Ray-based Tunnel Point Cloud Denoising Method and System

Abstract

The present invention discloses a tunnel point cloud denoising method and system based on the ray method. Based on the original point cloud data, a three-dimensional point cloud data model of the tunnel is constructed; along the direction of the tunnel central axis, the three-dimensional point cloud data model of the tunnel is equally spaced and divided into a number of tunnel point cloud slices; the point cloud data of each slice is projected onto the middle cross-section of the slice to obtain a tunnel contour line, and the tunnel contour line is regarded as the contour extraction object; a tunnel contour extractor is constructed; the tunnel contour extractor is matched with the geometric center point of the contour extraction object; any point on the tunnel contour extractor is used as the origin of the ray emission, and the geometric center of the contour extraction object is used as the target point of the ray, and a ray is emitted between the origin and the target point; it is judged whether there are point cloud sampling points on the tunnel point cloud in the direction of each ray. If so, the point cloud sampling points are saved, and finally all the saved point cloud sampling points are merged to obtain the tunnel point cloud data with the tunnel point cloud noise removed.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel contour point cloud data processing, and in particular to a tunnel point cloud denoising method and system based on the ray method. Background Technique

[0002] The statements in this part only mention the background techniques related to the present invention, and do not necessarily constitute prior art.

[0003] At present, as an emerging advanced vision technology, three-dimensional laser scanning technology is widely used in tunnel monitoring measurement and visualization digital management due to its characteristics of high precision, high efficiency, high resolution, fully automatic digital acquisition, and rich data volume. Currently, the measurement accuracy of emerging three-dimensional laser scanners is continuously improving, and the instrument cost is continuously decreasing, making it possible to replace the total station in many field applications and basically meet the accuracy requirements of tunnel monitoring measurement. However, because the obtained point cloud data is too rich and there are a large number of noise points, point cloud filtering is required before subsequent data application to remove noise points and structures that are not of interest, such as tunnel internal accessory structures, vehicles, pedestrians, etc. In addition, the quality of point cloud filtering will directly affect whether the three-dimensional laser point cloud meets the application requirements.

[0004] Traditional denoising methods for tunnel point clouds mainly include fitting methods, segmentation methods, triangulation methods, machine learning methods, etc., and all have certain limitations. The tunnel filtering and denoising method based on geometric shape fitting depends on the selection of interpolation methods, and the multi-level iterative method adopted is prone to error transmission and accumulation. The filtering effect of the filtering algorithm based on segmentation depends too much on the result of clustering segmentation. The method based on irregular triangulation requires a large amount of memory and is greatly affected by noise points of tunnel internal accessory structures, machinery, vehicles, etc. The filtering method based on machine learning requires a large number of training samples, and the samples must cover all possible tunnel shape features, which requires very high computer resources and is difficult to achieve good filtering effects.

[0005] The above tunnel denoising methods all focus on removing tunnel noise from the inside of the tunnel. However, the tunnel structure has a large size scale and the tunnel interior contains complex noise data. Especially during the construction period, the tunnel contains a large number of complex structures, and obtaining tunnel contour point clouds through tunnel point cloud filtering and denoising will bring a large amount of workload and technical difficulties. Summary of the Invention

[0006] In order to solve the deficiencies of the prior art, the present invention provides a tunnel point cloud denoising method and system based on the ray method; starting from the outside of the tunnel, it can efficiently identify the tunnel contour and extract the visualization data of the tunnel contour section, and can solve the problems of large difficulty and low accuracy in filtering and denoising the existing tunnel structure point cloud, and the unsatisfactory denoising effect.

[0007] In the first aspect, the present invention provides a tunnel point cloud denoising method based on the ray method;

[0008] The tunnel point cloud denoising method based on the ray method includes:

[0009] (1) Based on the three-dimensional laser scanning method, obtain the original point cloud data of the tunnel; based on the original point cloud data, construct a three-dimensional point cloud data model of the tunnel; perform denoising processing on the three-dimensional point cloud data model of the tunnel to remove the tunnel point cloud outliers outside the tunnel point cloud contour;

[0010] (2) Along the direction of the tunnel central axis, equally divide the three-dimensional point cloud data model of the denoised tunnel into several tunnel point cloud slices at equal intervals; project the point cloud data of each slice onto the middle cross-section of the slice to obtain a tunnel contour line, and regard the tunnel contour line as the contour extraction object;

[0011] (3) Construct a tunnel contour extractor; for the tunnel contour extractor, first magnify the lining contour of the designed size according to a set ratio, and then perform point sampling on the magnified tunnel lining contour. The sampling density is set according to the finally expected density of the denoised tunnel point cloud, and obtain the point cloud data representing the magnified tunnel lining contour, which consists of a group of points containing coordinate information, to obtain the tunnel contour extractor;

