A non-rigid registration method for tunnel deformation steel arch

Abstract

The application provides a non-rigid registration method for a tunnel-shaped deformed steel arch, and belongs to the cross field of point cloud processing technology and tunnel engineering. The non-rigid registration method mainly comprises three stages: (1) point cloud pretreatment; (2) steel arch space trend characteristic curve extraction; (3) point cloud non-rigid registration based on a space curve. Firstly, according to the distribution law of the nearest point polar radius, the separation of the rock wall part and the non-rock wall part, the steel arch part and the noise in the original scanning point cloud is realized in turn by using an iterative method; then the space trend characteristic curve extraction is carried out on the steel arch design model and the steel arch point cloud obtained after the point cloud pretreatment; finally, based on the characteristic curve, the non-rigid registration of the design model and the steel arch point cloud is carried out, and the recovery of the steel arch in the tunnel is realized. The application can realize efficient and accurate recovery of the steel arch for the stress-torsion deformed steel arch in the tunnel, provides technical support for the information perception of the tunnel rock wall support by intelligent equipment, and provides parameter basis for realizing automatic wet spraying operation.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of point cloud processing technology and tunnel engineering, and in particular relates to a non-rigid registration method for deformable steel arch frames in tunnels. Background Technology

[0002] During tunnel construction, tunnel excavation is typically carried out using methods such as drill-and-blast, shield tunneling, or a combination of both. Because shield tunneling has drawbacks such as high requirements for geological conditions and poor adaptability to cross-sectional changes, it generally lacks flexibility and adaptability in construction. Therefore, when tunnel construction faces extreme geological structures such as frequent changes in lithology, uneven hardness, and numerous faults, drill-and-blast methods are often necessary.

[0003] After excavation using the drill-and-blast method, reinforced concrete lining is required. Initial support for the tunnel is achieved through shotcreting and the use of steel arch supports, preventing surrounding rock collapse. In traditional drill-and-blast construction, wet shotcreting requires on-site operation by construction personnel. However, harsh environmental factors in the tunnel environment, such as rock bursts, large deformations, high temperatures and humidity, and inadequate ventilation, can all affect the construction process, leading to low efficiency, poor safety, and difficulty in ensuring work quality.

[0004] By developing new automated construction equipment, we can overcome the shortcomings of traditional drilling and blasting methods, such as insufficient mechanization, reduce the labor intensity of workers, and ensure their safety. This will enable safe, efficient, automated, and mechanized tunnel construction under various geological challenges, which is of great significance to the development of tunnel engineering technology.

[0005] During automated wet spraying operations, areas in the tunnel rock walls obscured by steel arches cannot be directly sprayed. Therefore, it is necessary to plan the path of the wet spraying machine's robotic arm so that it bypasses the steel arches within the tunnel, thus filling the gaps between the steel arches and the rock walls.

[0006] To guide the robotic arm's path planning, the steel arch frame point cloud extracted from the blasting point cloud needs to be reconstructed. This enables the intelligent equipment to perceive the position information of the steel arch frame, providing parameter support for the wet spraying machine's robotic arm path planning. Ultimately, this allows the intelligent equipment to replace on-site construction personnel in guiding the wet spraying machine to perform automated wet spraying operations. Simultaneously, the reconstructed steel arch frame results can also be used to remove the volume occupied by the steel arch frame in the blocks to be filled in the concrete spraying volume calculation results, thereby improving the accuracy of the concrete spraying volume calculation.

[0007] Due to the limitations of laser scanning equipment accuracy in actual engineering projects, the acquisition of information about the actual steel arch frame is insufficient. The local features of the scanned point cloud cannot be fully reflected in the original blasting point cloud file exported by the equipment, making it impossible to directly reconstruct the steel arch frame from the actual point cloud. By using non-rigid registration technology on the steel arch frame point cloud, and based on the registration results between the denoised steel arch frame design model and the steel arch frame point cloud, the steel arch frame design model is matched to the reconstructed rock wall model. This allows for the effective reconstruction of the tortuous and deformed steel arch frame within the tunnel, thus solving the problem of steel arch frame reconstruction. Summary of the Invention

[0008] The technical problem to be solved by this invention is to provide a method for restoring a steel arch frame subjected to torsional deformation under stress within a tunnel. To achieve the above technical objective, the technical solution adopted by this invention includes two parts: point cloud preprocessing and non-rigid registration. The point cloud preprocessing is the preprocessing part of the non-rigid registration method, which is further divided into two stages: extraction of the spatial trend characteristic curve of the steel arch frame and non-rigid registration of the point cloud based on the spatial curve. An embodiment provides a non-rigid registration method for a deformed steel arch frame in a tunnel.

[0009] Step 1, Point Cloud Preprocessing: Separate the rock walls from the non-rock walls, and the steel arches from the noisy parts in the original scanned point cloud;

[0010] Step 2: Curve extraction of the design model: Ignore the local features of the steel arch frame design model and retain only its overall trend. Extract and construct the spatial trend feature curve of the design model to approximate the steel arch frame design model.

[0011] Step 3: Curve extraction of steel arch point cloud: Ignore the local features of the steel arch point cloud and retain only its overall trend. Extract and construct the spatial trend feature curve of the steel arch point cloud to approximate the steel arch point cloud.

[0012] Step 4: Point cloud registration based on spatial curves: By matching the extracted spatial trend feature curves, non-rigid registration between the steel arch frame design model and the steel arch frame point cloud is further achieved.

