spatial propagation relationship and offset of point cloud registration result, thereby reducing the stability of point cloud registration and spatial alignment accuracy in complex spatial scenes. Summary of the Invention

[0004] One objective of this invention is to propose a laser scanning point cloud registration method based on image analysis. This invention employs spatial propagation association and propagation path optimization methods to achieve continuous and stable registration of laser scanning point clouds, possessing the advantages of high registration stability and high spatial alignment accuracy.

[0005] A laser scanning point cloud registration method based on image analysis according to an embodiment of the present invention includes the following steps:

[0006] Collect point cloud data, scene image data, spatial target sphere position data, and environmental parameter data, and preprocess them to generate a multi-source spatial observation dataset;

[0007] Spatial topological correlation analysis is performed on the multi-source spatial observation dataset to map the distance correlation, scanning angle correlation and image structure correlation between spatial target spheres into propagation edges, generating a set of spatial reference propagation relationships;

[0008] Propagation parameters are calculated on the set of spatial reference propagation relationships to obtain the propagation direction, propagation weight, and propagation perturbation parameters, and a compensated spatial reference propagation field is generated.

[0009] Perform local structure correlation analysis on point cloud data to extract spatial curvature features, edge change features, and spatial continuity features, and generate a set of local structure features of point cloud.

[0010] Map the local structure set of the point cloud to the compensated spatial reference propagation field, calculate the propagation stability value and the structure offset value, and generate the stable region and the drift region of the point cloud structure.

[0011] Extract continuous features of image edges and stable features of image texture from scene image data, and generate point cloud registration propagation paths based on continuous features of image edges, stable features of image texture, and stable regions of point cloud structure;

[0012] Spatial alignment processing is performed based on the point cloud registration propagation path and the point cloud structure drift region to generate laser scanning point cloud registration results.

[0013] Optionally, the point cloud data includes laser scan point coordinate data, point cloud reflection intensity data, point cloud normal vector data, and timestamp data; the scene image data includes scene texture image data and scene edge image data; the spatial target sphere position data includes spatial target sphere center coordinate data, spatial target sphere spacing data, and spatial target sphere spatial distribution data; the environmental parameter data includes environmental temperature data, environmental humidity data, and environmental air pressure data.

[0014] The preprocessing specifically includes performing noise point removal, outlier filtering, and coordinate unification processing on point cloud data; performing image denoising and image distortion correction processing on scene image data; performing spatial coordinate calibration processing on spatial target sphere position data; performing time synchronization and outlier data filtering processing on environmental parameter data; forming spatial observation units by associating local point cloud regions, local scene image regions, and environmental parameter data within the corresponding spatial regions based on the spatial region corresponding to the spatial target sphere; and performing spatial organization processing on multiple spatial observation units to generate a multi-source spatial observation dataset.

[0015] Optionally, the generation of the spatial reference propagation relation set specifically includes:

[0016] Read multiple space observation units from the multi-source space observation dataset, extract the center coordinate data of the space target sphere and the spatial distribution data of the space target sphere corresponding to each space observation unit, and calculate the spatial distance relationship between the space observation units.

[0017] Based on the spatial distance relationship, retrieve the coordinate data and timestamp data of the laser scanning points in the corresponding spatial observation unit, calculate the change in scanning direction and the change in scanning angle between adjacent spatial observation units, and generate scanning angle correlation results;

[0018] Read scene texture image data and scene edge image data from adjacent spatial observation units, extract texture continuous regions and edge continuous regions between adjacent spatial observation units, and generate image structure association results;

[0019] Based on spatial distance relationships, scanning angle correlation results, and image structure correlation results, topological organization processing is performed on the propagation correlation between multiple spatial observation units to generate a spatial topological correlation structure.

[0020] Each association structure in the spatial topological association structure is mapped to a corresponding propagation edge, and propagation organization processing is performed on multiple propagation edges to generate a set of spatial reference propagation relationships.

[0021] Optionally, the generation of the compensated spatial reference propagation field specifically includes:

[0022] Read multiple propagation edges from the spatial reference propagation relationship set, extract the spatial observation unit position relationship, scanning angle association result and image structure association result corresponding to each propagation edge, and calculate the spatial propagation direction corresponding to each propagation edge;

[0023] Around the spatial propagation direction corresponding to each propagation edge, the spatial distance relationship, scene texture image data and scene edge image data in the corresponding spatial observation unit are retrieved to calculate the propagation influence range corresponding to each propagation edge and generate propagation weights.

[0024] Read the ambient temperature, ambient humidity and ambient air pressure data from the environmental parameter data, calculate the environmental changes and spatial disturbance changes corresponding to each space observation unit, and generate propagation disturbance parameters;

[0025] Based on the propagation direction, propagation weight, and propagation perturbation parameters, spatial distribution organization processing is performed on the propagation parameter correlation relationship between multiple propagation edges to generate a spatial distribution structure of propagation parameters.

[0026] Perform propagation compensation update processing on multiple propagation edges in the spatial distribution structure of propagation parameters, correct the propagation direction and the propagation offset in the propagation weight, and generate propagation compensation results;

[0027] Based on the propagation compensation results, propagation field organization processing is performed on multiple propagation edges to generate a compensated spatial reference propagation field.

[0028] Optionally, the generation of the spatial distribution structure of the propagation parameters specifically includes:

[0029] Read the propagation direction, propagation weight, and propagation perturbation parameters corresponding to multiple propagation edges, and extract the spatial distance relationship corresponding to each propagation edge;

[0030] Based on spatial distance relationships, spatial association organization processing is performed on multiple propagation edges to establish spatial association relationships between them;

[0031] Based on the spatial relationship between propagation edges, perform directional continuity analysis on multiple propagation directions and calculate the directional change between adjacent propagation edges;

[0032] Read the propagation weights and propagation perturbation parameters corresponding to multiple propagation edges, and calculate the changes in propagation weights and perturbation.

[0033] Based on the changes in direction, propagation weight, and perturbation, propagation parameter space organization processing is performed on multiple propagation edges to generate propagation parameter space association results.

[0034] Spatial distribution fusion processing is performed on multiple propagation edges in the spatial association result of propagation parameters to generate a spatial distribution structure of propagation parameters.

[0035] Optionally, the execution of the propagation compensation update process specifically includes:

[0036] Read multiple propagation edges in the spatial distribution structure of propagation parameters, and extract the propagation direction, propagation weight, and propagation perturbation parameters corresponding to each propagation edge;

[0037] Around the propagation direction corresponding to multiple propagation edges, calculate the directional change between adjacent propagation edges and identify offset propagation edges with abnormal directional change.

[0038] Based on the propagation perturbation parameters corresponding to the offset propagation edge, the propagation direction of the offset propagation edge is corrected by direction offset, and the propagation direction parameters corresponding to the offset propagation edge are updated.

[0039] Around the propagation weights corresponding to multiple propagation edges, calculate the change in propagation weights between adjacent propagation edges, and identify abnormal propagation edges with abnormal changes in propagation weights;

[0040] Based on the propagation perturbation parameters corresponding to the abnormal propagation edges, the propagation weights of the abnormal propagation edges are corrected, and the propagation weight parameters corresponding to the abnormal propagation edges are updated.

[0041] Perform propagation continuity organization processing on multiple propagation edges after completing the direction offset correction and propagation weight correction to generate propagation compensation update results.

[0042] Optionally, the generation of the stable region and the drift region of the point cloud structure specifically includes:

[0043] Read multiple point cloud local structures from the point cloud local structure set, and extract the spatial curvature features, edge change features, and spatial continuity features corresponding to each point cloud local structure;

[0044] Based on the spatial location corresponding to each point cloud local structure, read multiple propagation edges in the compensated spatial reference propagation field, as well as their corresponding propagation directions and propagation weights. Associate the propagation directions and propagation weights corresponding to the multiple propagation edges with the corresponding point cloud local structure to establish the propagation mapping relationship between the point cloud local structure and the multiple propagation edges.

[0045] Based on the propagation mapping relationship, calculate the directional change between the propagation directions corresponding to multiple propagation edges and the weight change between propagation weights, calculate the propagation continuity between multiple propagation edges, and generate the propagation stability value corresponding to the local structure of the point cloud.

[0046] Read the spatial curvature features, edge change features, and spatial continuity features corresponding to multiple point cloud local structures, calculate the structural change between adjacent point cloud local structures, and generate structural offset values.

[0047] Based on the propagation stability value and the structure offset value, structural stability identification is performed on multiple local structures of point clouds to generate stable local structures of point clouds and drifting local structures of point clouds.

[0048] The stable point cloud local structure is processed to generate stable point cloud structure regions, and the drift point cloud local structure is processed to generate drift point cloud structure regions.

[0049] Optionally, the process of performing stable region organization processing and drift region organization processing specifically includes:

[0050] Read the local structures of multiple stable point clouds and multiple drifting point clouds, and extract the spatial location and spatial continuity features corresponding to each local structure of the point cloud.

[0051] Based on the spatial locations corresponding to multiple stable point cloud local structures, calculate the spatial distance and changes in spatial continuity features between adjacent stable point cloud local structures, and identify the connection relationships between stable point cloud local structures.

[0052] Based on the connection relationship of local structures in stable point clouds, spatial aggregation processing is performed on multiple interconnected local structures of stable point clouds to expand the continuous connection range between local structures of stable point clouds, extract the region boundary corresponding to the continuous connection range, and generate stable regions of point cloud structures.

[0053] Based on the spatial locations corresponding to multiple drift point cloud local structures, calculate the structural changes and spatial continuity feature changes between adjacent drift point cloud local structures, and identify the connection relationships of drift point cloud local structures.

[0054] Based on the connection relationship of the local structure of the drifting point cloud, drift space aggregation processing is performed on multiple interconnected local structures of the drifting point cloud to expand the drift connection range between the local structures of the drifting point cloud, extract the region boundary corresponding to the drift connection range, and generate the drift region of the point cloud structure.

[0055] Spatial location association processing is performed on stable and drifting regions of the point cloud structure to generate regional spatial distribution results.

[0056] Optionally, the generation of the point cloud registration propagation path specifically includes:

[0057] Read the continuous features of image edges and stable features of image texture in scene image data;

[0058] By associating the spatial location of the stable region of the point cloud structure with the continuous features of the image edge and the stable features of the image texture in the corresponding image region, a propagation relationship between the stable region of the point cloud structure and the image region is established.