[0012] (4) Match the geometric center point of the tunnel contour extractor with the contour extraction object;

[0013] (5) Use any point on the tunnel contour extractor as the origin of the ray, and use the geometric center of the contour extraction object as the target point of the ray, and emit a ray between the origin and the target point;

[0014] (6) Determine whether there are point cloud sampling points on the tunnel point cloud in each ray direction. If not, return to (4) and continue to detect the next contour extraction object. If so, save the point cloud sampling points, return to (5), change the origin coordinate position, re-emit the ray, repeat (5) several times until there are no new point cloud sampling points, return to (4) to detect the next contour extraction object, and finally merge all the saved point cloud sampling points to obtain the tunnel point cloud data with tunnel point cloud noise removed.

[0015] In the second aspect, the present invention provides a tunnel point cloud denoising system based on the ray method;

[0016] The tunnel point cloud denoising system based on the ray method includes:

[0017] An acquisition module, which is configured to: acquire the original point cloud data of the tunnel based on the three-dimensional laser scanning method; construct a three-dimensional point cloud data model of the tunnel based on the original point cloud data; perform denoising processing on the three-dimensional point cloud data model of the tunnel to remove the outlier point cloud of the tunnel outside the tunnel point cloud contour;

[0018] A slicing and segmentation module, which is configured to: evenly divide the three-dimensional point cloud data model of the denoised tunnel into a plurality of tunnel point cloud slices along the tunnel central axis direction; project the point cloud data of each slice onto the middle cross-section of the slice to obtain a tunnel contour line, and regard the tunnel contour line as an object for contour extraction;

[0019] A construction module, which is configured to: construct a tunnel contour extractor. The tunnel contour extractor first enlarges the lining contour of the design size according to a set ratio, and then samples points from the enlarged tunnel lining contour. The sampling density is set according to the final point cloud density of the denoised tunnel point cloud to obtain the point cloud data representing the enlarged tunnel lining contour, which consists of a group of points containing coordinate information, and obtain the tunnel contour extractor;

[0020] A matching module, which is configured to: match the geometric center point of the tunnel contour extractor with the object for contour extraction;

[0021] A ray emission module, which is configured to: use any point on the tunnel contour extractor as the origin of the ray emission, and use the geometric center of the object for contour extraction as the target point of the ray, and emit a ray between the origin and the target point;

[0022] A judgment module, which is configured to: judge whether there are point cloud sampling points on the tunnel point cloud in each ray direction. If not, return to the matching module to continue detecting the next object for contour extraction. If so, save the point cloud sampling points, return to the ray emission module to change the origin coordinate position and then re-emit the ray, repeat the work of the ray emission module several times until there are no new point cloud sampling points, return to the matching module to detect the next object for contour extraction, and finally merge all the saved point cloud sampling points to obtain the tunnel point cloud data after removing the tunnel point cloud noise.

[0023] In a third aspect, the present invention also provides an electronic device, including:

[0024] A memory for non-temporarily storing computer-readable instructions; and

[0025] A processor for running the computer-readable instructions,

[0026] wherein, when the computer-readable instructions are run by the processor, the method described in the first aspect above is executed.

[0027] Fourthly, the present invention also provides a storage medium that non - temporarily stores computer - readable instructions. When the non - temporary computer - readable instructions are executed by a computer, the instructions for executing the method described in the first aspect are executed.

[0028] Fifthly, the present invention also provides a computer program product, including a computer program, which is used to implement the method described in the above - mentioned first aspect when running on one or more processors.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0030] Existing common tunnel denoising techniques can be divided into those based on tunnel structure features and those based on point - cloud point features. The former focuses on the tunnel structure and uses methods such as geometric fitting or model filtering. The filtering effect is not thorough, and often multiple filtering methods need to be combined. The latter is based on the features of local neighborhood points of the point cloud, such as density, curvature, normal, etc. The computational amount is large, and the filtering efficiency is low. It often cannot remove irrelevant point - cloud data such as accessory structures, vehicles, and personnel in the tunnel. The present invention focuses on the outside of the tunnel cross - section, realizes the acquisition of the external contour of the tunnel, can effectively remove a large amount of noisy point - cloud data in the tunnel, especially for the point cloud of the tunnel during the construction period. The operation is relatively simple, and it has good processing efficiency and denoising effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The specification drawings forming a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention.