[0013] The non-rigid registration method for deformable steel arch frames in tunnels described in this invention, in a preferred embodiment, further includes the following steps in step 1:

[0014] Step 1.1, Coordinate Transformation and Coordinate System Conversion: Perform coordinate translation, coordinate rotation and coordinate system transformation on the original scanned point cloud in sequence to obtain the scanned point cloud after coordinate system transformation;

[0015] Step 1.2, Denoising the blasting point cloud: Separate the real rock wall part and the steel arch noise part in the scanned point cloud after the coordinate system transformation to obtain the denoised blasting point cloud and steel arch noise point cloud;

[0016] Step 1.3, Denoising the steel arch point cloud: Separate the steel arch portion from the noise portion in the non-rock wall portion of the steel arch noise point cloud to obtain the denoised steel arch point cloud, thus completing the point cloud preprocessing.

[0017] The coordinate transformation and coordinate system conversion consist of coordinate translation, coordinate rotation, and coordinate system conversion, ensuring that the centroid of the original scanned point cloud is located at the origin of the coordinate system and the tunnel's central axis is parallel to the x-axis. Finally, the coordinate values ​​are converted from a rectangular coordinate system... Convert to cylindrical coordinate system And the scanned point cloud in the cylindrical coordinate system is... The two-dimensional discrete function shown is used to represent the point cloud after coordinate system transformation.

[0018] The denoising of the explosive point cloud is based on iterative denoising according to the change in polar radius before and after weighting of nearest neighbor points, specifically including the following steps:

[0019] First, construct a point cloud of the hypothetical rock face: traverse each point in the scanned point cloud after coordinate system transformation, and calculate the hypothetical rock face radius corresponding to the current point based on its nearest neighbor information using a weighted average of the nearest neighbor points. The scanned point cloud after coordinate system transformation following the polar radius change is defined as the hypothetical rock wall point cloud. Specifically,

[0020]

[0021] in For the number of nearest neighbors, For point Its nearest neighbor The weight, specifically,

[0022]

[0023] The first part of the multiplication expression on the right side of the equation depends on the point. Its nearest neighbor Perpendicular to The Euclidean distance of the projection onto the plane of direction, the second part depends on the point. Its nearest neighbor At the polar radius Difference in direction;

[0024] Next, calculate the polar radius transformation: traverse each point in the scanned point cloud after the coordinate system transformation, and use the hypothetical rock wall polar radius of the current point. Replace the original extreme diameter r value and calculate the extreme diameter of the hypothetical rock wall. The difference between the polar radius and the original polar radius r value is defined as the change in polar radius of each point in the scanned point cloud after the coordinate system transformation. ;

[0025] Finally, iterative removal of non-rock wall point clouds: An iterative method is used to separate the real rock wall portion from the noisy steel arch portion in the scanned point cloud.

[0026] In some embodiments, the iterative removal of non-rock wall point clouds is performed on the scanned point cloud after coordinate system transformation based on the polar radius change. The numerical values ​​are iteratively denoised, specifically including the following steps: (a) using the scanned point cloud after coordinate system transformation as the initial input point cloud; (b) calculating the total number of points in the input point cloud and adjusting the polar radius change of the input point cloud. The mean, variance, and standard deviation are calculated sequentially; (c) the polar radius variation is calculated for each point in the input point cloud. Points whose difference from the mean exceeds a set multiple of the standard deviation are marked and removed from the input point cloud; (d) the ratio of the number of marked points to the total number of points in the input point cloud is calculated; (e) when the ratio is greater than a set value, the input point cloud after removing the marked points is used as a new input point cloud, and (b) is repeated; (f) when the ratio is less than a set value, the input point cloud after removing the marked points is used as the output point cloud, and the iterative denoising of the scanned point cloud after the coordinate system transformation is completed. The output point cloud is defined as the denoised explosion point cloud, and the difference between the initial input point cloud and the output point cloud is defined as the steel arch noise point cloud.

[0027] The point cloud denoising of the steel arch frame is based on the equidistant segmentation of the polar angle, followed by iterative denoising based on the magnitude of the mode of the polar radius. Specifically, it includes the following steps:

[0028] First, the point cloud is mapped: the noise point cloud of the steel arch frame is transferred from the cylindrical coordinate system. Mapped to a three-dimensional Cartesian coordinate system The two-dimensional plane below In the process, the planar point set after mapping the noise point cloud of the steel arch frame is obtained;

[0029] Next, the point cloud is grouped: the planar point set after mapping the steel arch noise point cloud is grouped according to the polar angle. The values ​​are divided into equal segments to obtain multiple sets of local point clouds;

[0030] Finally, iterative noise removal is performed: using an iterative method, patch noise is removed from each group of local point clouds to obtain the denoised steel arch point cloud.