[0059] Based on propagation correlation, identify the regional propagation connectivity between adjacent stable point cloud regions;

[0060] Perform propagation path extension processing on multiple stable point cloud regions with regional propagation connection relationships to establish regional propagation path connection relationships between multiple stable point cloud regions;

[0061] Based on the regional propagation path connectivity, candidate point cloud registration propagation paths are extracted between multiple point cloud structurally stable regions;

[0062] Around multiple stable point cloud structures in the candidate point cloud registration propagation path, the spatial intersection between the candidate point cloud registration propagation path and the point cloud structure drift region is detected, and the candidate point cloud registration propagation path that has spatial intersection with the point cloud structure drift region is eliminated.

[0063] The remaining candidate point cloud registration propagation paths are determined as the point cloud registration propagation paths.

[0064] Optionally, the elimination specifically includes:

[0065] Read multiple candidate point cloud registration propagation paths and point cloud structure drift regions, and extract the spatial location corresponding to each candidate point cloud registration propagation path.

[0066] Based on the spatial location corresponding to the candidate point cloud registration propagation path, detect the spatial intersection between the candidate point cloud registration propagation path and the point cloud structure drift region, and identify the candidate point cloud registration propagation path that has spatial intersection with the point cloud structure drift region;

[0067] For candidate point cloud registration propagation paths with spatial intersections, extract the corresponding spatial intersection positions between the candidate point cloud registration propagation paths and the point cloud structure drift regions;

[0068] Based on the spatial intersection location, the propagation path of the candidate point cloud registration is truncated to separate the propagation path part that has spatial intersection with the point cloud structure drift region;

[0069] Based on the candidate point cloud registration propagation path after the propagation path truncation process, the spatial continuity between the remaining propagation paths is detected, and candidate point cloud registration propagation paths that cannot form continuous propagation paths are identified.

[0070] Candidate point cloud registration propagation paths that cannot form a continuous propagation path are eliminated, and the remaining candidate point cloud registration propagation paths are determined as point cloud registration propagation paths.

[0071] The beneficial effects of this invention are:

[0072] This invention proposes a laser scanning point cloud registration method based on image analysis. By constructing a set of spatial reference propagation relationships, a spatial distribution structure of propagation parameters, and a compensated spatial reference propagation field, it transforms the discrete spatial correspondences in the traditional point cloud registration process into continuous spatial propagation relationships. This enables the organization of continuous propagation states and the analysis of propagation stability between different spatial regions. Compared with existing point cloud registration methods that rely solely on local feature matching or single spatial position correspondences, this invention not only establishes continuous propagation connections between spatial regions but also performs joint analysis on propagation direction, propagation weight, and propagation disturbance states. This effectively improves the continuity of spatial propagation and the stability of point cloud structures in complex spatial scenes, and reduces the propagation offset effects caused by local structural breaks, spatial disturbances, and changes in scanning angles on the point cloud registration process.

[0073] This invention further combines point cloud local structure analysis and scene image continuous feature analysis. By generating point cloud structure stable regions and point cloud structure drift regions, it achieves the differentiation and processing of spatially stable structure regions and structure drift regions. Based on the stable regions, it constructs point cloud registration propagation paths and performs propagation path truncation and continuous propagation screening for drift propagation paths with spatial intersections. This avoids the propagation interference of structure drift regions on the overall spatial alignment process. Through the above methods, this invention can maintain high point cloud propagation continuity and spatial registration stability in complex spatial environments, partially occluded environments, and structural change environments, improving the spatial alignment accuracy, structural continuity, and overall registration reliability of laser scanning point cloud registration results. Attached Figure Description

[0074] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0075] Figure 1 This is a flowchart of a laser scanning point cloud registration method based on image analysis proposed in this invention;

[0076] Figure 2This is a schematic diagram of the generation of the compensated spatial reference propagation field in a laser scanning point cloud registration method based on image analysis proposed in this invention.

[0077] Figure 3 This is a schematic diagram illustrating the point cloud registration propagation path generation of a laser scanning point cloud registration method based on image analysis proposed in this invention. Detailed Implementation

[0078] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0079] refer to Figures 1-3 A laser scanning point cloud registration method based on image analysis includes the following steps:

[0080] Collect point cloud data, scene image data, spatial target sphere position data, and environmental parameter data, and preprocess them to generate a multi-source spatial observation dataset;

[0081] Spatial topological correlation analysis is performed on the multi-source spatial observation dataset to map the distance correlation, scanning angle correlation and image structure correlation between spatial target spheres into propagation edges, generating a set of spatial reference propagation relationships;

[0082] Propagation parameters are calculated on the set of spatial reference propagation relationships to obtain the propagation direction, propagation weight, and propagation perturbation parameters, and a compensated spatial reference propagation field is generated.

[0083] Perform local structure correlation analysis on point cloud data to extract spatial curvature features, edge change features, and spatial continuity features, and generate a set of local structure features of point cloud.

[0084] Map the local structure set of the point cloud to the compensated spatial reference propagation field, calculate the propagation stability value and the structure offset value, and generate the stable region and the drift region of the point cloud structure.

[0085] Extract continuous features of image edges and stable features of image texture from scene image data, and generate point cloud registration propagation paths based on continuous features of image edges, stable features of image texture, and stable regions of point cloud structure.

[0086] Spatial alignment processing is performed based on the point cloud registration propagation path and the point cloud structure drift region to generate laser scanning point cloud registration results.

[0087] In this embodiment, the point cloud data includes laser scan point coordinate data, point cloud reflection intensity data, point cloud normal vector data, and timestamp data; the scene image data includes scene texture image data and scene edge image data; the spatial target ball position data includes spatial target ball center coordinate data, spatial target ball spacing data, and spatial target ball spatial distribution data; and the environmental parameter data includes environmental temperature data, environmental humidity data, and environmental air pressure data.

[0088] The spatial target sphere is a spherical reference target set in the laser scanning area. Each spatial target sphere has a corresponding spherical size and spatial position, which is used to form a spatial reference node in the laser scanning area. Multiple spatial target spheres form a spatial reference topology structure through spatial distance and spatial distribution relationships, which is used to constrain the spatial propagation relationship and spatial alignment process between point cloud data.

[0089] Preprocessing specifically includes removing noise points, filtering outliers, and unifying coordinates on point cloud data; denoising and correcting image distortion on scene image data; calibrating spatial coordinates on spatial target sphere position data; synchronizing time and filtering out abnormal data on environmental parameter data; forming spatial observation units by associating local point cloud regions, local scene image regions, and environmental parameter data within the corresponding spatial regions based on the spatial region corresponding to the spatial target sphere; and performing spatial organization processing on multiple spatial observation units to generate a multi-source spatial observation dataset.

[0090] In this embodiment, the generation of the spatial reference propagation relation set specifically includes:

[0091] Read multiple space observation units from the multi-source space observation dataset, extract the center coordinate data of the space target sphere and the spatial distribution data of the space target sphere corresponding to each space observation unit, and calculate the spatial distance relationship between the space observation units.

[0092] Based on the spatial distance relationship, retrieve the coordinate data and timestamp data of the laser scanning points in the corresponding spatial observation unit, calculate the change in scanning direction and the change in scanning angle between adjacent spatial observation units, and generate scanning angle correlation results;

[0093] Read scene texture image data and scene edge image data from adjacent spatial observation units, extract texture continuous regions and edge continuous regions between adjacent spatial observation units, and generate image structure association results;

[0094] Based on spatial distance relationships, scanning angle correlation results, and image structure correlation results, topological organization processing is performed on the propagation correlation between multiple spatial observation units to generate a spatial topological correlation structure.

[0095] The generation of spatial topological association structures specifically includes:

[0096] The system reads the spatial distance relationships, scanning angle correlation results, and image structure correlation results between multiple spatial observation units. Based on the spatial distance values ​​in the spatial distance relationships, it establishes spatial adjacency relationships for spatial observation units that are close in spatial distance. Based on the scanning direction change and scanning angle change in the scanning angle correlation results, it establishes scanning continuity relationships for spatial observation units with similar scanning directions and small scanning angle differences. Based on the texture continuity regions and edge continuity regions in the image structure correlation results, it establishes image structure continuity relationships for spatial observation units that have texture continuity region correspondences and edge continuity region correspondences. It establishes unit connection relationships for spatial observation units that simultaneously possess spatial adjacency relationships, scanning continuity relationships, and image structure continuity relationships. Based on the unit connection relationships established between multiple spatial observation units, it performs sequential connection on spatial observation units with continuous connection relationships to form multiple continuous propagation connection chains. It reads the starting spatial observation unit and the ending spatial observation unit in multiple continuous propagation connection chains, and performs link splicing on continuous propagation connection chains where the ending spatial observation unit is consistent with the starting spatial observation unit in other continuous propagation connection chains to generate a spatial topology correlation structure.

[0097] Spatial distance relationships, scanning angle correlation results, and image structure correlation results can only reflect the local spatial proximity state, local scanning continuity state, and local image structure continuity state between spatial observation units, respectively. Therefore, a single correlation result cannot directly establish the overall spatial connection structure between multiple spatial observation units. To establish continuous connection relationships between multiple spatial observation units, it is necessary to establish multiple continuous connection chains based on the spatial adjacency relationship, scanning continuity relationship, and image structure continuity relationship between spatial observation units to form local continuous connection paths between spatial observation units. There are still connection correspondences between different continuous connection chains. Therefore, it is necessary to perform link splicing on multiple continuous connection chains to establish the overall spatial connection structure between multiple spatial observation units, thereby generating a spatial topology correlation structure.

[0098] The spatial topology association structure is a spatial propagation connection network formed between multiple spatial observation units. The spatial topology association structure includes the corresponding unit connection relationships and connection order between multiple spatial observation units. The unit connection relationship is used to characterize the connection correspondence between spatial observation units, and the connection order is used to characterize the connection sequence in the connection path. The spatial topology association structure is used to provide the spatial propagation connection foundation for subsequent propagation edge generation and the establishment of a spatial reference propagation relationship set.

[0099] Each association structure in the spatial topological association structure is mapped to a corresponding propagation edge, and propagation organization processing is performed on multiple propagation edges to generate a set of spatial reference propagation relationships;

[0100] The generation of the set of spatial reference propagation relations specifically includes:

[0101] The process involves: acquiring multiple unit connection relationships in a spatial topology; extracting the connection direction and sequence of each unit connection relationship; determining the spatial propagation direction of each unit connection relationship based on the connection direction; determining the propagation connection sequence of each unit connection relationship based on the connection sequence; establishing directional connection edges for each unit connection relationship based on the spatial propagation direction to generate corresponding propagation edges; reading the spatial propagation direction and propagation connection sequence of multiple propagation edges; performing sequential connection on propagation edges with consistent spatial propagation directions and continuous propagation connection sequences to form a continuous propagation edge sequence; reading the starting and ending spatial observation units in multiple continuous propagation edge sequences; performing link splicing on continuous propagation edge sequences where the ending spatial observation unit is consistent with the starting spatial observation unit in other continuous propagation edge sequences to establish continuous propagation associations between multiple propagation edges and generate a set of spatial reference propagation relationships.