[0032] Figure 1 It is a flowchart of the method for tunnel point - cloud denoising based on the ray method in the embodiment of the present invention;

[0033] Figure 2 It is the original point - cloud data of a highway tunnel during the construction period obtained by the three - dimensional laser scanning technology in the embodiment of the present invention;

[0034] Figure 3 It is a schematic diagram of slicing the tunnel along the direction parallel to the tunnel central axis and projecting the sliced point cloud towards the center of the slice in the embodiment of the present invention;

[0035] Figure 4 It is a schematic diagram of the matching between the contour extractor and the tunnel contour extraction object in the embodiment of the present invention;

[0036] Figure 5 It is a schematic diagram of the radius neighborhood range of points on the ray in the embodiment of the present invention;

[0037] Figures 6(a) and 6(b) are diagrams showing the denoising effect of tunnel point cloud in the embodiment of the present invention;

[0038] Among them, 1. Tunnel point cloud contour; 2. Tunnel point cloud noise; 3. Tunnel point cloud outliers; 4. Tunnel point cloud slices; 5. Tunnel contour line; 6. Tunnel contour extractor; 7. Ray; 8. Radius neighborhood range of points on the ray; 9. Point cloud sampling points; 10. Original point cloud data of the tunnel; 11. Denoised tunnel point cloud data. Detailed implementation mode

[0039] It should be noted that the following detailed descriptions are all exemplary and are intended to provide further illustration of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0040] It should be noted that the terms used herein are only for describing specific implementation modes and are not intended to limit the exemplary implementation modes according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products or devices.

[0041] In the case of no conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0042] All data acquisition in this embodiment is based on compliance with laws, regulations and user consent, and is a legal application of the data.

[0043] Embodiment 1

[0044] This embodiment provides a method for denoising tunnel point clouds based on the ray method;

[0045] As Figure 1 shown, the method for denoising tunnel point clouds based on the ray method includes:

[0046] S101: Based on the three-dimensional laser scanning method, obtain the original point cloud data 10 of the tunnel; based on the original point cloud data, construct a three-dimensional point cloud data model of the tunnel; perform denoising processing on the three-dimensional point cloud data model of the tunnel to remove the tunnel point cloud outliers 3 outside the tunnel point cloud contour 1;

[0047] S102: Along the direction of the tunnel central axis, equally divide the three-dimensional point cloud data model of the denoised tunnel into several tunnel point cloud slices 4; project the point cloud data of each slice onto the middle cross-section of the slice to obtain the tunnel contour line 5, and regard the tunnel contour line as the contour extraction object;

[0048] S103: Construct a tunnel contour extractor 6. The tunnel contour extractor first magnifies the lining contour of the design dimensions according to a set ratio, and then samples points from the magnified tunnel lining contour. The sampling density is set according to the density of the denoised tunnel point cloud finally to be obtained, and point cloud data representing the magnified tunnel lining contour is obtained, which consists of a set of points containing coordinate information, thus obtaining the tunnel contour extractor;

[0049] S104: Match the tunnel contour extractor with the geometric center point of the contour extraction object;

[0050] S105: Use any point on the tunnel contour extractor as the origin from which a ray is emitted, and use the geometric center of the contour extraction object as the target point of the ray, and emit a ray 7 between the origin and the target point;

[0051] S106: Determine whether there are point cloud sampling points 9 on the tunnel point cloud in the direction of each ray. If not, return to S104 and continue to detect the next contour extraction object. If so, save the point cloud sampling points, return to S105, change the origin coordinate position, re-emit the ray, repeat S105 several times until there are no new point cloud sampling points, return to S104 to detect the next contour extraction object, and finally merge all the saved point cloud sampling points to obtain the tunnel point cloud data after removing the noise points 2 of the tunnel point cloud.

[0052] It should be understood that when cutting the tunnel point cloud data along the tunnel axis (the tunnel axis is a virtual curve parallel to its extension direction at the center position of the tunnel internal space, indicating the overall trend and attitude of the tunnel), in fact, the point cloud data within a certain range in the tunnel extension direction is selected as a slice (such as Z i ±delta / 2, Z i is the coordinate value of the extension direction of the tunnel contour line, and delta is the width of the slice). Therefore, project the point cloud of each slice onto the cross-section in the middle of each slice to obtain the contour of the tunnel cross-section, project each slice onto the middle plane of the slice to obtain the tunnel cross-section point cloud data tunnel contour line corresponding to each slice; regard the tunnel cross-section point cloud data tunnel contour line corresponding to each slice as the contour extraction object.

[0053] Regarding the slice width, if the slice is too thick, it will affect the representation of the actual space of the tunnel by the cross-section contour line, resulting in a decrease in accuracy; if the slice is too thin, it may result in uneven distribution of the point cloud within the slice and too large a proportion of noise points, thus leading to too large an error. Therefore, the slice thickness needs to be appropriate to ensure the accuracy of the slice and make the data representative.