[0031] In some embodiments, the iterative noise removal involves removing patchy noise from each group of local point clouds based on the density of noise distribution along the polar radius r. Specifically, this includes the following steps: (a) traversing each group of local point clouds; (b) using the local point cloud as the initial input point cloud; and (c) calculating the total number of points in the input point cloud and adjusting the polar radius change of the input point cloud. The mean, variance, and standard deviation are calculated sequentially; (d) the input point cloud is divided into equidistant segments based on the magnitude of the polar radius r, with each segment corresponding to a constant polar radius value, the magnitude of which is the median of the polar radius of the current segment's segmentation interval; (e) the number of point cloud points contained in each segment after equidistant segmentation is counted, and the constant polar radius value corresponding to the segment with the most points is set to the reference value r0 of the steel arch polar radius corresponding to the local point cloud; (f) each point in the input point cloud is traversed, and points whose difference between the polar radius r and the reference value r0 of the steel arch polar radius exceeds a set multiple of the standard deviation are marked and removed from the input point cloud; (g) (h) Calculate the ratio of the number of marker points to the total number of points in the input point cloud; (h) When the ratio is greater than a set value, use the input point cloud after removing the marker points as a new input point cloud, and repeat (c); (i) When the ratio is less than a set value, use the input point cloud after removing the marker points as an output point cloud, and complete the removal of patch noise from the current local point cloud; (j) Repeat (b) for the next group of local point clouds until the removal of patch noise from all local point clouds is completed, and define the set of output point clouds of all local point clouds as the denoised steel arch point cloud of the steel arch noise point cloud.

[0032] The non-rigid registration method for deformable steel arch frames in tunnels described in this invention, in a preferred embodiment, further includes the following steps in step 2:

[0033] Step 2.1, Point Cloud Construction: Read the design point cloud file of the steel arch frame and construct the point cloud corresponding to the design model of the steel arch frame from the read data, referred to as the design point cloud;

[0034] Step 2.2, Point Cloud Segmentation: Based on the design point cloud... The values ​​are divided into equally spaced segments, and each segment of the point cloud is represented by a feature point, which is the node corresponding to the current segment.

[0035] Step 2.3, Node Coordinate Assignment: Assign the x-coordinate of the node to zero, and assign the x-coordinate of the node to zero. The coordinates are assigned to all points contained in the segment corresponding to the node. The median of the distribution interval is used to assign the r coordinate of the node to the minimum value of r of all points contained in the corresponding segment of the node;

[0036] Step 2.4: Construction of the spatial curve of the design model: The spatial curve is represented by the extracted spatial point sequence. When the number of segments is large enough, the constructed spatial curve is smooth and matches the trend of the steel arch frame design model.

[0037] The non-rigid registration method for deformable steel arch frames in tunnels described in this invention, in a preferred embodiment, further includes the following steps in step 3:

[0038] Step 3.1, Point Cloud Mapping: Map the point cloud of the steel arch frame to a two-dimensional point set on a plane. It is required that the Euclidean distance between any two nearest neighbors in the local area of ​​the three-dimensional space is proportional to the Euclidean distance of the mapped plane, so as to reduce the error caused by the point cloud mapping operation to the final fitting curve result.

[0039] Step 3.2, Planar Curve Construction: Fit the two-dimensional point set to a two-dimensional plane curve;

[0040] Step 3.3: Construction of the spatial curve of the steel arch point cloud: Based on the above planar curve fitting results, construct the spatial trend characteristic curve of the steel arch point cloud.

[0041] The point cloud mapping includes two parts: mapping to an ideal cylindrical surface and mapping to an ideal plane. Specifically, it includes the following steps:

[0042] First, the point cloud is mapped onto an ideal cylindrical surface: the three-dimensional point cloud is mapped onto a spatial cylindrical surface with a constant mean polar radius, and the surface where the point cloud is located after mapping is defined as the ideal cylindrical surface of the steel arch point cloud.

[0043] Then the point cloud is further mapped onto the ideal plane: the cylindrical coordinate system that has been mapped to the ideal cylindrical surface. The 3D point cloud below is further mapped to a 3D Cartesian coordinate system. Two-dimensional plane In this process, the point cloud on the curved surface is mapped to a set of points on a plane, and the plane containing the mapped point cloud is defined as the ideal plane of the steel arch point cloud.

[0044] The two-dimensional plane curve fitting involves solving the plane curve function corresponding to the two-dimensional point set mapped to each tortuous steel arch. Specifically, it includes the following steps:

[0045] First, the point set is segmented: all points in the two-dimensional point set are segmented according to... The size is divided into equally spaced segments, and the distribution of each segment of points on the two-dimensional plane is roughly as several approximately parallel line segments, with each line segment corresponding to a different steel arch frame;

[0046] Next, two-dimensional line segment extraction is performed: planar line segment extraction is performed on each point set;

[0047] Finally, the planar curve is constructed: multiple planar line segments corresponding to two adjacent point sets are paired one by one according to their x-coordinate values. The paired adjacent line segments are then connected end to end to form a planar curve, thus completing the planar curve function. The structure.

[0048] The two-dimensional line segment extraction is based on Hough transform two-dimensional line recognition, and specifically includes the following steps:

[0049] First, straight line extraction is performed: two-dimensional straight line extraction based on Hough transform is performed on each point set, and the number of straight lines extracted is the number of steel arch frames identified in the current segment.

[0050] Next, the number of steel arch frames needs to be determined: the number of steel arch frames identified in each segment of the point set may not be the same, and the number of steel arch frames identified needs to be unified.

[0051] Finally, perform line segment extraction: for all extracted lines, group them according to their current segment size. Using boundary values ​​as a reference, line segments are extracted.

[0052] In some embodiments, the determination of the number of steel arches is based on statistical results of the number of steel arches identified in different segments, and mainly includes the following steps:

[0053] First, traverse each point set and count the number of times the identified steel arches appear. Take the number that appears most frequently as the final number of steel arches identified.