[0102] Spatial topological association structures are used to characterize the overall spatial connection state between multiple spatial observation units. However, the unit connection relationships in spatial topological association structures can only characterize the connection correspondence between spatial observation units and cannot further characterize the spatial propagation extension direction and the continuity of the propagation path. Therefore, it is necessary to establish propagation edges with spatial propagation directions based on the connection directions and connection order corresponding to each unit connection relationship to form continuous spatial propagation extension relationships between multiple spatial observation units. Furthermore, a single propagation edge can only characterize local spatial propagation relationships. Therefore, it is necessary to perform continuous connection on the propagation connection order between multiple propagation edges to establish continuous propagation association relationships between multiple propagation edges, thereby forming a set of spatial reference propagation relationships, which provides a continuous spatial propagation relationship basis for the subsequent establishment of the compensated spatial reference propagation field.

[0103] A propagation edge is a unit connection relationship with a spatial propagation direction. The spatial propagation direction in the propagation edge is used to characterize the spatial propagation extension direction between spatial observation units, and the connection order is used to characterize the connection sequence in the spatial propagation path. Multiple propagation edges establish continuous propagation association relationships according to their corresponding connection order, thereby forming a set of spatial reference propagation relationships. The set of spatial reference propagation relationships is a spatial propagation association network formed by the continuous connection of multiple propagation edges. It is used to establish the overall spatial propagation association relationship between multiple propagation edges and provide a continuous spatial propagation relationship foundation for the subsequent establishment of the compensated spatial reference propagation field.

[0104] In this embodiment, the generation of the compensated spatial reference propagation field specifically includes:

[0105] Read multiple propagation edges from the spatial reference propagation relationship set, extract the spatial observation unit position relationship, scanning angle association result and image structure association result corresponding to each propagation edge, and calculate the spatial propagation direction corresponding to each propagation edge;

[0106] Around the spatial propagation direction corresponding to each propagation edge, the spatial distance relationship, scene texture image data and scene edge image data in the corresponding spatial observation unit are retrieved to calculate the propagation influence range corresponding to each propagation edge and generate propagation weights.

[0107] The generation of propagation weights specifically includes:

[0108] Read the spatial propagation directions corresponding to multiple propagation edges, as well as the spatial distance relationships, scene texture image data, and scene edge image data in the corresponding spatial observation units. Using the spatial observation units corresponding to each propagation edge as the center, perform spatial expansion to surrounding spatial observation units based on the spatial distance values ​​in the spatial distance relationships, and read the texture continuous regions and edge continuous regions in the corresponding spatial observation units within the spatial expansion area. Calculate the pixel overlap area of ​​texture continuous regions between adjacent spatial observation units and detect whether there are pixel overlap regions between texture continuous regions. Calculate the edge overlap length of edge continuous regions between adjacent spatial observation units and detect whether there are edge connection regions between edge continuous regions. When there are no pixel overlap regions between texture continuous regions or no edge connection regions between edge continuous regions, stop the spatial expansion in the direction corresponding to the current propagation edge. Count the number of corresponding spatial observation units within the spatial propagation coverage area after spatial expansion, and generate the propagation influence range corresponding to each propagation edge. Count the total number of spatial observation units corresponding to multiple propagation edges and calculate the proportion of the number of spatial observation units corresponding to each propagation edge in the total number of spatial observation units, generating the propagation weight corresponding to each propagation edge based on the proportion. Execute corresponding records for the propagation weights corresponding to multiple propagation edges to form the propagation weight result.

[0109] The spatial propagation coverage ranges differ between different propagation edges, thus their spatial propagation influence capabilities vary. A larger spatial propagation coverage area indicates that the corresponding propagation edge can cover more spatial observation units, and its influence on the spatial propagation extension is stronger. Therefore, it is necessary to generate propagation weights for the corresponding propagation edges based on the propagation influence range to characterize the differences in spatial propagation influence capabilities between different propagation edges. Among them, the propagation influence range is used to characterize the spatial propagation coverage capability corresponding to the propagation edge.

[0110] During spatial expansion, when there is no pixel overlap between continuous texture regions, it indicates that the texture structure between adjacent spatial observation units has been broken; when there is no edge connection between continuous edge regions, it indicates that the edge structure between adjacent spatial observation units has been broken; texture structure breakage or edge structure breakage will cause the spatial propagation expansion in the direction corresponding to the current propagation edge to be unable to maintain continuous propagation, so the spatial expansion in the direction corresponding to the current propagation edge stops.

[0111] Read the ambient temperature, ambient humidity and ambient air pressure data from the environmental parameter data, calculate the environmental changes and spatial disturbance changes corresponding to each space observation unit, and generate propagation disturbance parameters;

[0112] Based on the propagation direction, propagation weight, and propagation perturbation parameters, spatial distribution organization processing is performed on the propagation parameter correlation relationship between multiple propagation edges to generate a spatial distribution structure of propagation parameters.

[0113] Perform propagation compensation update processing on multiple propagation edges in the spatial distribution structure of propagation parameters, correct the propagation direction and the propagation offset in the propagation weight, and generate propagation compensation results;

[0114] Based on the propagation compensation results, propagation field organization processing is performed on multiple propagation edges to generate a compensated spatial reference propagation field;

[0115] The generation of the compensated spatial reference propagation field specifically includes:

[0116] Read multiple propagation edges, their corresponding propagation directions, and propagation weights from the propagation compensation results, and detect the continuous propagation connection status between the multiple propagation edges; establish spatial propagation coverage relationships for multiple propagation edges with continuous propagation directions and connected regions in their spatial propagation ranges; based on the established spatial propagation coverage relationships between the multiple propagation edges, perform spatial propagation range connection on multiple propagation edges with connected relationships in their spatial propagation ranges to form a continuous spatial propagation distribution structure; and generate a compensated spatial reference propagation field based on the continuous spatial propagation distribution structure.

[0117] The multiple propagation edges in the propagation compensation result can only represent the local spatial propagation connection state, and therefore cannot establish the overall continuous spatial propagation state between multiple spatial regions. In order to establish the overall spatial propagation range and continuous spatial propagation relationship between multiple spatial regions, it is necessary to form a continuous spatial propagation distribution structure based on the spatial propagation coverage relationship between multiple propagation edges, thereby generating the compensated spatial reference propagation field.

[0118] The compensated spatial reference propagation field is the overall spatial propagation distribution structure formed after multiple propagation edges have completed propagation compensation and updates; the spatial propagation coverage relationship is used to characterize the overall spatial propagation range between multiple propagation edges, and the continuous spatial propagation distribution structure is used to characterize the continuous spatial propagation state between multiple propagation edges; the compensated spatial reference propagation field is used to provide a continuous spatial propagation reference basis for subsequent propagation stability state analysis corresponding to the local structure of the point cloud and identification of stable regions of the point cloud structure.

[0119] In this embodiment, the generation of the spatial distribution structure of propagation parameters specifically includes:

[0120] Read the propagation direction, propagation weight, and propagation perturbation parameters corresponding to multiple propagation edges, and extract the spatial distance relationship corresponding to each propagation edge;

[0121] Based on spatial distance relationships, spatial association organization processing is performed on multiple propagation edges to establish spatial association relationships between them;

[0122] The generation of spatial relationships specifically includes:

[0123] Read the spatial observation unit connection relationships corresponding to multiple propagation edges; traverse the spatial observation unit connection relationships corresponding to multiple propagation edges and detect whether there are the same spatial observation units between different propagation edges; when there are the same spatial observation units between different propagation edges, establish the spatial association relationship between the corresponding propagation edges;

[0124] Spatial correlation is used to characterize the adjacent spatial propagation relationship between different propagation edges. When there are the same spatial observation units between different propagation edges, it indicates that the corresponding propagation edges belong to adjacent spatial propagation regions, so a spatial correlation is established between the corresponding propagation edges. Through spatial correlation, the range of adjacent propagation edges between different propagation edges can be determined, thus providing the basis for subsequent propagation direction change analysis, propagation weight change analysis, and propagation disturbance change analysis.

[0125] Based on the spatial relationship between propagation edges, perform directional continuity analysis on multiple propagation directions and calculate the directional change between adjacent propagation edges;

[0126] Read the propagation weights and propagation perturbation parameters corresponding to multiple propagation edges, and calculate the changes in propagation weights and perturbation.

[0127] Based on the changes in direction, propagation weight, and perturbation, propagation parameter space organization processing is performed on multiple propagation edges to generate propagation parameter space association results.

[0128] The generation of propagation parameter spatial correlation results specifically includes:

[0129] Read the direction change, propagation weight change, and perturbation change corresponding to multiple propagation edges; traverse propagation edges with spatial relationships, compare the direction changes between adjacent propagation edges, and identify propagation edges with the same direction change or consistent direction change trend; compare the propagation weight changes between adjacent propagation edges, and identify propagation edges with the same propagation weight change or consistent propagation weight change trend; compare the perturbation changes between adjacent propagation edges, and identify propagation edges with the same perturbation change or consistent perturbation change trend; when adjacent propagation edges have consistent direction changes, consistent propagation weight changes, and consistent perturbation changes simultaneously, divide the corresponding propagation edges into the same propagation change region; count the number of propagation edges in the same propagation change region and the spatial distribution of the corresponding propagation edges, and generate the regional propagation state of the corresponding propagation change region; read the spatial distribution of multiple propagation change regions and detect whether there is regional boundary contact between different propagation change regions; when there is regional boundary contact between different propagation change regions, establish regional propagation relationship between the corresponding propagation change regions; based on the regional propagation relationship between multiple propagation change regions, generate spatial relationship results of propagation parameters.

[0130] When different regions of propagation change have contacting boundaries, it indicates that there is a spatial propagation extension boundary between the corresponding regions of propagation change. Therefore, the propagation states between different regions of propagation change will influence each other, thus establishing a regional propagation correlation between the corresponding regions of propagation change. The regional propagation correlation is used to characterize the spatial propagation connection correspondence between different regions of propagation change. Regional boundary contact refers to the existence of spatial contact areas or overlapping areas between the boundaries of the regions corresponding to different regions of propagation change. The regional propagation correlation is used to characterize the continuous spatial propagation extension relationship between different regions of propagation change, in order to establish the propagation connection range between different regions of propagation change.