[0054] Furthermore, the step S101: acquiring original point cloud data of the tunnel based on a three-dimensional laser scanning method; and constructing a three-dimensional point cloud data model of the tunnel based on the original point cloud data, specifically includes:

[0055] The tunnel structure is scanned using a 3D laser radar to obtain point cloud data of the existing tunnel structure, a 3D point cloud data model of the tunnel structure is obtained, and a local coordinate system is established for the 3D point cloud of the tunnel structure.

[0056] Exemplarily, the laser radar scanning operation in a tunnel scene: use a three-dimensional laser radar to scan the tunnel structure, scan tunnel objects, including but not limited to: highway tunnels and rail transit tunnels during operation and construction, obtain the point cloud data of the existing tunnel structure, and then obtain the basic three-dimensional point cloud data model of the tunnel structure after data solving and splicing, and establish a local coordinate system for the three-dimensional point cloud of the tunnel structure. Figure 2 It is the original point cloud data of the highway tunnel during the construction period obtained based on the three-dimensional laser scanning technology in the embodiment of the present invention.

[0057] Since tunnels are long and narrow structures, it is difficult for either a mounted or mobile laser radar to complete the tunnel structure scanning task in one operation. Therefore, the basic three-dimensional point cloud data model of the tunnel structure refers to the alignment and splicing of the point cloud data of the tunnel structure scanned by laser radar at multiple sites and in multiple phases.

[0058] Furthermore, the step S101: performing denoising on the three-dimensional point cloud data model of the tunnel to remove outlier points of the tunnel point cloud outside the tunnel point cloud contour specifically includes:

[0059] Based on the Gaussian distribution of the distances of all points in the point cloud, the mean and variance are set to remove obvious outlier noise points in the tunnel point cloud.

[0060] Furthermore, the step S102: dividing the denoised three-dimensional point cloud data model of the tunnel into a plurality of tunnel point cloud slices at equal intervals along the central axis of the tunnel, specifically includes:

[0061] Adjust the pose of the tunnel point cloud data in the Cartesian coordinate system: adjust the pose of the tunnel point cloud so that the tunnel section contour is parallel to the XOY plane, and the tunnel point cloud extends along the Z coordinate axis;

[0062] Calculate the minimum bounding box of the tunnel point cloud data and obtain the extreme value Z of the minimum bounding box on the Z coordinate axis min and Z max , set the thickness delta of the tunnel point cloud slice, the gap between adjacent slices is gap, and slice and divide the minimum bounding box of the tunnel point cloud data along the extension direction of the tunnel to obtain several tunnel point cloud slices.

[0063] It should be understood that in the Cartesian coordinate system, the tunnel cross-section contour is parallel to the XOY plane, and the tunnel point cloud extends along the positive direction of the Z coordinate axis.

[0064] Further, in S102, the point cloud data of each slice is projected onto the middle cross-section of the slice to obtain the tunnel contour line, and the tunnel contour line is regarded as the contour extraction object, which specifically includes:

[0065] After obtaining each slice, the point cloud data points in each slice are respectively projected in a dimension-reducing manner onto the middle cross-section of the slice, the Z-axis coordinates of all the point cloud data points in each slice are converted into the Z-axis coordinates of the center of the corresponding slice, the projected view of the obtained tunnel point cloud slice is regarded as the tunnel contour line, and the tunnel contour line will be used as the contour extraction object.

[0066] It should be understood that the beneficial effect of the above projection technical solution is: to prevent the number of point cloud data points in the thickness delta slice obtained from being too small, resulting in poor quality of the tunnel contour extracted from this slice.

[0067] The tunnel point cloud data is divided into equally spaced slices along the direction parallel to the tunnel central axis, and each slice is projected in a dimension-reducing manner onto the middle plane of the slice. The middle plane of the slice without thickness is used to replace the thickness slice, while retaining the point cloud data of the slice; Figure 3 The schematic diagram of slicing the tunnel along the direction parallel to the tunnel central axis and projecting the point cloud of the slice onto the center of the slice is given. In this embodiment, the tunnel point cloud is sliced and divided. The thickness of the slice is set to delta, and the slicing direction each time is along the tangent direction of the two-dimensional horizontal midline of the slice. The interval of the slice is taken as gap; the slice thickness delta should not be too small, otherwise the amount of slice point data will be too small; each slice is projected in a dimension-reducing manner onto the middle plane of the slice, that is, the Z coordinates of all points of each slice are converted into the Z coordinate z of the center of the slice. i The middle plane of the slice refers to the middle cross-section of each slice. The middle plane of the slice is perpendicular to the Z coordinate axis along which the tunnel point cloud extends.

[0068] Further, the slice is in the shape of the tunnel contour, and its cross-section shape can be divided into rectangle, circle, multi-centered circle, straight-wall arch, horseshoe shape, etc. Taking the three-centered circle tunnel as an example, its center point is the point on the central axis.

