A nonlinear registration optimization method for fastener point clouds

Through the nonlinear registration optimization method, combined with time, spatial continuity and local differences, the ellipsoid spatial diffusion technology is used to solve the problem of point cloud data deviation in fastener detection, and improve the accuracy and calculation efficiency of the data. Advanced understanding of fastener inspection.

CN118674756BActive Publication Date: 2024-11-12CHINA RAILWAY DESIGN GRP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411155366.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-22
Publication Date
2024-11-12
Estimated Expiration
2044-08-22

AI Technical Summary

Technical Problem

In the fastener detection, point cloud data is deviated due to environmental noise, hardware distortion, manual errors, etc., which affects the accuracy of fastener status evaluation and the timeliness of maintenance decisions. The existing point cloud optimization algorithm has a single optimization dimension. The degree of coupling between process and target is low and the time efficiency is low.

Method used

A nonlinear registration optimization method of fastener point cloud is adopted to obtain point cloud data containing spatial coordinates and timestamps through the acquisition device, perform data preprocessing, frame image conversion, registration, and streamlining, and use ellipsoid spatial diffusion to Achieve efficient and precise optimization of point cloud data, and combine time, spatial continuity and local differences to propose nonlinear target optimization equations.

Benefits of technology

It improves the accuracy and confidence of point cloud data, provides more reliable data input, avoids local optimization traps, has high computing efficiency and strong universality, and is suitable for subsequent point cloud processing and advanced understanding.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118674756B_ABST
    Figure CN118674756B_ABST
Patent Text Reader

Abstract

The present invention discloses a nonlinear registration optimization method for fastener point clouds, comprising the following steps: S1, original data acquisition; S2, data preprocessing; S3, converting the frame image into a deformation map; S4, registration between frame images; S5, streamlining the optimized point cloud. Aiming at the original point cloud quality problem that is currently neglected in fastener evaluation work, the present invention proposes a nonlinear registration optimization algorithm, which further improves the data accuracy and confidence of the original point cloud, provides more reliable data input for subsequent point cloud processing and high-level understanding processes, and thus ensures the realization rate of point cloud processing goals; the point cloud optimization process of the present invention has high robustness, retains the effective features of the original point cloud data as much as possible on the basis of eliminating noise and disordered points, and has strong detail retention. At the same time, the nonlinear optimization process avoids falling into the trap of local optimization, has high computational efficiency, strong universality, and high value for popularization and application.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of rapid comprehensive detection of rail transit, and in particular relates to a nonlinear registration optimization method for a fastener point cloud. Background Art

[0002] The supporting fasteners of the track are an important part of the entire rail transit project, and their status directly affects the operational safety of the train. After long-term operation and maintenance of rail transit lines, the service status of the fasteners will change, affecting the geographical spatial location and smoothness requirements of the line. With the development of high-speed railways and urban rail transit, the demand for detecting abnormal fastener conditions is increasing. However, in the fastener detection process, the point cloud data, which is the most critical input, often has a higher degree of deviation than the general redundant range due to environmental noise, hardware distortion, human errors, partial failure of fasteners, and complex conditions in the target area, which directly affects the accuracy of fastener status evaluation and the timeliness of maintenance decisions, thereby affecting railway driving safety. It is urgent to correct the data and optimize the features of the collected fastener point cloud.

[0003] There are three main types of existing targeted point cloud optimization algorithms:

[0004] (1) Point cloud optimization algorithm based on mathematical statistics concepts;

[0005] (2) Point cloud optimization algorithm based on features of each dimension;

[0006] (3) Point cloud optimization algorithm based on spatial domain state.

[0007] The above method has the following problems:

[0008] (1) The algorithm has a single optimization dimension, has a weak description coverage capability, and is prone to introducing additional errors or mistakes;

[0009] (2) The coupling degree between the algorithm process and the target point cloud is low, and the targeting is poor;

[0010] (3) The solution process is repetitive and wastes a lot of computing power, resulting in low time efficiency and heavy workload.