[0131] The directional change state, propagation weight change state, and perturbation change state between adjacent propagation edges are used to characterize the changes in propagation direction, propagation influence capability, and propagation perturbation between different propagation edges, respectively. When adjacent propagation edges simultaneously exhibit consistent directional changes, consistent propagation weight changes, and consistent perturbation changes, it indicates that the corresponding propagation edges have the same propagation change trend, and therefore, the corresponding propagation edges are divided into the same propagation change region. Different propagation change regions correspond to the propagation change states in different spatial regions. Therefore, by using multiple propagation change regions, the overall propagation change correlation between different spatial regions can be established, thereby generating spatial correlation results for propagation parameters.

[0132] The spatial correlation results of propagation parameters are used to characterize the overall propagation state distribution relationship between different propagation change regions. Among them, different propagation change regions correspond to different propagation direction change states, propagation weight change states, and propagation disturbance change states, respectively, to reflect the overall propagation change situation in different spatial regions. The number of propagation edges is used to characterize the propagation coverage scale corresponding to the propagation change region, and the spatial distribution location is used to characterize the spatial propagation distribution range corresponding to the propagation change region.

[0133] Perform spatial distribution fusion processing on multiple propagation edges in the spatial association result of propagation parameters to generate a spatial distribution structure of propagation parameters;

[0134] The generation of the spatial distribution structure of propagation parameters specifically includes:

[0135] The process involves reading multiple propagation change regions, regional propagation relationships, and corresponding propagation edges from the spatial association results of propagation parameters, and extracting the spatial distribution location corresponding to each propagation change region. Based on the regional propagation relationships, the spatial distance between the boundaries of different propagation change regions is calculated. When the spatial distance between the boundaries of different propagation change regions is less than the regional connection distance, a connection relationship is established between the propagation edges of the boundaries of different propagation change regions. Based on the connection relationship between the propagation edges, continuous connections are performed on the propagation edges in different propagation change regions to form multiple propagation edge connection sequences. The starting and ending propagation edges in the multiple propagation edge connection sequences are read, and it is checked whether there are identical propagation edges between different propagation edge connection sequences. When there are identical propagation edges between different propagation edge connection sequences, spatial splicing is performed on the corresponding propagation edge connection sequences to form continuous propagation connection chains. The spatial coverage and propagation direction corresponding to multiple continuous propagation connection chains are read, and a link connection relationship is established between continuous propagation connection chains with overlapping spatial coverage and continuous propagation directions. Based on the link connection relationship, multiple continuous propagation connection chains are merged to generate the spatial distribution structure of propagation parameters.

[0136] The spatial distribution structure of propagation parameters is the overall spatial propagation distribution structure formed by merging multiple continuous propagation connection chains. It is used to establish the overall spatial propagation distribution relationship between multiple propagation change regions, providing a spatial propagation distribution basis for subsequent propagation compensation updates. The propagation edge connection relationship is used to characterize the spatial connection correspondence between the propagation edges at the boundaries of different propagation change regions. The continuous propagation connection chain is the spatial propagation extension structure formed by continuously connecting multiple propagation edges. It is used to characterize the continuous propagation range of the spatial propagation path. The continuous propagation direction means that the propagation direction change trend corresponding to different continuous propagation connection chains is consistent.

[0137] In this embodiment, the propagation compensation update process specifically includes:

[0138] Read multiple propagation edges in the spatial distribution structure of propagation parameters, and extract the propagation direction, propagation weight, and propagation perturbation parameters corresponding to each propagation edge;

[0139] Around the propagation direction corresponding to multiple propagation edges, calculate the directional change between adjacent propagation edges and identify offset propagation edges with abnormal directional change.

[0140] The change in direction is used to characterize the difference in propagation direction between adjacent propagation edges. When the change in direction between adjacent propagation edges is significantly greater than the change in direction between surrounding adjacent propagation edges, it indicates that the propagation direction between the corresponding propagation edges has deviated from the continuous propagation direction. Therefore, the corresponding propagation edge is identified as an offset propagation edge.

[0141] Based on the propagation perturbation parameters corresponding to the offset propagation edge, the propagation direction of the offset propagation edge is corrected by direction offset, and the propagation direction parameters corresponding to the offset propagation edge are updated.

[0142] The specific steps for performing direction offset correction include:

[0143] Based on the propagation perturbation parameters corresponding to the offset propagation edge, read the propagation direction corresponding to the offset propagation edge and the propagation direction corresponding to the adjacent propagation edge, calculate the direction difference between the propagation direction corresponding to the offset propagation edge and the propagation direction corresponding to the adjacent propagation edge, and generate the propagation direction offset; based on the propagation direction offset, perform a direction adjustment to move the propagation direction corresponding to the offset propagation edge closer to the propagation direction corresponding to the adjacent propagation edge, and generate the updated propagation direction parameters.

[0144] The propagation perturbation parameter is used to characterize the propagation perturbation state in the spatial region corresponding to the propagation edge; the more obvious the propagation perturbation state, the greater the degree to which the propagation direction corresponding to the propagation edge deviates from the continuous propagation direction; the updated propagation direction parameter is used to characterize the propagation direction state after the direction offset correction is completed.

[0145] Around the propagation weights corresponding to multiple propagation edges, calculate the change in propagation weights between adjacent propagation edges, and identify abnormal propagation edges with abnormal changes in propagation weights;

[0146] Based on the propagation perturbation parameters corresponding to the abnormal propagation edges, the propagation weights of the abnormal propagation edges are corrected, and the propagation weight parameters corresponding to the abnormal propagation edges are updated.

[0147] The specific steps for performing propagation weight adjustments include:

[0148] Based on the propagation perturbation parameters and propagation weight offset corresponding to the abnormal propagation edge, determine the weight adjustment amount corresponding to the abnormal propagation edge; based on the weight adjustment amount, perform weight adjustment on the propagation weight corresponding to the abnormal propagation edge to move closer to the propagation weight corresponding to the adjacent propagation edge, and generate the updated propagation weight parameters.

[0149] The more pronounced the propagation disturbance, the greater the deviation of the propagation influence capability corresponding to the propagation edge from the continuous propagation state; the updated propagation weight parameters are used to characterize the propagation influence capability state after the propagation weight correction is completed.

[0150] Perform propagation continuity organization processing on multiple propagation edges after completing the direction offset correction and propagation weight correction to generate propagation compensation update results;

[0151] The specific procedures for implementing continuity of communication include:

[0152] Read multiple propagation edges after completing the direction offset correction and propagation weight correction, and detect the continuous state of propagation direction and propagation weight between adjacent propagation edges; establish propagation continuity connection relationship for propagation edges with continuous propagation direction and continuous propagation weight; perform continuous propagation connection on multiple propagation edges based on multiple propagation continuity connection relationship, and generate propagation compensation update result;

[0153] The propagation direction correction and propagation weight correction can only correct the propagation direction state and propagation influence state of a single propagation edge, respectively. Therefore, there may still be a break in the propagation continuity relationship between multiple propagation edges after the propagation direction correction and propagation weight correction are completed. In order to restore the overall continuous propagation state between multiple propagation edges, it is necessary to re-establish the propagation continuity connection relationship between multiple propagation edges to generate the propagation compensation update result. The propagation compensation update result is the continuous propagation structure formed after multiple propagation edges have completed the propagation direction correction and propagation weight correction, which is used to establish the overall continuous propagation state after compensation.

[0154] In this embodiment, the generation of the point cloud local structure set specifically includes:

[0155] Read multiple point cloud data points from the point cloud data, and detect the spatial distribution status between adjacent point cloud data points around the spatial position corresponding to each point cloud data point;

[0156] Based on the spatial distribution of adjacent point cloud data points, calculate the spatial curvature change in the corresponding spatial region of each point cloud data point and generate spatial curvature features;

[0157] Based on the spatial location corresponding to each point cloud data point, detect the edge distribution changes between adjacent point cloud data points and generate edge change features;

[0158] Based on the spatial connection status between adjacent point cloud data points, detect the spatial continuity changes in the spatial region corresponding to each point cloud data point and generate spatial continuity features;

[0159] Based on spatial curvature characteristics, edge change characteristics, and spatial continuity characteristics, local structure partitioning is performed on multiple point cloud data points to generate a set of local structure points.

[0160] In this embodiment, the generation of stable and drifting regions of the point cloud structure specifically includes:

[0161] Read multiple point cloud local structures from the point cloud local structure set, and extract the spatial curvature features, edge change features, and spatial continuity features corresponding to each point cloud local structure;

[0162] Based on the spatial location corresponding to each point cloud local structure, read multiple propagation edges in the compensated spatial reference propagation field, as well as their corresponding propagation directions and propagation weights. Associate the propagation directions and propagation weights corresponding to the multiple propagation edges with the corresponding point cloud local structures to establish a propagation mapping relationship between the point cloud local structures and multiple propagation edges.

[0163] The establishment of propagation mapping relationships specifically includes:

[0164] Based on the spatial location corresponding to each point cloud local structure, detect the spatial overlap relationship between the spatial region corresponding to the point cloud local structure and the spatial propagation range corresponding to multiple propagation edges; associate the propagation direction and propagation weight corresponding to the propagation edge with spatial overlap relationship with the corresponding point cloud local structure, and establish the propagation mapping relationship between the point cloud local structure and multiple propagation edges.

[0165] The propagation edges in the compensated spatial reference propagation field can only represent the spatial propagation state, and therefore cannot directly reflect the propagation continuity state corresponding to the local structure of the point cloud. In order to establish the correspondence between the local structure of the point cloud and the spatial propagation state, it is necessary to associate the local structure of the point cloud with the propagation edges in the corresponding spatial region to establish the propagation mapping relationship between the local structure of the point cloud and the propagation edges.

[0166] The propagation mapping relationship is used to characterize the spatial propagation association state between the local structure of the point cloud and the corresponding propagation edge; the propagation direction is used to characterize the propagation direction state in the spatial region corresponding to the local structure of the point cloud; and the propagation weight is used to characterize the propagation influence state in the spatial region corresponding to the local structure of the point cloud.

[0167] Based on the propagation mapping relationship, calculate the directional change between the propagation directions corresponding to multiple propagation edges and the weight change between propagation weights, calculate the propagation continuity between multiple propagation edges, and generate the propagation stability value corresponding to the local structure of the point cloud.