[0069] Further, in S103: constructing a tunnel contour extractor specifically includes:

[0070] The tunnel contour of the original design size of the tunnel is enlarged as a whole according to a set ratio to obtain the enlarged tunnel contour; the set ratio is selected as 1:1.2 or 1:1.5;

[0071] Based on the enlarged tunnel contour and the set point cloud density, generate the point cloud data of the enlarged tunnel contour; regard the point cloud data of the enlarged tunnel contour as the tunnel contour extractor;

[0072] The cross-sectional contour of the tunnel contour extractor is larger than the cross-sectional contour of the tunnel's original design size.

[0073] Further, if the original tunnel is circular, the tunnel contour extractor is a concentric circle of the original tunnel, but the radius of the tunnel contour extractor is larger than the radius of the original tunnel.

[0074] Further, if the original tunnel is non-circular, the tunnel contour extractor is a proportional enlargement of the original tunnel contour. Here, proportional enlargement means that if the ratio of the length to the height of the original tunnel is r, then the ratio of the length to the height of the tunnel contour extractor is still r; and if the ratio of the length to the width of the original tunnel is z, then the ratio of the length to the width of the tunnel contour extractor is still z.

[0075] Further, the tunnel contour extractor is a point cloud data file that encloses the cross-sectional contour of the tunnel. The tunnel contour extractor is equidistant tunnel contour point cloud data established according to the tunnel design size. Here, equidistant means that the distance between all the point cloud data points of the tunnel contour extractor and the tunnel contour surface of the original design size is a fixed value q.

[0076] Further, the tunnel contour extractor is a point cloud format file, composed of several coordinate points. The point cloud density of the contour extractor is higher than that of the original tunnel point cloud data obtained by 3D laser scanning; the point cloud density of the tunnel contour extractor is higher than the point density of the tunnel contour line; the cross-sectional contour of the tunnel contour extractor is larger than the cross-sectional contour of the tunnel's original design size.

[0077] Carrying XYZ coordinate information, it can directly generate a point set based on the tunnel design parameters, or generate a point cloud file after establishing a line sketch through CAD-related software. The resolution of the points of the contour extractor should be as large as possible to meet the contour extraction requirements.

[0078] The contour extractor is a point cloud data file that encloses the cross-sectional contour of the tunnel. It is equidistant tunnel contour point cloud data established according to the tunnel design size. The contour extractor is a point cloud format file, composed of high-density coordinate points, carrying XYZ coordinate information. It can directly generate a point set based on the tunnel design parameters, or generate a point cloud file after establishing a line sketch through CAD software. The resolution of the points of the contour extractor should be as large as possible to meet the contour extraction requirements.

[0079] If equidistant tunnel contour point cloud data is established according to the tunnel design dimensions, assuming the design radius of the tunnel is R1, then R1 + β is taken as the radius of the tunnel contour extractor; β is a set value.

[0080] Further, the matching of the tunnel contour extractor with the geometric center point of the contour extraction object means: matching and coinciding the geometric center of the tunnel contour extractor with the geometric center of the tunnel slice, i.e., the contour extraction object.

[0081] Further, S106: Judging whether there are point cloud sampling points on the tunnel point cloud in each ray direction specifically includes:

[0082] The ray is a point line composed of several points, with the points on the tunnel contour extractor as its endpoints, and points are iteratively generated along the direction pointing to the geometric center of the tunnel slice, i.e., the contour extraction object, and then this ray is formed; the step size for generating a new point A on the ray is L i , that is, the distance between point A and the endpoint of this ray is L i , then a circular detection area is constructed with A as the center and r as the radius, and it is judged whether there are point cloud data points in the detection area. If there are, the point with the closest perpendicular distance to the ray is found and saved as the point cloud sampling point, and the process ends. Then, the next point on the tunnel contour extractor is taken as the new ray endpoint and the above operations are repeated to construct the ray and detect the point cloud sampling point; if not, a new point B is generated for judgment; the value of L i is dynamic, L i = i * step, step is the distance between adjacent points set on this ray, and i represents the number of newly generated points. For example: L B -L A = step

[0083] Affected by the accuracy of scanning devices such as lidar and the scanning environment, the collected tunnel point cloud data has a "thickness". For example Figure 5 , when there are 5 tunnel point cloud data in the neighborhood radius r range of the latest emission point on the ray, the ray emission ends. At this time, the "intersection point" is a cluster of points, and then the projection distance between each point in the cluster and the ray is calculated respectively, and the one with the smallest projection distance is taken as the point cloud sampling point.