[0011] In response to the above problems and weaknesses, it is urgent to develop a nonlinear registration optimization technology for fastener point clouds, which can comprehensively, efficiently and accurately process and optimize the collected fastener point cloud data, and provide more reliable and trustworthy data input for subsequent point cloud understanding and advanced processing. Summary of the invention

[0012] In order to solve the problems existing in the prior art, the present invention provides a nonlinear registration optimization method for fastener point clouds, which can comprehensively, efficiently and accurately process and optimize the collected fastener point cloud data, and provide more reliable and trustworthy data input for subsequent point cloud understanding and advanced processing.

[0013] To this end, the present invention adopts the following technical solutions:

[0014] A nonlinear registration optimization method for a fastener point cloud comprises the following steps:

[0015] S1, raw data acquisition: using an acquisition device to acquire point cloud data containing spatial coordinates and timestamp information in an active axis, a depth axis and an offset axis; the active axis is a tangent direction of a curve along which the acquisition device moves, and the active axis, the depth axis and the offset axis are perpendicular to each other;

[0016] S2, data preprocessing: dividing the point cloud data collected in S1 into a plurality of frame images in the form of spatial slices according to the time axis according to the timestamp information;

[0017] S3, converting the frame image into a deformation image:

[0018] ;

[0019] ;

[0020] ;

[0021] in, For the The frame image of the frame is composed of point sets and edge sets composition; For the The total number of points in the frame image of the frame, For the The total number of edges in the frame image of the frame, is the frame image number, , is the total number of frame images;

[0022] S4, registration between frame images:

[0023] First, build the registration mapping relationship between frame images ,have:

[0024] ;

[0025] in: To align the description items; is the regularization description term; is the local transformation term; and is the indicator weight, and stated The value of depends on the degree of differentiation between the frame images. and The value range of is (0, 1),

[0026] Then according to the registration mapping relationship Obtaining an optimal registration relationship, and registering all frame images obtained in S2 according to the optimal registration relationship to obtain corrected point cloud data;

[0027] S5, simplifying the corrected point cloud data, including the following steps:

[0028] S51, establish the initial ellipsoid space:

[0029] First, a point in the corrected point cloud data obtained by S4 As the center of the ellipse in the ellipsoid space; the initial value of the length of the major axis of the ellipse in the ellipsoid space is twice the distance between the center of the ellipse and the nearest neighbor point projected on the offset axis, and the initial value of the length of the minor axis of the ellipse in the ellipsoid space is twice the distance between the center of the ellipse and the nearest neighbor point projected on the depth axis;

[0030] The nearest neighbor point is the point in the corrected point cloud data with the smallest Euclidean distance from the center of the ellipse;

[0031] Then, the direction of the major axis of the ellipse is parallel to the offset axis, the direction of the minor axis is parallel to the depth axis, and the two end points of the major axis of the ellipse are respectively and The two end points of the minor axis of the ellipse are and ; Get the initial ellipsoid space;

[0032] Finally, construct an ellipsoid space set and add the initial ellipsoid space as the first element;

[0033] S52, ellipsoidal spatial diffusion, comprises the following steps:

[0034] Adding or reconstructing the ellipsoid space in the ellipsoid space concentration, and simplifying the S4 corrected point cloud data covered by the ellipsoid space;

[0035] S53, repeating step S52 until the ellipsoid space in the ellipsoid space set covers all the point cloud data corrected in S4, and outputting the obtained point cloud as high-quality point cloud data after nonlinear registration.

[0036] Alignment description items as described in S4 Calculated by:

[0037] ;

[0038] ;

[0039] in, for function; For the The points within the frame image of the frame, For the The number of the point in the frame image of the frame; ; yes The nearest target point in the frame to be registered is set as the first frame, ; is greater than 0 and makes the A monotonically increasing constant on the positive semi-axis. The regularization term described in S4 Calculated by:

[0040] ;

[0041] in, for The updated constant term in the function, is greater than 0 and makes the A monotonically increasing constant on the positive semi-axis; Indicates the midpoint of the frame image to be registered The set of adjacent points of For the The number of the point in the frame image of the frame, , For the The total number of points in the frame image of the frame; is the transformation form of each point in the deformation graph taking into account spatial cross-coupling.