[0168] The changes in direction and weight among multiple propagation edges reflect the changes in propagation direction and propagation influence, respectively, and thus can characterize the propagation continuity among multiple propagation edges. When the changes in direction and weight among multiple propagation edges remain continuous, it indicates that the propagation state in the corresponding spatial region of the local structure of the point cloud is stable, and the local structure of the point cloud has a stable propagation state.

[0169] The propagation stability value is used to characterize the propagation stability in the spatial region corresponding to the local structure of the point cloud; the higher the propagation stability value, the more stable the propagation state of the corresponding local structure of the point cloud; the propagation stability value is used to provide a basis for the propagation stability state for subsequent identification of the stability of the local structure of the point cloud.

[0170] Read the spatial curvature features, edge change features, and spatial continuity features corresponding to multiple point cloud local structures, calculate the structural change between adjacent point cloud local structures, and generate structural offset values.

[0171] Based on the propagation stability value and the structure offset value, structural stability identification is performed on multiple local structures of point clouds to generate stable local structures of point clouds and drifting local structures of point clouds.

[0172] The propagation stability value can only reflect the propagation stability in the spatial region corresponding to the local structure of the point cloud, while the structure offset value can only reflect the structural changes between local structures of the point cloud. Therefore, it is not possible to identify the local structural stability state of the point cloud on its own. In order to establish both the propagation stability state and the structural change state corresponding to the local structure of the point cloud at the same time, it is necessary to combine the propagation stability value and the structure offset value to perform structural stability identification.

[0173] When the propagation stability value corresponding to the local structure of the point cloud remains continuous and stable and the structure offset value is small, it indicates that the propagation state and structure state in the corresponding spatial region of the local structure of the point cloud remain stable, and the local structure of the point cloud is identified as a stable local structure of the point cloud; when the propagation stability value corresponding to the local structure of the point cloud changes significantly or the structure offset value increases significantly, it indicates that the propagation state or structure state in the corresponding spatial region of the local structure of the point cloud has changed, and the local structure of the point cloud is identified as a drifting local structure of the point cloud.

[0174] Perform stable region organization processing on the local structure of the stable point cloud to generate stable regions of the point cloud structure, and perform drift region organization processing on the local structure of the drifting point cloud to generate drift regions of the point cloud structure.

[0175] Read the spatial location and spatial continuity features corresponding to multiple stable point cloud local structures, perform region connection on stable point cloud local structures that are both spatially continuous and maintain continuous spatial continuity features to form stable structure connection regions; generate stable point cloud structure regions based on multiple stable structure connection regions; read the spatial location and spatial continuity features corresponding to multiple drifting point cloud local structures, perform region connection on drifting point cloud local structures that are both spatially continuous and have continuous structural change states to form drifting structure connection regions; generate drifting point cloud structure regions based on multiple drifting structure connection regions.

[0176] A single point cloud local structure can only reflect the structural stability or structural drift state in a local spatial region, and therefore cannot establish the overall structural continuity state in a continuous spatial region. In order to establish the overall stable structural state and the overall drift structural state between multiple spatial regions, it is necessary to perform region connection on multiple adjacent point cloud local structures to form a point cloud structural stability region and a point cloud structural drift region.

[0177] The stable region of the point cloud structure is a stable spatial region formed by the continuous connection of multiple stable local structures of the point cloud. It is used to characterize the continuous stable structural state in the spatial region and to provide a stable propagation spatial region for the subsequent point cloud registration propagation path generation. The drift region of the point cloud structure is a drift spatial region formed by the continuous connection of multiple drifting local structures of the point cloud. It is used to characterize the structural change state in the spatial region and to provide a drift structure region reference for the subsequent point cloud registration propagation path filtering.

[0178] In this embodiment, performing stable region organization processing and performing drift region organization processing specifically includes:

[0179] Read the local structures of multiple stable point clouds and multiple drifting point clouds, and extract the spatial location and spatial continuity features corresponding to each local structure of the point cloud.

[0180] Based on the spatial locations corresponding to multiple stable point cloud local structures, calculate the spatial distance and changes in spatial continuity features between adjacent stable point cloud local structures, and identify the connection relationships between stable point cloud local structures.

[0181] The changes in spatial distance and spatial continuity characteristics reflect the spatial proximity and structural continuity states between local structures of stable point clouds, respectively, and thus can characterize the continuous and stable relationship between multiple local structures of stable point clouds. When the spatial distance between adjacent local structures of stable point clouds remains continuous and the changes in spatial continuity characteristics are small, it indicates that the structural continuity state between the corresponding local structures of stable point clouds remains stable, and thus a connection relationship between local structures of stable point clouds can be established.

[0182] The connection relationship of local structures in stable point clouds is used to characterize the continuous and stable connection state between multiple local structures in stable point clouds; it provides a stable structural connection basis for the subsequent generation of stable regions in point clouds; the change in spatial continuity features is used to characterize the difference in spatial continuity features between adjacent local structures in stable point clouds; the smaller the change in spatial continuity features, the more continuous and stable the spatial continuity state between adjacent local structures in stable point clouds.

[0183] Based on the connection relationship of local structures in stable point clouds, spatial aggregation processing is performed on multiple interconnected local structures of stable point clouds to expand the continuous connection range between local structures of stable point clouds, extract the region boundary corresponding to the continuous connection range, and generate stable regions of point cloud structures.

[0184] Based on the connection relationship of the local structure of the stable point cloud, read the spatial positions corresponding to multiple interconnected local structures of the stable point cloud; perform spatial region connection on multiple stable local structures with continuous spatial positions to form a continuous connection range of stable structures; read the peripheral spatial positions corresponding to the continuous connection range of stable structures, perform boundary connection processing on the peripheral spatial positions to form a region boundary; generate a stable region of point cloud structure based on the continuous connection range of stable structures and the region boundary.

[0185] A single stable point cloud local structure can only reflect the structural stability state in a local spatial region, and therefore cannot establish the overall stable structural state in a continuous spatial region. In order to establish the overall stable structural range in a continuous spatial region, it is necessary to perform spatial region connection on multiple interconnected stable point cloud local structures to form a continuous stable structural connection range, and generate a stable point cloud structure region based on the continuous stable structural connection range.

[0186] The continuous connection range of stable structures is used to characterize the continuous stable spatial range formed after the continuous connection of multiple stable point cloud local structures; the region boundary is used to characterize the outer spatial boundary corresponding to the continuous connection range of stable structures; the stable region of point cloud structure is the stable spatial region formed after the continuous connection of multiple stable point cloud local structures, used to characterize the overall stable structural state in the continuous spatial region, and to provide a stable spatial region basis for the subsequent point cloud registration and propagation path generation.

[0187] Based on the spatial locations corresponding to multiple drift point cloud local structures, calculate the structural changes and spatial continuity feature changes between adjacent drift point cloud local structures, and identify the connection relationships of drift point cloud local structures.

[0188] Based on the connection relationship of the local structure of the drifting point cloud, drift space aggregation processing is performed on multiple interconnected local structures of the drifting point cloud to expand the drift connection range between the local structures of the drifting point cloud, extract the region boundary corresponding to the drift connection range, and generate the drift region of the point cloud structure.

[0189] Based on the connection relationship of the local structure of the drifting point cloud, read the spatial positions corresponding to multiple interconnected local structures of the drifting point cloud; perform spatial region connection on multiple drifting point cloud local structures with continuous spatial positions to form a continuous connection range of drifting structures; read the outer spatial positions corresponding to the continuous connection range of drifting structures, perform boundary connection processing on the outer spatial positions to form region boundaries; generate the drifting region of the point cloud structure based on the continuous connection range of drifting structures and the region boundaries.

[0190] A single drift point cloud local structure can only reflect the structural drift state in a local spatial region, and therefore cannot establish the overall drift structure state in a continuous spatial region. In order to establish the overall drift structure range in a continuous spatial region, it is necessary to perform spatial region connection on multiple interconnected drift point cloud local structures to form a continuous drift structure connection range, and generate the point cloud structure drift region based on the continuous drift structure connection range.

[0191] The continuous connection range of the drift structure is used to characterize the continuous drift space range formed after the continuous connection of multiple drift point cloud local structures; the drift region of the point cloud structure is the drift space region formed after the continuous connection of multiple drift point cloud local structures, used to characterize the overall structural drift state in the continuous space region, and to provide a drift space region reference for subsequent point cloud registration propagation path screening.

[0192] Spatial location association processing is performed on stable and drifting regions of the point cloud structure to generate regional spatial distribution results.

[0193] The generation of regional spatial distribution results specifically includes:

[0194] Read the spatial locations corresponding to the stable and drifting regions of the point cloud structure; based on the spatial locations corresponding to the stable and drifting regions of the point cloud structure, perform stable region marking and drift region marking on different spatial regions to generate regional spatial distribution results;

[0195] The stable and drifting regions of a point cloud structure can only reflect the stable and drifting structural states in local spatial regions, respectively, and therefore cannot establish the regional distribution state in the overall spatial region. In order to establish the stable and drifting regional distribution states in the overall spatial region, it is necessary to generate regional spatial distribution results based on the spatial locations corresponding to the stable and drifting regions of the point cloud structure.

[0196] The regional spatial distribution results are used to characterize the regional distribution of stable and drifting regions of the point cloud structure in the overall space, providing a spatial regional distribution reference basis for the subsequent point cloud registration and propagation path generation.

[0197] In this embodiment, the generation of the point cloud registration propagation path specifically includes:

[0198] Read the continuous features of image edges and stable features of image texture in scene image data;

[0199] By associating the spatial location of the stable region of the point cloud structure with the continuous features of the image edge and the stable features of the image texture in the corresponding image region, a propagation relationship between the stable region of the point cloud structure and the image region is established.

[0200] Based on the spatial location corresponding to the stable region of the point cloud structure, the spatial position correspondence between the stable region of the point cloud structure and the corresponding image region is detected; the continuous features of the image edge and the stable features of the image texture in the corresponding image region are read and associated with the corresponding stable region of the point cloud structure, and the propagation correlation between the stable region of the point cloud structure and the image region is established.

[0201] A stable point cloud structure region can only reflect the structural stability state in a spatial region, and therefore cannot establish a continuous propagation state in an image region. In order to establish a continuous propagation state in an image region corresponding to a stable point cloud structure region, it is necessary to associate the stable point cloud structure region with the continuous image edge features and stable image texture features in the corresponding image region, so as to establish a propagation association between the stable point cloud structure region and the image region.