[0084] Taking a point greater than the tunnel contour extractor as the origin of the ray emission and the center of the tunnel section as the target point, a ray is emitted between these two points, and finally the point cloud sampling point is obtained; the above operations are repeated for all tunnel slices, and finally the point cloud sampling points on all tunnel slice sections are merged to obtain the tunnel contour point cloud data, realizing the denoising of the tunnel point cloud. Figure 4 The schematic diagram shows the slicing of the tunnel along the direction parallel to the tunnel central axis and the projection of the slice point cloud towards the slice center.

[0085] Further, the matching of the contour extractor and the contour extraction object specifically includes:

[0086] Calculate the geometric centers of the contour extractor and the contour extraction object respectively, and perform Euclidean transformation on the contour extractor to make the geometric centers of the two coincide and match.

[0087] Taking any point on the tunnel contour extractor as the origin from which a ray is emitted, and taking the center of the tunnel cross-section as the target point of the ray, a ray is emitted between the origin and the target point, which specifically includes:

[0088] Set the sampling resolution of the contour extractor. Taking the coordinate points on the contour extractor as the origin from which the ray is emitted, and taking the geometric center of the contour extraction object as the target point, a ray is emitted between these two points based on this direction. Finally, the "intersection point" of this ray and the contour extraction object is obtained. The so-called "intersection point" is the termination point of this ray, that is, the sampling point;

[0089] The ray is composed of points. Set the step size step for generating points on the ray; build a KD tree for the point cloud of the contour extraction object, perform a nearest neighbor search within the radius neighborhood range 8 of the points on the ray, and then calculate the distances from each nearest neighbor point in the nearest neighbor point cluster to the ray, and take the one with the minimum distance as the tunnel point cloud contour sampling point. Figure 5 This is a schematic diagram of the radius neighborhood range of the points on the ray in the embodiment of the present invention.

[0090] Merge all the saved point cloud sampling points to obtain the tunnel point cloud data after removing the noise 2 of the tunnel point cloud, which specifically includes:

[0091] Repeat the sampling operation for all contour extraction objects of the tunnel. Finally, merge all the point cloud sampling points to obtain the denoised tunnel point cloud data 11. The denoised tunnel point cloud data only contains the tunnel contour data and does not contain various noise information inside the tunnel.

[0092] The contour extractor can set different sampling resolutions according to the tunnel contour distribution. Specifically, for circular tunnels such as common subway and other rail transit tunnels, its contour extractor can be divided into the segment lining part and the track part, and different resolutions can be set; for tunnels with two-centered circles, three-centered circles, etc., such as highway tunnels, its contour extractor can be divided into the crown part, the shoulder part, the bottom part of the tunnel, etc., and different resolutions can be set. Figures 6(a) and 6(b) are the display diagrams of the tunnel point cloud denoising effect in the embodiment of the present invention.

[0093] Embodiment 2

[0094] This embodiment provides a tunnel point cloud denoising system based on the ray method;

[0095] The tunnel point cloud denoising system based on the ray method includes:

[0096] An acquisition module, which is configured to: acquire the original point cloud data of the tunnel based on a three-dimensional laser scanning method; construct a three-dimensional point cloud data model of the tunnel based on the original point cloud data; perform denoising processing on the three-dimensional point cloud data model of the tunnel to remove the outlier point cloud of the tunnel outside the tunnel point cloud contour.

[0097] A slicing and segmentation module, which is configured to: evenly segment the three-dimensional point cloud data model of the denoised tunnel into a number of tunnel point cloud slices along the tunnel central axis direction; project the point cloud data of each slice onto the middle cross-section of the slice to obtain a tunnel contour line, and regard the tunnel contour line as an object for contour extraction.

[0098] A construction module, which is configured to: construct a tunnel contour extractor. The tunnel contour extractor first magnifies the lining contour of the design size according to a set ratio, and then samples points from the magnified tunnel lining contour. The sampling density is set according to the final point cloud density of the denoised tunnel point cloud to obtain the point cloud data representing the magnified tunnel lining contour, which consists of a set of points containing coordinate information, and obtains the tunnel contour extractor.

[0099] A matching module, which is configured to: match the geometric center point of the tunnel contour extractor with the object for contour extraction.

[0100] A ray emission module, which is configured to: use any point on the tunnel contour extractor as the origin of the ray emission, and use the geometric center of the object for contour extraction as the target point of the ray, and emit a ray between the origin and the target point.

[0101] A judgment module, which is configured to: judge whether there are point cloud sampling points on the tunnel point cloud in each ray direction. If not, return to the matching module to continue detecting the next object for contour extraction. If so, save the point cloud sampling points, return to the ray emission module to change the origin coordinate position and then re-emit the ray, repeat the work of the ray emission module several times until there are no new point cloud sampling points, return to the matching module to detect the next object for contour extraction, and finally merge all the saved point cloud sampling points to obtain the tunnel point cloud data with the tunnel point cloud noise removed.