[0042] Said Calculated by:

[0043] ;

[0044] in, is the affine rotation matrix of the embedded deformation graph; and Respectively Frame and Affine translation vector of the embedded deformation map corresponding to the frame.

[0045] The local transformation term Calculated by:

[0046] ;

[0047] in, is a modification factor that takes into account the severity of local transformations, and its value range is (0, 1); the operator symbol After considering the differential local rigid state Norm solution.

[0048] S52 specifically includes the following steps: S521, in the ellipsoid space set, for each ellipsoid space, a point is randomly determined outside the ellipsoid space. , is the point in the corrected point cloud data obtained in S4. For each point , when it satisfies the following formula, execute S522, otherwise execute S523,

[0049] ;

[0050] in, is the normal vector of the plane formed by the major axis and the minor axis of the ellipse and ; For point The distance threshold to the projection surface, is the vector deflection threshold; and The value of is determined according to the fastener to be registered, and the projection surface is The center of the ellipse in the corresponding ellipsoid space The plane where the corresponding frame is located;

[0051] S522, Reconstruct the ellipsoid space:

[0052] Point Integrate into its corresponding ellipsoid space, and the center of the ellipse remains unchanged, and the length of the major axis of the ellipse in the corresponding ellipsoid space is changed to , the length of the minor axis of the ellipse is changed to , we get the reconstructed ellipsoid space:

[0053] ;

[0054] in, is the endpoint of the major axis of the ellipsoid space with centered at . For is the endpoint of the minor axis of the ellipsoid space with the center of the circle;

[0055] Change all point cloud data covered in the reconstructed ellipsoid space to the ellipse center point , the two endpoints of the major axis of the reconstructed ellipsoid space, the two endpoints of the minor axis of the reconstructed ellipsoid space and the ;

[0056] Execute step S53;

[0057] S523, add ellipsoid space:

[0058] First, according to the above point With the center of the ellipse The diffusion direction of the ellipsoid space is determined by the position of:

[0059] Then, the point is determined in the diffusion direction The nearest neighbor of ; An additional ellipsoid space is constructed as the ellipse center of the ellipsoid space, and the length of the major axis of the ellipse of the additional ellipsoid space is the point The length of the major axis projection of the ellipse in the additional ellipsoid space; the length of the minor axis of the ellipse in the additional ellipsoid space is the length of the minor axis of the ellipse at the point The length of the ellipse minor axis projection of the additional ellipsoid space retains all point cloud data in the additional ellipsoid space;

[0060] The additional ellipsoid space is added to the ellipsoid space set, and step S53 is executed.

[0061] The diffusion direction in S523 is determined by the following formula:

[0062] ;

[0063] in, Indicate point The quadrant in which it is located.

[0064] Preferably, the acquisition device in S1 is a self-moving rail transit mobile three-dimensional scanning system.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] 1. Aiming at the problem of original point cloud quality which is often neglected in the current fastener evaluation work, the present invention proposes a nonlinear registration optimization algorithm. The data correction process of the point cloud is quantified into three types of description items, which further improves the data accuracy and confidence of the original point cloud, provides more reliable data input for subsequent point cloud processing and advanced understanding process, and thus ensures the realization rate of point cloud processing goals;

[0067] 2. The method of the present invention combines the temporal and spatial continuity of the collected point cloud data, and focuses on the local difference and global unity, the rigid distinction of details and the overall non-rigid composition, proposes a nonlinear target optimization equation, and uses three types of descriptive terms to further improve the accuracy, coverage and confidence of the optimization process;

[0068] 3. The method of the present invention designs an ellipsoid inclusion space, realizes point cloud simplification and optimization in the form of elliptical diffusion, further improves the availability, simplicity and descriptiveness of point cloud data, and meets the needs of subsequent point cloud processing and advanced understanding processes; the point cloud optimization process of the present invention is highly robust, and retains the effective features of the original point cloud data as much as possible on the basis of removing noise and disordered points, and has strong detail retention. At the same time, the nonlinear optimization process avoids falling into the trap of local optimization, has high computational efficiency, strong universality, and high value of popularization and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 is a flow chart of the present invention;

[0070] Figure 2 It is a schematic diagram of point cloud data composed of each subdivided frame data in the data preprocessing process of the present invention;

[0071] Figure 3 It is a schematic diagram of the ellipsoid space set of the present invention. DETAILED DESCRIPTION

[0072] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings and embodiments.