[0202] Propagation association is used to characterize the propagation correspondence between stable point cloud regions and corresponding image regions, providing a basis for subsequent region propagation connection recognition; image edge continuity features are used to characterize the edge continuity propagation state in image regions, and image texture stability features are used to characterize the texture stability propagation state in image regions.

[0203] Based on propagation correlation, identify the regional propagation connectivity between adjacent stable point cloud regions;

[0204] Based on the propagation correlation, the image edge continuity state and image texture stability state of the corresponding image regions of adjacent stable point cloud structures are detected; regional propagation connection relationship is established for adjacent stable point cloud structures that maintain continuous image edge continuity state and image texture stability state.

[0205] Propagation correlation can reflect the edge continuous propagation state and texture stable propagation state in the image region corresponding to the stable point cloud structure region. Therefore, it can characterize the propagation continuity state between adjacent stable point cloud structure regions. When the propagation state between adjacent stable point cloud structure regions remains continuous, it indicates that there is a continuous propagation state between the corresponding stable point cloud structure regions. Therefore, a region propagation connection relationship is established.

[0206] Regional propagation connectivity is used to characterize the continuous propagation connectivity state between adjacent stable regions of point cloud structure, providing a regional propagation connectivity basis for subsequent point cloud registration propagation path expansion;

[0207] Perform propagation path extension processing on multiple stable point cloud regions with regional propagation connection relationships to establish regional propagation path connection relationships between multiple stable point cloud regions;

[0208] The establishment of regional propagation path connectivity specifically includes:

[0209] Read the spatial locations and propagation directions of multiple stable point cloud structures with regional propagation connections; based on the propagation direction, perform regional propagation connection expansion on the multiple stable point cloud structures that are mutually propagated to form a continuous regional propagation connection chain; establish the regional propagation path connection relationship between the multiple stable point cloud structures based on the continuous regional propagation connection chain.

[0210] The regional propagation connection relationship can only reflect the local propagation connection state between adjacent stable point cloud structures, and therefore cannot establish the continuous propagation path state between multiple spatial regions. In order to establish a continuous regional propagation path between multiple spatial regions, it is necessary to perform regional propagation connection extension on multiple stable point cloud structures based on the regional propagation connection relationship to form a continuous regional propagation connection chain and establish the regional propagation path connection relationship.

[0211] The continuous region propagation connection chain is used to characterize the continuous propagation path structure formed after continuous propagation connection of multiple stable point cloud structures; the region propagation path connection relationship is used to characterize the continuous propagation path connection state between multiple stable point cloud structures, providing a propagation path connection basis for subsequent candidate point cloud registration propagation path extraction.

[0212] Based on the regional propagation path connectivity, candidate point cloud registration propagation paths are extracted between multiple point cloud structurally stable regions;

[0213] Around multiple stable point cloud structures in the candidate point cloud registration propagation path, the spatial intersection between the candidate point cloud registration propagation path and the point cloud structure drift region is detected, and the candidate point cloud registration propagation path that has spatial intersection with the point cloud structure drift region is eliminated.

[0214] Read the spatial propagation range corresponding to the candidate point cloud registration propagation path and the spatial region range corresponding to the point cloud structure drift region; detect whether there is a spatial overlap between the spatial propagation range corresponding to the candidate point cloud registration propagation path and the spatial region range corresponding to the point cloud structure drift region; when there is a spatial overlap, remove the corresponding candidate point cloud registration propagation path.

[0215] The structural state in the spatial region corresponding to the point cloud structure drift region is subject to drift changes, thus it cannot maintain a continuous and stable propagation state. When the candidate point cloud registration propagation path intersects with the point cloud structure drift region, it indicates that the corresponding candidate point cloud registration propagation path passes through the structure drift region, so the corresponding candidate point cloud registration propagation path needs to be eliminated.

[0216] Spatial intersection is used to characterize the spatial overlap between the spatial propagation range corresponding to the candidate point cloud registration propagation path and the spatial region corresponding to the point cloud structure drift region; candidate point cloud registration propagation paths that have spatial intersection with the point cloud structure drift region are eliminated to improve the propagation stability of the point cloud registration propagation path.

[0217] The remaining candidate point cloud registration propagation paths are determined as the point cloud registration propagation paths.

[0218] In this embodiment, the elimination specifically includes:

[0219] Read multiple candidate point cloud registration propagation paths and point cloud structure drift regions, and extract the spatial location corresponding to each candidate point cloud registration propagation path.

[0220] Based on the spatial location corresponding to the candidate point cloud registration propagation path, detect the spatial intersection between the candidate point cloud registration propagation path and the point cloud structure drift region, and identify the candidate point cloud registration propagation path that has spatial intersection with the point cloud structure drift region;

[0221] For candidate point cloud registration propagation paths with spatial intersections, extract the corresponding spatial intersection positions between the candidate point cloud registration propagation paths and the point cloud structure drift regions;

[0222] The generation of spatial intersection positions specifically includes:

[0223] Read the spatial propagation range corresponding to the candidate point cloud registration propagation path and the spatial region range corresponding to the point cloud structure drift region; detect the spatial overlap between the spatial propagation range and the spatial region range; extract the spatial position corresponding to the spatial overlap region and generate the spatial intersection position;

[0224] The candidate point cloud registration propagation path and the point cloud structure drift region can only reflect the existence of spatial intersection, so the corresponding drift propagation position in the propagation path cannot be determined. In order to determine the corresponding drift propagation region in the candidate point cloud registration propagation path, it is necessary to extract the corresponding spatial intersection position between the candidate point cloud registration propagation path and the point cloud structure drift region.

[0225] The spatial intersection position is used to characterize the spatial region where the spatial propagation range corresponding to the candidate point cloud registration propagation path and the spatial region corresponding to the point cloud structure drift region overlap; it provides the basis for the path truncation position in the subsequent propagation path truncation process.

[0226] Based on the spatial intersection location, the propagation path of the candidate point cloud registration is truncated to separate the propagation path part that has spatial intersection with the point cloud structure drift region;

[0227] The specific steps of performing propagation path truncation include:

[0228] Based on the spatial intersection position, the corresponding cross-propagation part in the candidate point cloud registration propagation path is located; the spatial intersection position is used as the boundary position of the propagation path, and path separation processing is performed on the cross-propagation part to form the truncated candidate point cloud registration propagation path; the cross-propagation part is used to characterize the propagation path part in the candidate point cloud registration propagation path that enters the spatial region corresponding to the point cloud structure drift region.

[0229] When the candidate point cloud registration propagation path enters the point cloud structure drift region, the corresponding propagation path part cannot maintain a continuous and stable propagation state; by performing path separation processing on the cross propagation part that enters the point cloud structure drift region, the stable propagation part in the candidate point cloud registration propagation path is preserved.

[0230] The propagation path truncation process is used to remove the drift propagation part in the candidate point cloud registration propagation path, so as to improve the propagation stability of the candidate point cloud registration propagation path.

[0231] Based on the candidate point cloud registration propagation path after the propagation path truncation process, the spatial continuity between the remaining propagation paths is detected, and candidate point cloud registration propagation paths that cannot form continuous propagation paths are identified.

[0232] Read the spatial positions corresponding to the multiple remaining propagation path segments after the propagation path truncation process is completed; detect whether the spatial propagation connection state is maintained between the multiple remaining propagation path segments; identify the candidate point cloud registration propagation path that cannot maintain the spatial propagation connection state as the candidate point cloud registration propagation path that cannot form a continuous propagation path;

[0233] After the propagation path is truncated, the remaining propagation path parts in the candidate point cloud registration propagation path may be spatially separated; when the remaining propagation path parts cannot maintain a spatial propagation connection state, it means that the corresponding candidate point cloud registration propagation path can no longer form a continuous propagation path.

[0234] Spatial continuity is used to characterize the spatial propagation connection state between multiple remaining propagation path parts; candidate point cloud registration propagation paths that cannot form continuous propagation paths are used to characterize candidate propagation paths that cannot maintain a continuous propagation connection state between the remaining propagation path parts after propagation path truncation processing; identifying candidate point cloud registration propagation paths that cannot form continuous propagation paths is used to provide a basis for path continuity determination for subsequent stable propagation path screening.

[0235] Candidate point cloud registration propagation paths that cannot form a continuous propagation path are eliminated, and the remaining candidate point cloud registration propagation paths are determined as point cloud registration propagation paths.

[0236] In this embodiment, the generation of laser scanning point cloud registration results specifically includes:

[0237] Read the stable regions of the point cloud structure that constitute the point cloud registration propagation path and their corresponding spatial locations;

[0238] Calculate the spatial offset between adjacent stable point cloud structures based on their corresponding spatial locations.

[0239] Based on the spatial position offset and the propagation connection order corresponding to the point cloud registration propagation path, continuous spatial position alignment processing is performed on adjacent stable point cloud regions.

[0240] The propagation connection order corresponding to the point cloud registration propagation path is used to characterize the continuous propagation order between multiple stable point cloud structures; continuous spatial position alignment processing is performed on multiple stable point cloud structures according to the propagation connection order to maintain the continuous propagation correspondence between multiple stable point cloud structures.

[0241] Based on the spatial location corresponding to the point cloud structure drift region, detect whether there is a spatial intersection region between the stable region of the point cloud structure and the drift region of the point cloud structure after spatial alignment processing.

[0242] After spatial alignment, the stable region of the point cloud structure may enter the spatial region corresponding to the drift region of the point cloud structure; when there is a spatial intersection between the stable region of the point cloud structure and the drift region of the point cloud structure after spatial alignment, it indicates that the corresponding spatial alignment region has entered the drift region.

[0243] For stable regions of point cloud structures with spatial intersections, perform spatial position adjustment processing;

[0244] Perform spatial position adjustment processing to ensure that the stable region of the corresponding point cloud structure avoids the spatial region corresponding to the drift region of the point cloud structure.

[0245] Based on the stable regions of multiple point cloud structures after spatial position adjustment processing, laser scanning point cloud registration results are generated.