[0054] Secondly, for point sets where the number of identified steel arches is greater than the number of identified steel arches, the identified straight lines with lower confidence are removed based on the confidence level of the identified straight line parameters.

[0055] Finally, for the point set where the number of identified steel arches is less than the number of identified steel arches, the threshold for straight line discrimination is appropriately reduced to supplement new identified straight lines, so that the number of identified steel arches is equal to the number of identified steel arches.

[0056] The spatial curve construction of the point cloud of the steel arch frame is combined with the planar curve function. The deformation results of each tortuous steel arch in the extreme radial direction. The solution process includes the following steps:

[0057] First, for the plane curve function in accordance with Size is sampled at equal intervals;

[0058] Secondly, iterate through each sampling point. The sampling points are searched in the two-dimensional point set on the ideal plane. Several nearest neighbor points;

[0059] Then, for each of the nearest neighbor points, its polar radius value in the original cylindrical coordinate system before mapping is found according to the point cloud index, and the minimum value of the polar radius values ​​in the original cylindrical coordinate system of the nearest neighbor points is assigned to... The deformation results in the radial direction are completed. Solving for;

[0060] Finally, combining the aforementioned planar curve function and the deformation results in the radial direction. By traversing This completes the construction of the spatial curve of the point cloud of the steel arch frame.

[0061] The non-rigid registration method for deformable steel arch frames in tunnels described in this invention, in a preferred embodiment, further includes the following steps in step 4:

[0062] Step 4.1: Compare the spatial trend characteristic curve of the steel arch frame point cloud with the spatial trend characteristic curve of the design model. Represent the independent variable;

[0063] Step 4.2: Calculate the difference between the spatial trend characteristic curve of the steel arch frame point cloud and the spatial trend characteristic curve of the design model to obtain the change in curve coordinates. and ;

[0064] Step 4.3: Traverse each point in the designed point cloud. Superimpose vectors on this point Perform displacement;

[0065] Step 4.4: Store the displacement results of each point in the design point cloud as a new point cloud and output it to complete the non-rigid registration of the steel arch frame design model and the steel arch frame point cloud.

[0066] This invention addresses the problem of restoring steel arch supports within tunnels. Based on the polar radius distribution pattern of tunnel point clouds in cylindrical coordinates, it achieves separation of the rock wall portion from the non-rock wall portion in the tunnel scanning point cloud by constructing a hypothetical rock wall surface through weighted polar radius calculation of nearest neighbor points. Furthermore, it separates the steel arch support portion from noise by segmenting the point cloud according to polar angles and statistically analyzing the mode of the polar radius distribution of each local point cloud. Finally, it achieves non-rigid registration of the stressed and torsional deformed steel arch support by constructing a spatial trend characteristic curve of the steel arch support. Thus, this invention completes the preprocessing of the tunnel point cloud and the non-rigid registration of the steel arch support point cloud, enabling efficient and accurate restoration of stressed and torsional deformed steel arch supports within tunnels. This provides technical support for intelligent equipment to perceive information about tunnel rock wall supports and provides parameter support for guiding the path planning of the robotic arm of a new type of automated wet spraying machine. Attached Figure Description

[0067] Figure 1 This is a flowchart of a non-rigid registration method for deformable steel arch frames in tunnels;

[0068] Figure 2 This is a flowchart of a point cloud preprocessing method for deformed steel arch frames in tunnels;

[0069] Figure 3A This is a schematic diagram of the scanned point cloud after coordinate system transformation;

[0070] Figure 3B This is a schematic diagram of the construction result of the hypothetical rock wall point cloud;

[0071] Figure 3C This is a schematic diagram showing the separation results between the actual rock face and the noisy steel arch.

[0072] Figure 4A This is a schematic diagram of the planar point set after mapping the noise point cloud of the steel arch frame;

[0073] Figure 4B The point cloud of the denoised steel arch frame in a three-dimensional Cartesian coordinate system The diagram below;

[0074] Figure 5A The design model of the steel arch frame and the point cloud of the steel arch frame in a three-dimensional rectangular coordinate system The diagram below;

[0075] Figure 5B The spatial feature trend curve extraction results in a three-dimensional rectangular coordinate system The diagram below;

[0076] Figure 5C The non-rigid registration results between the steel arch frame design model and the steel arch frame point cloud in a three-dimensional rectangular coordinate system The diagram below;

[0077] Figure 6This is a schematic diagram of the steel arch frame design model and the point cloud of the steel arch frame before and after registration. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] The non-rigid registration method for deformable steel arch frames in tunnels provided by this invention can be implemented using computer software technology. The overall technical flowchart of the non-rigid registration method is shown below. Figure 1 This includes the following steps:

[0080] Step 1, Point Cloud Preprocessing: This step separates the rock face from the non-rock face, and the steel arch from the noise components in the original scanned point cloud. See the overall technical flowchart for the point cloud preprocessing. Figure 2 The specific implementation process of this embodiment is described as follows:

[0081] Step 1.1, Coordinate Transformation and Coordinate System Conversion: The original scanned point cloud is sequentially subjected to coordinate translation, coordinate rotation, and coordinate system transformation to obtain the scanned point cloud after coordinate system transformation, as shown in the figure. Figure 3A As shown;

[0082] Step 1.2, Denoising the blasting point cloud: Separate the real rock wall part and the steel arch noise part in the scanned point cloud after the coordinate system transformation to obtain the denoised blasting point cloud and steel arch noise point cloud;

[0083] Step 1.3, Denoising the steel arch point cloud: Separate the steel arch portion from the noise portion in the non-rock wall portion of the steel arch noise point cloud to obtain the denoised steel arch point cloud, thus completing the point cloud preprocessing.