[0102] It should be noted here that the above acquisition module, slicing and segmentation module, construction module, matching module, ray emission module and judgment module correspond to steps S101 to S106 in Embodiment 1. The examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules can be executed in a computer system such as a set of computer executable instructions as part of the system.

[0103] In the above embodiments, the descriptions of the various embodiments have their own focuses. For parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0104] The proposed system can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of the above modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0105] Embodiment III

[0106] This embodiment also provides an electronic device, including: one or more processors, one or more memories, and one or more computer programs; wherein, the processor is connected to the memory, and the above one or more computer programs are stored in the memory. When the electronic device runs, the processor executes the one or more computer programs stored in the memory, so that the electronic device executes the method described in Embodiment I above.

[0107] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0108] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A part of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0109] In the implementation process, the steps of the above method can be completed by the integrated logic circuit in the hardware of the processor or the instructions in the form of software.

[0110] The method in Embodiment I can be directly embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory, and the processor reads the information in the memory and combines its hardware to complete the steps of the above method. To avoid repetition, it will not be described in detail here.

[0111] Those of ordinary skill in the art will realize that the units and algorithm steps of the examples described in conjunction with this embodiment can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0112] Embodiment 4

[0113] This embodiment also provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the method described in Embodiment 1 is completed.

[0114] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A tunnel point cloud denoising method based on the ray method, characterized in that, Including: (1) Obtain the original point cloud data of the tunnel based on the three-dimensional laser scanning method; Construct a three-dimensional point cloud data model of the tunnel based on the original point cloud data; Perform denoising processing on the three-dimensional point cloud data model of the tunnel to remove the outlier point cloud of the tunnel outside the tunnel point cloud contour; (2) Along the tunnel central axis direction, equally divide the denoised three-dimensional point cloud data model of the tunnel into several tunnel point cloud slices; project the point cloud data of each slice onto the middle cross-section of the slice to obtain the tunnel contour line, and regard the tunnel contour line as the contour extraction object; (3) Construct a tunnel contour extractor. The tunnel contour extractor first magnifies the lining contour of the designed size according to a set ratio, and then samples points from the magnified tunnel lining contour. The sampling density is set according to the density of the denoised tunnel point cloud finally to be obtained, and the point cloud data representing the magnified tunnel lining contour is obtained to get the tunnel contour extractor; (4) Match the geometric center point of the tunnel contour extractor with the contour extraction object; (5) Use any point on the tunnel contour extractor as the origin of the ray emission, and use the geometric center of the contour extraction object as the target point of the ray, and emit a ray between the origin and the target point; (6) Determine whether there are point cloud sampling points on the tunnel point cloud in the direction of each ray. If not, return to (4) and continue to detect the next contour extraction object. If so, save the point cloud sampling points, return to (5), change the origin coordinate position, and re-emit the ray. Repeat (5) several times until there are no new point cloud sampling points, return to (4) to detect the next contour extraction object, and finally merge all the saved point cloud sampling points to obtain the tunnel point cloud data with tunnel point cloud noise removed.

2. The ray method-based tunnel point cloud denoising method according to claim 1, characterized in that Obtain the original point cloud data of the tunnel based on the three-dimensional laser scanning method; Construct a three-dimensional point cloud data model of the tunnel based on the original point cloud data, specifically including: Use a three-dimensional lidar to scan the tunnel structure to obtain the point cloud data of the existing tunnel structure, obtain a three-dimensional point cloud data model of the tunnel structure, and establish a local coordinate system for the three-dimensional point cloud of the tunnel structure.

3. The tunnel point cloud denoising method based on the ray method according to claim 1, characterized in that, Perform denoising processing on the three-dimensional point cloud data model of the tunnel to remove the outlier point cloud of the tunnel outside the tunnel point cloud contour, specifically including: Based on the Gaussian distribution of the distances of all points in the point cloud, set the mean and variance to remove the obvious outlier noise points of the tunnel point cloud.

4. The ray-based tunnel point cloud denoising method according to claim 1, wherein, Along the tunnel central axis direction, equally divide the denoised three-dimensional point cloud data model of the tunnel into several tunnel point cloud slices, specifically including: Adjust the pose of the tunnel point cloud in the Cartesian coordinate system: adjust the pose of the tunnel point cloud so that the tunnel cross-section contour is parallel to the XOY plane, and the tunnel point cloud extends along the Z coordinate axis direction; Calculate the minimum bounding box of the tunnel point cloud data and obtain the extreme values Z of the minimum bounding box on the Z coordinate axis min and Z max , set the thickness delta of the tunnel point cloud slice and the gap between adjacent slices as gap. For the minimum bounding box of the tunnel point cloud data, perform slicing and dissection along the tunnel extension direction to obtain several tunnel point cloud slices.