[0246] Example 1: To verify the feasibility of the present invention in practice, it was applied to the laser scanning point cloud registration scenario of a large underground utility tunnel. This underground utility tunnel is located in a coastal high-humidity environment and includes multiple complex spatial areas such as power channels, water supply channels, and communication channels. The overall spatial structure is distributed over a long distance. Due to the presence of numerous bending areas, equipment obstruction areas, and locally repetitive structural areas within the underground utility tunnel, it is easily affected by changes in environmental humidity, local spatial obstruction, and scanning angle during laser scanning at different times. This leads to problems such as point cloud structure drift, local spatial propagation breaks, and spatial registration offset between different scanning areas. Especially in some densely equipped areas, due to the high similarity between pipe edge structures and spatial texture structures, traditional point cloud registration methods are prone to misidentifying different spatial areas as the same structural area, resulting in local misalignment and propagation path interruption during the overall spatial alignment process. This seriously affects the continuity of the three-dimensional spatial reconstruction results of the underground utility tunnel and the stability of the spatial structure.

[0247] In practical applications, laser scanning equipment is first used to continuously scan the interior space of the underground utility tunnel, simultaneously acquiring scene image data, spatial target sphere position data, and environmental parameter data. Spatial target spheres are deployed in corner areas, equipment connection areas, and long-distance passage areas within the underground utility tunnel to form spatial reference nodes. During scanning, unified time synchronization processing is performed on the laser scanning point coordinate data, scene texture image data, and ambient temperature data in different scanning areas to generate spatial observation units for corresponding spatial areas, further forming a multi-source spatial observation dataset. Subsequently, based on the spatial distance relationships, scanning angle variation relationships, and image structure continuity relationships among multiple spatial observation units, spatial topological correlation analysis is performed on the spatial propagation state between different spatial observation units to establish continuous spatial propagation relationships among multiple spatial observation units. Due to the existence of long-distance continuous passages within the underground utility tunnel, a clear continuous propagation direction can be formed between different spatial areas. However, in areas partially obstructed by equipment, the spatial propagation direction will show significant shifts. By using the joint analysis method of propagation direction, propagation weight, and propagation disturbance parameters in this invention, propagation compensation updates can be performed on local propagation offset regions, thereby restoring the continuous propagation state between multiple spatial regions and forming a compensated spatial reference propagation field.

[0248] In densely equipped areas within underground utility tunnels, traditional point cloud registration methods often rely on local point cloud features for spatial alignment due to numerous structural repetitions and local edge occlusion. This can lead to mismatches in local structures. This invention addresses this by performing local structural correlation analysis on point cloud data, extracting spatial curvature, edge variation, and spatial continuity features. The local point cloud structure is then mapped to a compensated spatial reference propagation field. The propagation stability and structural offset values ​​of the corresponding local point cloud structures are calculated, thus dividing the underground utility tunnel into stable and drifting point cloud structures. The stable point cloud structure is formed by continuous wall areas, continuous pipe areas, and stable edge areas, while drifting point cloud structures are formed by locally occluded areas, equipment intersection areas, and areas with significant environmental disturbances. This approach effectively avoids interference from drifting point cloud structures in the overall spatial registration process.

[0249] Subsequently, this invention further combines the continuous features of image edges and the stable features of image texture in the scene image data to establish regional propagation connections between multiple stable point cloud structures, and forms point cloud registration propagation paths based on the regional propagation connections. After forming candidate point cloud registration propagation paths, this invention continues to detect the spatial intersection between the candidate point cloud registration propagation paths and the point cloud structure drift areas. When the candidate propagation path passes through a local device occlusion area or an area with obvious environmental disturbance, the corresponding propagation path will form a spatial intersection state with the point cloud structure drift area. At this time, this invention further extracts the corresponding spatial intersection position and performs path truncation processing on the cross propagation part in the propagation path, retaining only the stable propagation part that can maintain a continuous propagation state. For candidate propagation paths that cannot continue to form a continuous propagation state after the propagation path truncation processing is completed, path elimination processing is further performed, thereby finally retaining a stable and continuous point cloud registration propagation path. Through the above method, even if there are local structural drift areas and complex occlusion areas inside the underground integrated pipe gallery, this invention can still maintain the continuous and stable state of the overall spatial propagation path.

[0250] During continuous scanning of underground utility tunnels at different time periods, this invention further performs continuous spatial alignment processing on multiple stable point cloud structural regions based on the propagation connection sequence corresponding to the point cloud registration propagation path. During the spatial alignment process, it continuously detects whether there is spatial intersection between the corresponding spatial region and the point cloud structural drift region. When some stable regions enter the structural drift region during the spatial alignment process, this invention continues to perform position adjustment processing on the corresponding spatial position, thereby avoiding propagation interference of local structural drift regions on the overall spatial alignment result. After long-term continuous scanning and repeated spatial alignment verification between different scanning regions, the laser scanning point cloud registration result formed by this invention can maintain a high degree of spatial continuity and stability. Continuous and stable spatial connection relationships can be formed between the continuous wall area, pipe connection area, and equipment edge area inside the underground utility tunnel.

[0251] During practical application, multiple rounds of continuous scanning verification were performed on the area where the underground utility tunnel is located during the high humidity period in spring and the significant temperature difference period in summer. Repeated point cloud registration processing was performed on multiple continuous spatial regions within the underground utility tunnel on different dates, at different times, and in different scanning areas. The results show that the present invention can maintain high spatial propagation continuity and spatial registration stability even in areas with partial occlusion, structural repetition, and environmental disturbance. Compared with traditional methods that rely solely on local feature matching for spatial registration, the present invention can significantly reduce the probability of local spatial structural breaks, propagation path drift, and spatial alignment offsets, and effectively improve the continuity of point cloud structures and the reliability of spatial registration in complex spatial scenes.

[0252] Table 1. Comparison of overall registration performance of different point cloud registration methods in complex spatial scenes.

[0253] Spatial registration offset / m 1.84 1.37 1.12 0.68 Number of local structural fractures / times 17 13 10 4 Continuous propagation path retention rate / % 52.6 61.8 69.4 83.7 Number of mismatches in the drift area / times 14 11 9 3 Spatial structure continuity index 48.3 57.6 64.1 81.5 Point cloud edge region overlap rate / % 55.8 63.2 71.6 86.4 Registration completion rate in complex occluded areas / % 46.7 58.4 66.9 82.1 Multi-round scanning propagation stability index 51.2 60.7 68.3 84.6 Number of spatial path drifts / times 16 12 8 3 Point cloud overall registration time / min 43.8 39.6 36.2 31.4

[0254] As shown in Table 1, the traditional ICP registration method is easily affected by local structural repetition areas and spatial occlusion areas in complex spatial environments, with a spatial registration offset of 1.84m and 17 instances of local structural breakage. This indicates that the traditional method has poor propagation stability in continuous spatial areas, especially in underground integrated pipe corridors, long-distance passages, and complex equipment connection areas, where spatial propagation relationship interruptions and local structural misalignments are prone to occur. Although the feature matching-based point cloud registration method can reduce the spatial registration offset to some extent, it still mainly relies on local point cloud features to establish spatial correspondence. Therefore, the registration completion rate in complex occlusion areas is only 58.4%, and the number of mismatches in drift areas is still high, indicating that it lacks effective identification and propagation path filtering capabilities for structural drift areas.

[0255] Image-assisted point cloud registration methods improve the overlap of point cloud edge regions to 71.6% and the spatial structure continuity index to 64.1 by introducing image continuity features. This indicates that image continuity features can improve the structural correspondence in local spatial regions to a certain extent. However, this method still lacks analysis of the overall spatial propagation continuity. When there is environmental disturbance or structural drift in the local region, the number of spatial path drifts still reaches 8, and the multi-round scanning propagation stability index is only 68.3. This shows that existing image-assisted methods are still difficult to establish a continuous and stable overall spatial propagation path.

[0256] In contrast, this invention transforms the originally discrete spatial correspondence into a continuous spatial propagation relationship by constructing a set of spatial reference propagation relationships, a spatial distribution structure of propagation parameters, and a compensated spatial reference propagation field. Furthermore, it combines the stable region of the point cloud structure, the drift region of the point cloud structure, and the point cloud registration propagation path to perform joint analysis on the continuous propagation state in complex spatial regions. As shown in Table 1, the spatial registration offset of this invention is reduced to 0.68m, the number of local structural breaks is reduced to 4, and the continuous propagation path retention rate reaches 83.7%, indicating that this invention can effectively maintain the continuous propagation stability state in complex spatial scenes.

[0257] Meanwhile, this invention performs propagation path truncation and drift propagation filtering on point cloud structure drift regions, reducing the number of mismatches in drift regions to 3 times and the number of spatial path drifts to 3 times. This shows that this invention can effectively avoid the propagation interference of local drift regions on the overall spatial alignment process. In addition, this invention achieves a registration completion rate of 82.1% in complex occluded regions and a point cloud edge region overlap of 86.4%, indicating that this invention can not only maintain spatial propagation continuity, but also maintain high spatial structure correspondence accuracy in complex structural environments.

[0258] Further analysis of the overall point cloud registration time in Table 1 reveals that the overall registration time of this invention is reduced to 31.4 minutes. This is because the invention uses a point cloud structure stable region and propagation path filtering mechanism to eliminate a large number of unstable propagation regions and drift propagation paths in advance, reducing the number of repeated propagation corrections in the subsequent spatial alignment process, thereby improving the overall point cloud registration efficiency. Therefore, this invention can not only improve the continuity of point cloud structure and spatial registration stability in complex spatial scenes, but also improve the overall registration efficiency and propagation path reliability in complex spatial regions.

[0259] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A laser scanning point cloud registration method based on image analysis, characterized in that, Includes the following steps: Collect point cloud data, scene image data, spatial target sphere position data, and environmental parameter data, and preprocess them to generate a multi-source spatial observation dataset; Spatial topological correlation analysis is performed on the multi-source spatial observation dataset to map the distance correlation, scanning angle correlation and image structure correlation between spatial target spheres into propagation edges, generating a set of spatial reference propagation relationships; Propagation parameters are calculated on the set of spatial reference propagation relationships to obtain the propagation direction, propagation weight, and propagation perturbation parameters, and a compensated spatial reference propagation field is generated. Perform local structure correlation analysis on point cloud data to extract spatial curvature features, edge change features, and spatial continuity features, and generate a set of local structure features of point cloud. Map the local structure set of the point cloud to the compensated spatial reference propagation field, calculate the propagation stability value and the structure offset value, and generate the stable region and the drift region of the point cloud structure. Extract continuous features of image edges and stable features of image texture from scene image data, and generate point cloud registration propagation paths based on continuous features of image edges, stable features of image texture, and stable regions of point cloud structure; Spatial alignment processing is performed based on the point cloud registration propagation path and the point cloud structure drift region to generate laser scanning point cloud registration results.