[0084] The coordinate transformation and coordinate system conversion consist of coordinate translation, coordinate rotation, and coordinate system conversion, ensuring that the centroid of the original scanned point cloud is located at the origin of the coordinate system and the tunnel's central axis is parallel to the x-axis. Finally, the coordinate values ​​are converted from a rectangular coordinate system... - - Convert to cylindrical coordinate system - - And the scanned point cloud in the cylindrical coordinate system is... The two-dimensional discrete function shown is used to represent the point cloud after coordinate system transformation.

[0085] The denoising of the explosive point cloud is based on iterative denoising according to the change in polar radius before and after weighting of nearest neighbor points, specifically including the following steps:

[0086] First, construct a point cloud of the hypothetical rock face: traverse each point in the scanned point cloud after coordinate system transformation, and calculate the hypothetical rock face radius corresponding to the current point based on its nearest neighbor information using a weighted average of the nearest neighbor points. The scanned point cloud after coordinate system transformation following the polar radius change is defined as the hypothetical rock wall point cloud. Specifically,

[0087]

[0088] in For the number of nearest neighbors, For point Its nearest neighbor The weight, specifically,

[0089]

[0090] The first part of the multiplication expression on the right side of the equation depends on the point. Its nearest neighbor Perpendicular to The Euclidean distance of the projection onto the plane of direction, the second part depends on the point. Its nearest neighbor At the polar radius The difference in direction, defining the scanned point cloud after coordinate system transformation following the polar radius change as the hypothetical rock wall point cloud, the result is as follows: Figure 3B As shown;

[0091] Next, calculate the polar radius transformation: traverse each point in the scanned point cloud after the coordinate system transformation, and use the hypothetical rock wall polar radius of the current point. Replace the original extreme diameter r value and calculate the extreme diameter of the hypothetical rock wall. The difference between the polar radius and the original polar radius r value is defined as the change in polar radius of each point in the scanned point cloud after the coordinate system transformation. ;

[0092] Finally, iterative removal of non-rock wall point clouds: An iterative method is used to separate the real rock wall portion from the noisy steel arch portion in the scanned point cloud, as shown in the following results. Figure 3C As shown.

[0093] In this embodiment, the iterative removal of non-rock wall point clouds is performed on the scanned point cloud after coordinate system transformation based on the polar radius change. The numerical values ​​are iteratively denoised, specifically including the following steps: (a) using the scanned point cloud after coordinate system transformation as the initial input point cloud; (b) calculating the total number of points in the input point cloud and adjusting the polar radius change of the input point cloud. The mean, variance, and standard deviation are calculated sequentially; (c) the polar radius variation is calculated for each point in the input point cloud. Points whose difference from the mean exceeds a set multiple of the standard deviation are marked and removed from the input point cloud; (d) the ratio of the number of marked points to the total number of points in the input point cloud is calculated; (e) when the ratio is greater than a set value, the input point cloud after removing the marked points is used as a new input point cloud, and (b) is repeated; (f) when the ratio is less than a set value, the input point cloud after removing the marked points is used as the output point cloud, and the iterative denoising of the scanned point cloud after the coordinate system transformation is completed. The output point cloud is defined as the denoised explosion point cloud, and the difference between the initial input point cloud and the output point cloud is defined as the steel arch noise point cloud.

[0094] The point cloud denoising of the steel arch frame is based on the equidistant segmentation of the polar angle, followed by iterative denoising based on the magnitude of the mode of the polar radius. Specifically, it includes the following steps:

[0095] First, the point cloud is mapped: the noise point cloud of the steel arch frame is transferred from the cylindrical coordinate system. Mapped to a three-dimensional Cartesian coordinate system The two-dimensional plane below In the process, the planar point set after mapping the noise point cloud of the steel arch frame is obtained, and the results are as follows. Figure 4A As shown;

[0096] Next, the point cloud is grouped: the planar point set after mapping the steel arch noise point cloud is grouped according to the polar angle. The values ​​are divided into equal segments to obtain multiple sets of local point clouds;

[0097] Finally, iterative noise removal was performed: an iterative method was used to remove patchy noise from each group of local point clouds, resulting in the denoised steel arch point cloud. Figure 4B As shown.

[0098] In this embodiment, the iterative noise removal is performed by removing patchy noise from each group of local point clouds based on the density of noise distribution along the polar radius r. Specifically, this includes the following steps: (a) traversing each group of local point clouds; (b) using the local point cloud as the initial input point cloud; and (c) calculating the total number of points in the input point cloud and adjusting the polar radius change of the input point cloud. The mean, variance, and standard deviation are calculated sequentially; (d) the input point cloud is divided into equidistant segments based on the magnitude of the polar radius r, with each segment corresponding to a constant polar radius value, the magnitude of which is the median of the polar radius of the current segment's segmentation interval; (e) the number of point cloud points contained in each segment after equidistant segmentation is counted, and the constant polar radius value corresponding to the segment with the most points is set to the reference value r0 of the steel arch polar radius corresponding to the local point cloud; (f) each point in the input point cloud is traversed, and points whose difference between the polar radius r and the reference value r0 of the steel arch polar radius exceeds a set multiple of the standard deviation are marked and removed from the input point cloud; (g) (h) Calculate the ratio of the number of marker points to the total number of points in the input point cloud; (h) When the ratio is greater than a set value, use the input point cloud after removing the marker points as a new input point cloud, and repeat (c); (i) When the ratio is less than a set value, use the input point cloud after removing the marker points as an output point cloud, and complete the removal of patch noise from the current local point cloud; (j) Repeat (b) for the next group of local point clouds until the removal of patch noise from all local point clouds is completed, and define the set of output point clouds of all local point clouds as the denoised steel arch point cloud of the steel arch noise point cloud.