[0073] During the acquisition process of digital point cloud, the point cloud coordinate system is defined. The point cloud coordinate system includes three coordinate axes, namely the depth axis, the offset axis and the active axis. The active axis is the tangent direction of the curve of the acquisition system movement, the depth axis and the offset axis are perpendicular to the active axis, and the depth axis and the offset axis are also in a vertical relationship.

[0074] like Figure 1 As shown, the nonlinear registration optimization method of the fastener point cloud of the present invention comprises the following steps:

[0075] S1, original data acquisition: use a self-mobile rail transit mobile 3D scanning system, or use other acquisition systems and equipment to collect and obtain the point cloud data containing spatial coordinates and timestamp information in the aforementioned point cloud coordinate system.

[0076] S2, data preprocessing: According to the timestamp information obtained in S1, the complete point cloud is divided into several frame images along the time axis in the form of spatial slices, and the plane composed of the depth axis and the offset axis is as follows Figure 2 The plane of the point cloud slice shown;

[0077] S3, converting the frame image into a deformation graph: introducing an embedded deformation graph in graph theory into the frame image, that is, Frame image of frame Described as a point set and edge sets The collection composed of , Respectively The total number of points and edges of the frame image of the frame, is the frame image number, , is the total number of frame images, then:

[0078] ;

[0079] ;

[0080] ;

[0081] S4, registration between frame images: the frame images divided by S2 are registered with a nonlinear mapping relationship. Considering the spatial continuity and subtle distinction of the apparent features during the data acquisition process, a registration process with a higher difference tolerance and an optimization goal should be considered, including the following steps:

[0082] S41, establish the registration mapping equation: in Frame and Establish a registration mapping relationship between frame images , No. Frame No. The frames to be registered are:

[0083] ;

[0084] in, To align the description items; is the regularization description term; is the local transformation term; and is the indicator weight, which is determined by the degree of differentiation between frame images. and The value range of is (0, 1), ;

[0085] S42, calculate alignment description items ,have:

[0086] ;

[0087] ;

[0088] in, for function; For the The points within the frame image of the frame, y For the The number of the point in the frame image of the frame; ; yes The nearest target point in the frame image to be registered; is greater than 0 and makes A monotonically increasing constant on the positive semi-axis;

[0089] S43, calculate the regularization description term ,have:

[0090] ;

[0091] in, for The updated constant term in the function, is greater than 0 and makes A monotonically increasing constant on the positive semi-axis; Indicates the midpoint of the frame image to be registered The set of adjacent points of For the The number of the point in the frame image of the frame, , For the The total number of points in the frame image of the frame; is the transformation form of each point in the deformation graph considering spatial cross coupling;

[0092] ;

[0093] in, is the affine rotation matrix of the embedded deformation graph; and Respectively Frame and Affine translation vector of the embedded deformation map corresponding to the frame.

[0094] S44, calculate local transformation terms ,have:

[0095] ;

[0096] in, is a modification factor that takes into account the severity of local transformations, and its value range is (0, 1); the operator symbol After considering the differential local rigid state Norm solving;

[0097] S45, nonlinear optimization solution: according to the registration mapping relationship of step S41 , solve the best registration relationship between all the frame images to be registered of the fastener point cloud with the highest confidence, and correct the positions of adjacent frame images based on this relationship; obtain the corrected point cloud data.