2. The laser scanning point cloud registration method based on image analysis according to claim 1, characterized in that, The point cloud data includes laser scan point coordinate data, point cloud reflection intensity data, point cloud normal vector data, and timestamp data; the scene image data includes scene texture image data and scene edge image data; the spatial target sphere position data includes spatial target sphere center coordinate data, spatial target sphere spacing data, and spatial target sphere spatial distribution data; the environmental parameter data includes environmental temperature data, environmental humidity data, and environmental air pressure data. The preprocessing specifically includes performing noise point removal, outlier filtering, and coordinate unification on point cloud data; performing image denoising and image distortion correction on scene image data; and performing spatial coordinate calibration on spatial target sphere position data. The system performs time synchronization and abnormal data filtering on environmental parameter data; based on the spatial region corresponding to the spatial target sphere, it associates the local area of ​​the point cloud, the local area of ​​the scene image, and the environmental parameter data in the corresponding spatial region to form a spatial observation unit; it performs spatial organization processing on multiple spatial observation units to generate a multi-source spatial observation dataset.

3. The laser scanning point cloud registration method based on image analysis according to claim 1, characterized in that, The generation of the spatial reference propagation relation set specifically includes: Read multiple space observation units from the multi-source space observation dataset, extract the center coordinate data of the space target sphere and the spatial distribution data of the space target sphere corresponding to each space observation unit, and calculate the spatial distance relationship between the space observation units. Based on the spatial distance relationship, retrieve the coordinate data and timestamp data of the laser scanning points in the corresponding spatial observation unit, calculate the change in scanning direction and the change in scanning angle between adjacent spatial observation units, and generate scanning angle correlation results; Read scene texture image data and scene edge image data from adjacent spatial observation units, extract texture continuous regions and edge continuous regions between adjacent spatial observation units, and generate image structure association results; Based on spatial distance relationships, scanning angle correlation results, and image structure correlation results, topological organization processing is performed on the propagation correlation between multiple spatial observation units to generate a spatial topological correlation structure. Each association structure in the spatial topological association structure is mapped to a corresponding propagation edge, and propagation organization processing is performed on multiple propagation edges to generate a set of spatial reference propagation relationships.

4. The laser scanning point cloud registration method based on image analysis according to claim 1, characterized in that, The generation of the compensated spatial reference propagation field specifically includes: Read multiple propagation edges from the spatial reference propagation relationship set, extract the spatial observation unit position relationship, scanning angle association result and image structure association result corresponding to each propagation edge, and calculate the spatial propagation direction corresponding to each propagation edge; Around the spatial propagation direction corresponding to each propagation edge, the spatial distance relationship, scene texture image data and scene edge image data in the corresponding spatial observation unit are retrieved to calculate the propagation influence range corresponding to each propagation edge and generate propagation weights. Read the ambient temperature, ambient humidity and ambient air pressure data from the environmental parameter data, calculate the environmental changes and spatial disturbance changes corresponding to each space observation unit, and generate propagation disturbance parameters; Based on the propagation direction, propagation weight, and propagation perturbation parameters, spatial distribution organization processing is performed on the propagation parameter correlation relationship between multiple propagation edges to generate a spatial distribution structure of propagation parameters. Perform propagation compensation update processing on multiple propagation edges in the spatial distribution structure of propagation parameters, correct the propagation direction and the propagation offset in the propagation weight, and generate propagation compensation results; Based on the propagation compensation results, propagation field organization processing is performed on multiple propagation edges to generate a compensated spatial reference propagation field.

5. The laser scanning point cloud registration method based on image analysis according to claim 4, characterized in that, The generation of the spatial distribution structure of the propagation parameters specifically includes: Read the propagation direction, propagation weight, and propagation perturbation parameters corresponding to multiple propagation edges, and extract the spatial distance relationship corresponding to each propagation edge; Based on spatial distance relationships, spatial association organization processing is performed on multiple propagation edges to establish spatial association relationships between them; Based on the spatial relationship between propagation edges, perform directional continuity analysis on multiple propagation directions and calculate the directional change between adjacent propagation edges; Read the propagation weights and propagation perturbation parameters corresponding to multiple propagation edges, and calculate the changes in propagation weights and perturbation. Based on the changes in direction, propagation weight, and perturbation, propagation parameter space organization processing is performed on multiple propagation edges to generate propagation parameter space association results. Spatial distribution fusion processing is performed on multiple propagation edges in the spatial association result of propagation parameters to generate a spatial distribution structure of propagation parameters.

6. The laser scanning point cloud registration method based on image analysis according to claim 4, characterized in that, The specific steps of performing the propagation compensation update process include: Read multiple propagation edges in the spatial distribution structure of propagation parameters, and extract the propagation direction, propagation weight, and propagation perturbation parameters corresponding to each propagation edge; Around the propagation direction corresponding to multiple propagation edges, calculate the directional change between adjacent propagation edges and identify offset propagation edges with abnormal directional change. Based on the propagation perturbation parameters corresponding to the offset propagation edge, the propagation direction of the offset propagation edge is corrected by direction offset, and the propagation direction parameters corresponding to the offset propagation edge are updated. Around the propagation weights corresponding to multiple propagation edges, calculate the change in propagation weights between adjacent propagation edges, and identify abnormal propagation edges with abnormal changes in propagation weights; Based on the propagation perturbation parameters corresponding to the abnormal propagation edges, the propagation weights of the abnormal propagation edges are corrected, and the propagation weight parameters corresponding to the abnormal propagation edges are updated. Perform propagation continuity organization processing on multiple propagation edges after completing the direction offset correction and propagation weight correction to generate propagation compensation update results.

7. The laser scanning point cloud registration method based on image analysis according to claim 1, characterized in that, The generation of the stable region and the drift region of the point cloud structure specifically includes: Read multiple point cloud local structures from the point cloud local structure set, and extract the spatial curvature features, edge change features, and spatial continuity features corresponding to each point cloud local structure; Based on the spatial location corresponding to each point cloud local structure, read multiple propagation edges in the compensated spatial reference propagation field, as well as their corresponding propagation directions and propagation weights. Associate the propagation directions and propagation weights corresponding to the multiple propagation edges with the corresponding point cloud local structures to establish a propagation mapping relationship between the point cloud local structures and multiple propagation edges. Based on the propagation mapping relationship, calculate the directional change between the propagation directions corresponding to multiple propagation edges and the weight change between propagation weights, calculate the propagation continuity between multiple propagation edges, and generate the propagation stability value corresponding to the local structure of the point cloud. Read the spatial curvature features, edge change features, and spatial continuity features corresponding to multiple point cloud local structures, calculate the structural change between adjacent point cloud local structures, and generate structural offset values. Based on the propagation stability value and the structure offset value, structural stability identification is performed on multiple local structures of point clouds to generate stable local structures of point clouds and drifting local structures of point clouds. The stable point cloud local structure is processed to generate stable point cloud structure regions, and the drift point cloud local structure is processed to generate drift point cloud structure regions.

8. The laser scanning point cloud registration method based on image analysis according to claim 7, characterized in that, The specific implementation of stable region organization processing and drift region organization processing includes: Read the local structures of multiple stable point clouds and multiple drifting point clouds, and extract the spatial location and spatial continuity features corresponding to each local structure of the point cloud. Based on the spatial locations corresponding to multiple stable point cloud local structures, calculate the spatial distance and changes in spatial continuity features between adjacent stable point cloud local structures, and identify the connection relationships between stable point cloud local structures. Based on the connection relationship of local structures in stable point clouds, spatial aggregation processing is performed on multiple interconnected local structures of stable point clouds to expand the continuous connection range between local structures of stable point clouds, extract the region boundary corresponding to the continuous connection range, and generate stable regions of point cloud structures. Based on the spatial locations corresponding to multiple drift point cloud local structures, calculate the structural changes and spatial continuity feature changes between adjacent drift point cloud local structures, and identify the connection relationships of drift point cloud local structures. Based on the connection relationship of the local structure of the drifting point cloud, drift space aggregation processing is performed on multiple interconnected local structures of the drifting point cloud to expand the drift connection range between the local structures of the drifting point cloud, extract the region boundary corresponding to the drift connection range, and generate the drift region of the point cloud structure. Spatial location association processing is performed on stable and drifting regions of the point cloud structure to generate regional spatial distribution results.

9. The laser scanning point cloud registration method based on image analysis according to claim 1, characterized in that, The generation of the point cloud registration propagation path specifically includes: Read the continuous features of image edges and stable features of image texture in scene image data; By associating the spatial location of the stable region of the point cloud structure with the continuous features of the image edge and the stable features of the image texture in the corresponding image region, a propagation relationship between the stable region of the point cloud structure and the image region is established. Based on propagation correlation, identify the regional propagation connectivity between adjacent stable point cloud regions; Perform propagation path extension processing on multiple stable point cloud regions with regional propagation connection relationships to establish regional propagation path connection relationships between multiple stable point cloud regions; Based on the regional propagation path connectivity, candidate point cloud registration propagation paths are extracted between multiple point cloud structurally stable regions; Around multiple stable point cloud structures in the candidate point cloud registration propagation path, the spatial intersection between the candidate point cloud registration propagation path and the point cloud structure drift region is detected, and the candidate point cloud registration propagation path that has spatial intersection with the point cloud structure drift region is eliminated. The remaining candidate point cloud registration propagation paths are determined as the point cloud registration propagation paths.

10. The laser scanning point cloud registration method based on image analysis according to claim 9, characterized in that, The elimination specifically includes: Read multiple candidate point cloud registration propagation paths and point cloud structure drift regions, and extract the spatial location corresponding to each candidate point cloud registration propagation path. Based on the spatial location corresponding to the candidate point cloud registration propagation path, detect the spatial intersection between the candidate point cloud registration propagation path and the point cloud structure drift region, and identify the candidate point cloud registration propagation path that has spatial intersection with the point cloud structure drift region; For candidate point cloud registration propagation paths with spatial intersections, extract the corresponding spatial intersection positions between the candidate point cloud registration propagation paths and the point cloud structure drift regions; Based on the spatial intersection location, the propagation path of the candidate point cloud registration is truncated to separate the propagation path part that has spatial intersection with the point cloud structure drift region; Based on the candidate point cloud registration propagation path after the propagation path truncation process, the spatial continuity between the remaining propagation paths is detected, and candidate point cloud registration propagation paths that cannot form continuous propagation paths are identified. Candidate point cloud registration propagation paths that cannot form a continuous propagation path are eliminated, and the remaining candidate point cloud registration propagation paths are determined as point cloud registration propagation paths.

Images