[0099] The denoised steel arch point cloud obtained in step 1 is referred to as the steel arch point cloud. The steel arch design model and the steel arch point cloud are then compared using a cylindrical coordinate system based on their coordinate values. Mapped to a three-dimensional Cartesian coordinate system The results are as follows Figure 5A As shown, the extraction of spatial characteristic curves of the steel arch frame design model and the point cloud of the steel arch frame will be performed sequentially in steps 2 and 3.

[0100] Step 2: Curve Extraction from the Design Model: Ignoring local features of the steel arch frame design model and retaining only its overall trend, extract and construct the spatial trend characteristic curve of the design model to approximately describe the steel arch frame design model. The result is as follows: Figure 5B As shown. The specific implementation process of this embodiment is described below:

[0101] Step 2.1, Point Cloud Construction: Read the design point cloud file of the steel arch frame and construct the point cloud corresponding to the design model of the steel arch frame from the read data, referred to as the design point cloud;

[0102] Step 2.2, Point Cloud Segmentation: Based on the design point cloud... The values ​​are divided into equally spaced segments, and each segment of the point cloud is represented by a feature point, which is the node corresponding to the current segment.

[0103] Step 2.3, Node Coordinate Assignment: Assign the x-coordinate of the node to zero, and assign the x-coordinate of the node to zero. The coordinates are assigned to all points contained in the segment corresponding to the node. The median of the distribution interval is used to assign the r coordinate of the node to the minimum value of r of all points contained in the corresponding segment of the node;

[0104] Step 2.4: Construction of the spatial curve of the design model: The spatial curve is represented by the extracted spatial point sequence. When the number of segments is large enough, the constructed spatial curve is smooth and matches the trend of the steel arch frame design model.

[0105] Step 3: Curve extraction of the steel arch point cloud: Ignoring local features of the steel arch point cloud and retaining only its overall trend, extract and construct the spatial trend feature curve of the steel arch point cloud to approximate its description. The result is as follows: Figure 5B As shown. The specific implementation process of this embodiment is described below:

[0106] Step 3.1, Point Cloud Mapping: Map the point cloud of the steel arch frame to a two-dimensional point set on a plane. It is required that the Euclidean distance between any two nearest neighbors in the local area of ​​the three-dimensional space is proportional to the Euclidean distance of the mapped plane, so as to reduce the error caused by the point cloud mapping operation to the final fitting curve result.

[0107] Step 3.2, Planar Curve Construction: Fit the two-dimensional point set to a two-dimensional plane curve;

[0108] Step 3.3: Construction of the spatial curve of the steel arch point cloud: Based on the above planar curve fitting results, construct the spatial trend characteristic curve of the steel arch point cloud.

[0109] The point cloud mapping includes two parts: mapping to an ideal cylindrical surface and mapping to an ideal plane. Specifically, it includes the following steps:

[0110] First, the point cloud is mapped onto an ideal cylindrical surface: the three-dimensional point cloud is mapped onto a spatial cylindrical surface with a constant mean polar radius, and the surface where the point cloud is located after mapping is defined as the ideal cylindrical surface of the steel arch point cloud.

[0111] Then the point cloud is further mapped onto the ideal plane: the cylindrical coordinate system that has been mapped to the ideal cylindrical surface. The 3D point cloud below is further mapped to a 3D Cartesian coordinate system. Two-dimensional plane In this process, the point cloud on the curved surface is mapped to a set of points on a plane, and the plane containing the mapped point cloud is defined as the ideal plane of the steel arch point cloud.

[0112] The two-dimensional plane curve fitting involves solving the plane curve function corresponding to the two-dimensional point set mapped to each tortuous steel arch. Specifically, it includes the following steps:

[0113] First, the point set is segmented: all points in the two-dimensional point set are segmented according to... The size is divided into equally spaced segments, and the distribution of each segment of points on the two-dimensional plane is roughly as several approximately parallel line segments, with each line segment corresponding to a different steel arch frame;

[0114] Next, two-dimensional line segment extraction is performed: planar line segment extraction is performed on each point set;

[0115] Finally, the planar curve is constructed: multiple planar line segments corresponding to two adjacent point sets are paired one by one according to their x-coordinate values. The paired adjacent line segments are then connected end to end to form a planar curve, thus completing the planar curve function. The structure.

[0116] The two-dimensional line segment extraction is based on Hough transform two-dimensional line recognition, and specifically includes the following steps:

[0117] First, straight line extraction is performed: two-dimensional straight line extraction based on Hough transform is performed on each point set, and the number of straight lines extracted is the number of steel arch frames identified in the current segment.