[0098] S5, simplifying the corrected point cloud data: After the nonlinear optimization of the fastener point cloud as described in step S4, additional point cloud topological structures, point-line-edge relationships, and possible spatial outliers, data jitter, local burrs, etc. will be introduced, affecting the accuracy of subsequent point cloud processing. Therefore, it is necessary to use the ellipsoid space to simplify the point cloud results processed by S4, convert the spatial dimension optimization process of the point cloud data into a growth process in which multiple base spaces grow into feature inclusion spaces, and perform generalized elliptical parameter diffusion, including the following sub-steps:

[0099] S51, establish the initial ellipsoid space: take a point in the point cloud data corrected in S4 as As the center of the ellipse in the ellipsoid space; the initial values ​​of the lengths of the major and minor axes of the ellipse in the ellipsoid space are respectively twice the distance between the center of the ellipse and the nearest neighbor point on the offset axis and the depth axis projection, and the nearest neighbor point is the point in the point cloud data with the smallest Euclidean distance from the center of the ellipse, so that the major axis direction of the ellipse in the ellipsoid space is parallel to the offset axis, and the minor axis direction is parallel to the active axis, and the two end points of the major axis of the ellipsoid space are respectively and ; The two end points of the minor axis of the ellipsoid space are and ; Get the initial ellipsoid space.

[0100] Construct an ellipsoid space set and add the initial ellipsoid space as the first element.

[0101] S52, ellipsoidal spatial diffusion, comprises the following steps:

[0102] S521, randomly determine a point in the point cloud data processed by S4 ,and It is a point outside the ellipsoid space in the ellipsoid space concentration. For each point , when it satisfies the following formula, execute S522, otherwise execute S523,

[0103] ;

[0104] in, is the normal vector to the plane formed by the major and minor axes of the ellipse and ; for The distance threshold to the projection surface, is the vector deflection threshold; and The value is determined according to the fastener to be registered, and the projection surface is The center of the ellipse in the corresponding ellipsoid space The plane where the corresponding frame is located; The center of the ellipse A point outside the ellipsoid space. If there are multiple ellipsoid spaces in the ellipsoid space concentration, then a point outside each ellipsoid space must be taken. , so each There are corresponding ellipse centers .

[0105] S522, Reconstruct the ellipsoid space:

[0106] Point Integrate into its corresponding ellipsoid space, and the center of the ellipse Remain unchanged, and the lengths of the major and minor axes of the corresponding ellipsoid space are changed to points The length of the projection on the major axis of the ellipse and the length of the projection on the minor axis , and obtain the reconstructed ellipsoid space; among them,

[0107] ;

[0108] for The endpoint of the major axis of the ellipsoid space with the center of the circle, For is the endpoint of the minor axis of the ellipsoid space with the center of the circle;

[0109] For the reconstructed ellipsoid space, Located on its surface, all point cloud data covered in the reconstructed ellipsoid space are changed to the center point of the ellipse , the two endpoints of the major axis of the reconstructed ellipsoid space, the two endpoints of the minor axis of the reconstructed ellipsoid space and .

[0110] Execute step S53;

[0111] S523, add ellipsoid space:

[0112] According to the point The diffusion direction of the ellipsoid space is determined by the position of:

[0113] ;

[0114] in, Indicate point The quadrant in which it is located.

[0115] Determine in the obtained diffusion direction The nearest neighbor of ;Will An additional ellipsoid space is constructed as the center of the ellipse in the ellipsoid space, and the lengths of the major axis and the minor axis of the ellipse in the additional ellipsoid space are the points All point cloud data in the additional ellipsoid space are retained at the length of the major axis and minor axis projections of the additional ellipsoid space;

[0116] The additional ellipsoid space is added to the ellipsoid space set, and step S53 is executed.

[0117] S53, repeating step S52 until the ellipsoid space in the ellipsoid space set covers all the point cloud data corrected in S4, and outputting the obtained point cloud as high-quality point cloud data after nonlinear registration. Figure 3 shown.

[0118] The method of the present invention realizes the registration and alignment of local details of the collected point cloud and the averaging of data errors by establishing three types of description items from the perspective of global optimization. On the one hand, the collected point cloud is divided into multiple frame point clouds, which assist each other in modification, avoiding the blurring of local details and the appearance of data holes, and improving the data accuracy of the collected point cloud; on the other hand, the point cloud registration process guided by the nonlinear optimization idea realizes the error averaging between multiple frame point cloud data, and improves the data confidence of the collected point cloud.