5. The tunnel point cloud denoising method based on the ray method according to claim 1, characterized in that Project the point cloud data of each slice onto the middle cross-section of the slice to obtain the tunnel contour line, and regard the tunnel contour line as the contour extraction object, specifically including: After obtaining each slice, the point cloud data points within each slice are respectively projected for dimensionality reduction onto the middle cross-section of the slice. The Z-axis coordinates of all the point cloud data points within each slice are converted to the Z-axis coordinates corresponding to the center of the slice. The obtained projected tunnel point cloud slice is regarded as the tunnel contour line, and the tunnel contour line will be used as the object for contour extraction.

6. The ray-based tunnel point cloud denoising method according to claim 1, characterized in that Construct a tunnel contour extractor, specifically including: Overall magnify the tunnel contour of the original tunnel design size according to a set ratio to obtain an enlarged tunnel contour; Generate the point cloud data of the enlarged tunnel contour based on the enlarged tunnel contour and the set point cloud density; regard the point cloud data of the enlarged tunnel contour as the tunnel contour extractor; The cross-section contour of the tunnel contour extractor is larger than the cross-section contour of the tunnel of the original tunnel design size.

7. The ray method-based tunnel point cloud denoising method according to claim 1, characterized in that Judge whether there are point cloud sampling points on the tunnel point cloud in each ray direction, specifically including: The ray is a point line composed of several points, with the points on the tunnel contour extractor as its endpoints, and points are iteratively generated along the direction pointing to the geometric center of the tunnel slice, i.e., the tunnel contour extraction object, and then this ray is formed; the step size for generating a new point A on the ray is L i , that is, the distance between point A and the endpoint of this ray is L i , and then a circular detection area is constructed with A as the center and r as the radius. It is judged whether there are point cloud data points in the detection area. If there are, the point with the closest perpendicular distance to the ray is found and saved as the point cloud sampling point, and the process ends. Then, the next point on the tunnel contour extractor is used as the new ray endpoint, and the above operations are repeated to construct the ray and detect the point cloud sampling point; if not, a new point B is continuously generated for judgment; the value of L i is dynamic. L i = i * step, where step is the distance between adjacent points on this ray, and i represents the number of newly generated points.

8. Tunnel point cloud denoising system based on the ray method, characterized in that, Including: An acquisition module configured to: obtain the original point cloud data of the tunnel based on the three-dimensional laser scanning method; Based on the original point cloud data, construct a three-dimensional point cloud data model of the tunnel; perform denoising processing on the three-dimensional point cloud data model of the tunnel to remove the outlier tunnel point clouds outside the tunnel point cloud contour; A slice segmentation module configured to: evenly segment the three-dimensional point cloud data model of the denoised tunnel into a number of tunnel point cloud slices along the tunnel central axis direction; project the point cloud data of each slice onto the middle cross-section of the slice to obtain the tunnel contour line, and regard the tunnel contour line as the object for contour extraction; A construction module configured to: construct a tunnel contour extractor. The tunnel contour extractor first magnifies the lining contour of the design size according to a set ratio, and then performs point sampling on the enlarged tunnel lining contour. The sampling density is set according to the finally expected denoised tunnel point cloud density to obtain the point cloud data representing the enlarged tunnel lining contour, which consists of a set of points containing coordinate information, to obtain the tunnel contour extractor; A matching module configured to: match the geometric center points of the tunnel contour extractor and the contour extraction object; A ray emission module configured to: use any point on the tunnel contour extractor as the origin of the ray emission, and use the geometric center of the contour extraction object as the target point of the ray to emit a ray between the origin and the target point; A judgment module configured to: judge whether there are point cloud sampling points on the tunnel point cloud in each ray direction. If not, return to the matching module to continue detecting the next contour extraction object. If so, save the point cloud sampling points, return to the ray emission module to change the origin coordinate position and then re-emit the ray, repeat the work of the ray emission module several times until there are no new point cloud sampling points, return to the matching module to detect the next contour extraction object, and finally merge all the saved point cloud sampling points to obtain the tunnel point cloud data after removing the tunnel point cloud noise.

9. An electronic device, characterized by including: A memory for non-temporarily storing computer-readable instructions; And A processor for running the computer-readable instructions Wherein, when the computer-readable instructions are run by the processor, the method according to any one of claims 1-7 above is executed.

10. A storage medium, characterized in that, Non-transitory computer-readable instructions are stored, wherein when the non-transitory computer-readable instructions are executed by a computer, the instructions for executing the method according to any one of claims 1-7 are executed.

Images