[0118] Next, the number of steel arch frames needs to be determined: the number of steel arch frames identified in each segment of the point set may not be the same, and the number of steel arch frames identified needs to be unified.

[0119] Finally, perform line segment extraction: for all extracted lines, group them according to their current segment size. Using boundary values ​​as a reference, line segments are extracted.

[0120] In this embodiment, the determination of the number of steel arch frames is based on the statistical results of the number of steel arch frames identified in different segments, and mainly includes the following steps:

[0121] First, traverse each point set and count the number of times the identified steel arches appear. Take the number that appears most frequently as the final number of steel arches identified.

[0122] Secondly, for point sets where the number of identified steel arches is greater than the number of identified steel arches, the identified straight lines with lower confidence are removed based on the confidence level of the identified straight line parameters.

[0123] Finally, for the point set where the number of identified steel arches is less than the number of identified steel arches, the threshold for straight line discrimination is appropriately reduced to supplement new identified straight lines, so that the number of identified steel arches is equal to the number of identified steel arches.

[0124] The spatial curve construction of the point cloud of the steel arch frame is combined with the planar curve function. The deformation results of each tortuous steel arch in the extreme radial direction. The solution process includes the following steps:

[0125] First, for the plane curve function in accordance with Size is sampled at equal intervals;

[0126] Secondly, iterate through each sampling point. The sampling points are searched in the two-dimensional point set on the ideal plane. Several nearest neighbor points;

[0127] Then, for each of the nearest neighbor points, its polar radius value in the original cylindrical coordinate system before mapping is found according to the point cloud index, and the minimum value of the polar radius values ​​in the original cylindrical coordinate system of the nearest neighbor points is assigned to... The deformation results in the radial direction are completed. Solving for;

[0128] Finally, combining the aforementioned planar curve function and the deformation results in the radial direction. By traversing This completes the construction of the spatial curve of the point cloud of the steel arch frame.

[0129] Step 4: Point cloud registration based on spatial curves: By matching the extracted spatial trend feature curves, non-rigid registration between the steel arch frame design model and the steel arch frame point cloud is further achieved. Results Figure 5C As shown. The specific implementation process of this embodiment is described below:

[0130] Step 4.1: Compare the spatial trend characteristic curve of the steel arch frame point cloud with the spatial trend characteristic curve of the design model. Represent the independent variable;

[0131] Step 4.2: Calculate the difference between the spatial trend characteristic curve of the steel arch frame point cloud and the spatial trend characteristic curve of the design model to obtain the change in curve coordinates. and ;

[0132] Step 4.3: Traverse each point in the designed point cloud. Superimpose vectors on this point Perform displacement;

[0133] Step 4.4: Store the displacement results of each point in the design point cloud as a new point cloud and output it to complete the non-rigid registration of the steel arch frame design model and the steel arch frame point cloud.

[0134] The design model of the steel arch frame before and after registration and the point cloud of the steel arch frame are as follows: Figure 6As shown, the registration effect of the steel arch frame can be evaluated visually. By observing and evaluating the registration results of the steel arch frame from multiple angles, it can be concluded that the technical method in this embodiment can meet the needs of restoring the steel arch frame subjected to stress and torsion deformation in actual engineering projects.

Claims

1. A point cloud preprocessing method for deformed steel arch frames in tunnels, characterized in that, The method includes the following steps: Step 1, Point Cloud Preprocessing: Separating rock walls from non-rock walls, and steel arches from noisy parts in the original scanned point cloud; Step 2: Curve extraction of the design model: Ignore the local features of the steel arch frame design model and retain only its overall trend. Extract and construct the spatial trend feature curve of the steel arch frame design model to approximate the steel arch frame design model. Step 3: Curve extraction of steel arch point cloud: Ignore the local features of the steel arch point cloud and retain only its overall trend. Extract and construct the spatial trend feature curve of the steel arch point cloud to approximate the steel arch point cloud. Step 4, Point Cloud Registration Based on Spatial Curves: By matching the spatial trend feature curves output in Step 2 and Step 3, non-rigid registration between the steel arch frame design model and the steel arch frame point cloud is further achieved. Step 1 further includes the following steps: Step 1.1, Coordinate Transformation and Coordinate System Conversion: Perform coordinate translation, coordinate rotation and coordinate system transformation on the original scanned point cloud in sequence to obtain the scanned point cloud after coordinate system transformation; Step 1.2, Denoising the blasting point cloud: Divide the point cloud output in Step 1.1 into two parts: the real rock wall and the noise of the steel arch frame; Step 1.3, Denoising the point cloud of the steel arch frame: The noisy point cloud of the steel arch frame output in Step 1.2 is divided into two parts: the steel arch frame and the noise. Step 1.2 further includes the following steps: First, construct a point cloud of the hypothetical rock face: For each point in the point cloud output in step 1.1, calculate the hypothetical rock face radius corresponding to the current point based on its nearest neighbor information using a weighted average of the nearest neighbor points. Specifically, ， in For the number of nearest neighbors, For point Its nearest neighbor The weight, specifically, ， The first part of the multiplication expression on the right side of the equation depends on the point. Its nearest neighbor Perpendicular to The Euclidean distance of the projection onto the plane of direction, the second part depends on the point. Its nearest neighbor At the polar radius Difference in direction; Next, the change in polar diameter is calculated: the polar diameter of the hypothetical rock wall is calculated for each point in the point cloud output in step 1.