[0119] In the process of nonlinear optimization, the collected point cloud is regarded as a complex rigid body set of finite subdivisions composed of multiple simple structural rigid bodies by referring to the spatiotemporal continuity of the point cloud data during the acquisition process. Therefore, in view of the two major contradictions in the quality of point clouds, namely, local differences and global unity, and detail rigid distinction and overall non-rigid composition, the optimization process focuses on the following issues: (1) poor tolerance of data differences introduced by measurement errors, (2) low resolution of gradient descent introduced by a single optimization perspective, and (3) weak spatial structure similarity introduced by point cloud sparsity. The items to be fitted in the nonlinear optimization objective equation are further deconstructed and quantified into three types of descriptive items, which respectively improve the accuracy, coverage and confidence indicators in the optimization process.

[0120] During the optimization process, this algorithm establishes a deep coupling relationship between the collected point cloud and the target point cloud through the three additionally designed descriptive items, and ensures the robustness of the algorithm with an independent optimization structure. At the same time, the algorithm retains the effective features of the collected point cloud data as much as possible on the basis of eliminating noise and disordered points, and has strong detail retention. In addition, during the iterative optimization process of the algorithm, the traditional single equation gradient descent process is parsed into a three-way calculation process, and some intermediate results can be reused, further improving the calculation efficiency and avoiding falling into the local optimization trap. The above processes are all feature representation and optimization descriptions within the point cloud. There are no additional requirements and restrictions on the basic properties and data distribution of the point cloud. Therefore, they are highly universal and can be further extended to non-rigid target registration scenarios to achieve human body reconstruction, high-resolution reconstruction, high-detail virtual reality, etc.

Claims

1. A nonlinear registration optimization method for fastener point clouds, characterized in that: The following steps are involved: S1, raw data acquisition: using an acquisition device to acquire point cloud data containing spatial coordinates and timestamp information in an active axis, a depth axis and an offset axis; the active axis is a tangent direction of a curve along which the acquisition device moves, and the active axis, the depth axis and the offset axis are perpendicular to each other; S2, data preprocessing: dividing the point cloud data collected in S1 into a plurality of frame images in the form of spatial slices according to the time axis according to the timestamp information; S3, converting the frame image into a deformation image: ; ; ; in, For the The frame image of the frame is composed of point sets and edge sets composition; For the The total number of points in the frame image of the frame, For the The total number of edges in the frame image of the frame, is the frame image number, , is the total number of frame images; S4, registration between frame images: First, build the registration mapping relationship between frame images ,have: ; in: To align the description items; is the regularization description term; is the local transformation term; and is the indicator weight, and stated The value of depends on the degree of differentiation between the frame images. and The value range of is (0, 1), Then according to the registration mapping relationship Obtaining an optimal registration relationship, and registering all frame images obtained in S2 according to the optimal registration relationship to obtain corrected point cloud data; S5, simplifying the corrected point cloud data, including the following steps: S51, establish the initial ellipsoid space: First, a point in the corrected point cloud data obtained by S4 As the center of the ellipse in the ellipsoid space; the initial value of the length of the major axis of the ellipse in the ellipsoid space is twice the distance between the center of the ellipse and the nearest neighbor point projected on the offset axis, and the initial value of the length of the minor axis of the ellipse in the ellipsoid space is twice the distance between the center of the ellipse and the nearest neighbor point projected on the depth axis; The nearest neighbor point is the point in the corrected point cloud data with the smallest Euclidean distance from the center of the ellipse; Then, the direction of the major axis of the ellipse is parallel to the offset axis, the direction of the minor axis is parallel to the depth axis, and the two end points of the major axis of the ellipse are respectively and The two end points of the minor axis of the ellipse are and ; Get the initial ellipsoid space; Finally, construct an ellipsoid space set and add the initial ellipsoid space as the first element; S52, ellipsoidal spatial diffusion, comprises the following steps: Adding or reconstructing the ellipsoid space in the ellipsoid space concentration, and simplifying the S4 corrected point cloud data covered by the ellipsoid space; S53, repeating step S52 until the ellipsoid space in the ellipsoid space set covers all the point cloud data corrected in S4, and outputting the obtained point cloud as high-quality point cloud data after nonlinear registration.

2. The nonlinear registration optimization method of fastener point cloud according to claim 1, characterized in that: Alignment description items as described in S4 Calculated by: ; ; in, for function; For the The points within the frame image of the frame, For the The number of the point in the frame image of the frame; ; yes The nearest target point in the frame to be registered is set as the first frame, ; is greater than 0 and makes the A monotonically increasing constant on the positive semi-axis.

3. The nonlinear registration optimization method of fastener point cloud according to claim 2, characterized in that: The regularization descriptor described in S4 Calculated by: ; in, for The updated constant term in the function, is greater than 0 and makes the A monotonically increasing constant on the positive semi-axis; Indicates the midpoint of the frame image to be registered The set of adjacent points of For the The number of the point in the frame image of the frame, , For the The total number of points in the frame image of the frame; is the transformation form of each point in the deformation graph taking into account spatial cross-coupling.

4. The nonlinear registration optimization method of fastener point cloud according to claim 3, characterized in that: Said Calculated by: ; in, is the affine rotation matrix of the embedded deformation graph; and Respectively Frame and Affine translation vector of the embedded deformation map corresponding to the frame.

5. The nonlinear registration optimization method of fastener point cloud according to claim 4, characterized in that: The local transformation term Calculated by: ; in, is a modification factor that takes into account the severity of local transformations, and its value range is (0, 1); the operator symbol After considering the differential local rigid state Norm solution.

6. The nonlinear registration optimization method of fastener point cloud according to claim 1, characterized in that: S52 specifically includes the following steps: S521, in the ellipsoid space set, for each ellipsoid space, a point is randomly determined outside the ellipsoid space. , is the point in the corrected point cloud data obtained in S4. For each point , when it satisfies the following formula, execute S522, otherwise execute S523, ; in, is the normal vector of the plane formed by the major axis and the minor axis of the ellipse and ; For point The distance threshold to the projection surface, is the vector deflection threshold; and The value of is determined according to the fastener to be registered, and the projection surface is The center of the ellipse in the corresponding ellipsoid space The plane where the corresponding frame is located; S522, Reconstruct the ellipsoid space: Point Integrate into its corresponding ellipsoid space, and the center of the ellipse remains unchanged, and the length of the major axis of the ellipse in the corresponding ellipsoid space is changed to , the length of the minor axis of the ellipse is changed to , we get the reconstructed ellipsoid space: ; in, is the endpoint of the major axis of the ellipsoid space with centered at . For is the endpoint of the minor axis of the ellipsoid space with the center of the circle; Change all point cloud data covered in the reconstructed ellipsoid space to the ellipse center point , the two endpoints of the major axis of the reconstructed ellipsoid space, the two endpoints of the minor axis of the reconstructed ellipsoid space and the ; Execute step S53; S523, add ellipsoid space: First, according to the above point With the center of the ellipse The diffusion direction of the ellipsoid space is determined by the position of: Then, the point is determined in the diffusion direction The nearest neighbor of ; An additional ellipsoid space is constructed as the ellipse center of the ellipsoid space, and the length of the major axis of the ellipse of the additional ellipsoid space is the point The length of the major axis projection of the ellipse in the additional ellipsoid space; the length of the minor axis of the ellipse in the additional ellipsoid space is the length of the minor axis of the ellipse at the point The length of the ellipse minor axis projection of the additional ellipsoid space retains all point cloud data in the additional ellipsoid space; The additional ellipsoid space is added to the ellipsoid space set, and step S53 is executed.

7. The nonlinear registration optimization method of fastener point cloud according to claim 6, characterized in that: The diffusion direction in S523 is determined by the following formula: ; in, Indicate point The quadrant in which it is located.

8. The nonlinear registration optimization method of fastener point cloud according to claim 1, characterized in that: The acquisition device in S1 is a self-moving rail transit mobile three-dimensional scanning system.