1. The difference in polar radius between the stated polar radius r value; Finally, iterate to remove non-rock wall point clouds: using an iterative method, the point cloud output in step 1.1 is processed according to the aforementioned polar radius change. Iteratively denoise the numerical values; Step 1.3 further includes the following steps: First, the point cloud is mapped: the point cloud output in step 1.2 is mapped from cylindrical coordinates. Mapped to a three-dimensional rectangular coordinate system The two-dimensional plane below middle; Then the point cloud is grouped: the planar point sets after mapping the point cloud are grouped according to the polar angle. The values ​​are divided into equal segments to obtain multiple sets of local point clouds; Finally, iterative noise removal is performed: using an iterative method, patch noise is removed from each group of local point clouds based on the density of the distribution along the polar radius r.

2. The point cloud preprocessing method for a deformed steel arch frame in a tunnel according to claim 1, characterized in that: Step 2 further includes the following steps: Step 2.1, Point Cloud Construction: Read the point cloud file of the steel arch frame design and construct the corresponding point cloud of the steel arch frame design model from the read data; Step 2.2, Point Cloud Segmentation: Based on the corresponding point cloud of the steel arch frame design model. The values ​​are divided into equally spaced segments, and each segment of the point cloud is represented by a feature point, which is the node corresponding to the current segment. Step 2.3, Node Coordinate Assignment: Assign the x-coordinate of the node to zero, and assign the x-coordinate of the node to zero. The coordinates are assigned to all points contained in the segment corresponding to the node. The median of the distribution interval is used to assign the r coordinate of the node to the minimum value of r of all points contained in the corresponding segment of the node; Step 2.4: Design the spatial curve construction of the model: Represent the spatial curve using the extracted spatial point sequence.

3. The point cloud preprocessing method for a deformed steel arch frame in a tunnel according to claim 1, characterized in that: Step 3 further includes the following steps: Step 3.1, Point Cloud Mapping: Map the point cloud of the steel arch frame into a two-dimensional point set on a plane; Step 3.2, Plane Curve Construction: Fit the two-dimensional point set to a two-dimensional plane curve, and solve for the plane curve function corresponding to the mapped two-dimensional point set for each tortuous steel arch. ; Step 3.3, Spatial Curve Construction of the Point Cloud of the Steel Arch Frame: Combining the planar curve function output in Step 2.2 The deformation results of each tortuous steel arch in the extreme radial direction. Solve the problem.

4. The point cloud preprocessing method for a deformed steel arch frame in a tunnel according to claim 3, characterized in that: Step 3.2 further includes the following steps: First, the point set is segmented: all points in the two-dimensional point set are segmented according to... The size is divided into equally spaced segments, and the distribution of each segment of points on the two-dimensional plane is a number of approximately parallel line segments, each line segment corresponding to a different steel arch frame; Next, two-dimensional line segment extraction is performed: planar line segment extraction is performed on each point set; Finally, the planar curve is constructed: multiple planar line segments corresponding to two adjacent point sets are paired one by one according to the magnitude of the x coordinate value. The paired adjacent line segments are then connected end to end to form a planar curve, thus completing the planar curve function. The structure.

5. A point cloud preprocessing method for a deformed steel arch frame in a tunnel according to claim 4, characterized in that: The two-dimensional line segment extraction, based on Hough transform two-dimensional line recognition, further includes the following steps: First, straight line extraction is performed: two-dimensional straight line extraction based on Hough transform is performed on each point set, and the number of straight lines extracted is the number of steel arch frames identified in the current segment. Next, the number of steel arch frames needs to be determined: the number of steel arch frames identified in each segment of the point set may not be the same, and the number of steel arch frames identified needs to be unified. Finally, perform line segment extraction: for all extracted lines, group them according to their current segment size. Using boundary values ​​as a reference, line segments are extracted.

6. The point cloud preprocessing method for a deformed steel arch frame in a tunnel according to claim 3, characterized in that: Step 3.3 further includes the following steps: First, for the plane curve function in accordance with Size is sampled at equal intervals; Secondly, iterate through each sampling point. The sampling points are searched in a two-dimensional point set on an ideal plane. Several nearest neighbor points; Then, for each of the nearest neighbor points, its polar radius value in the original cylindrical coordinate system before mapping is found according to the point cloud index, and the minimum value of the polar radius values ​​in the original cylindrical coordinate system of the nearest neighbor points is assigned to... The deformation results in the radial direction are completed. Solve for; Finally, combining the aforementioned planar curve function and the deformation results in the radial direction. By traversing This completes the construction of the spatial curve of the point cloud of the steel arch frame.

7. The point cloud preprocessing method for a deformed steel arch frame in a tunnel according to claim 1, characterized in that: Step 4 further includes the following steps: Step 4.1: Compare the spatial trend characteristic curve of the steel arch point cloud with the spatial trend characteristic curve of the steel arch design model. Represent the independent variable; Step 4.2: Calculate the difference between the spatial trend characteristic curve of the steel arch point cloud and the spatial trend characteristic curve of the steel arch design model to obtain the change in curve coordinates. and ; Step 4.3: Traverse each point in the corresponding point cloud of the steel arch frame design model. Superimpose vectors on this point Perform displacement; Step 4.4: Store the displacement results of each point in the corresponding point cloud of the steel arch frame design model as a new point cloud and output it.

Images