Tunnel vehicle-mounted laser radar positioning and deviation correction method and system

Abstract

The application discloses a tunnel vehicle-mounted laser radar positioning and deviation rectification method and system, which comprises the following steps: acquiring the coordinate true value of a target point arranged on a tunnel site; extracting target point cloud from complete tunnel point cloud data according to tunnel characteristics, and proposing a boundary barycentric method to identify the coordinates of the target center point; one-to-one corresponding the coordinate true value of the target point with the identified target center point coordinates, and selecting a set number of points for registration operation through a method of random sampling and then traversing combination to obtain a coordinate conversion matrix; calculating the coordinates of the target center points after rectification based on the coordinate conversion matrix, calculating the standard deviation of each target center point coordinate after rectification and the corresponding coordinate true value, selecting the coordinate conversion matrix corresponding to the minimum standard deviation, and performing laser radar positioning and deviation rectification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel vehicle-mounted laser radar detection, and particularly relates to a tunnel vehicle-mounted laser radar positioning and deviation correction method and system. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute the prior art.

[0003] The tunnel vehicle-mounted laser radar detection technology mainly comprises the following steps: a laser radar is mounted on a vehicle, the vehicle carrying the laser radar is driven to scan a tunnel to be detected by means of laser radar three-dimensional data, a series of point cloud data are obtained, a motion trajectory composed of instantaneous positions and poses of the device is obtained, and three-dimensional point cloud data of the tunnel as a whole are calculated. Tunnel diseases are identified by means of registration analysis of multi-period tunnel point cloud data or comparative analysis of point cloud and tunnel modeling data, and therefore high-precision absolute coordinate positioning of tunnel point cloud needs to be realized.

[0004] The tunnel has the characteristics of complex geographical environment, closed and narrow space, and long longitudinal distance, and the shell of the tunnel is generally composed of a mountain rock wall or reinforced concrete, which causes a very serious shielding effect on signals. The traditional GPS and Beidou positioning systems cannot penetrate the shell of the tunnel to reach the inside of the tunnel. At present, the mainstream way of vehicle-mounted tunnel point cloud positioning is to position outside the tunnel according to the GPS combined inertial navigation, directly calculate the absolute coordinates of the tunnel point cloud according to the GPS position information outside the tunnel and the inertial navigation data inside the tunnel, or to improve the detection accuracy, to realize the arrangement of target points inside the tunnel, to take the measured absolute coordinate data of the target point center as a reference point, to register and correct the target center coordinate data of the point cloud, and to assign accurate absolute coordinates to the tunnel point cloud. However, the above-mentioned means still has the problem of low positioning accuracy. On the one hand, the GPS coordinate positioning has fluctuation, and the inertial navigation data error will continuously accumulate with the increase of the tunnel length, and the point cloud data of the same tunnel scanned multiple times has errors, which need to be registered and corrected. On the other hand, the registration and correction usually adopts a manual registration method of the target center absolute coordinates and the measured target point cloud center coordinates, but in the actual operation process, there are phenomena such as large manual point selection registration error and different registration effects caused by different selected corresponding registration points. SUMMARY

[0005] In order to solve the above-mentioned problems, the present application provides a tunnel vehicle-mounted laser radar positioning and deviation correction method and system, which can improve the accuracy of tunnel detection by extracting coordinates, iteratively selecting the best registration combination of targets, and finally correcting the whole point cloud.

[0006] In some embodiments, the following technical solutions are adopted:

[0007] A tunnel vehicle-mounted laser radar positioning and deviation correction method comprises the following steps:

[0008] Obtaining a coordinate true value of a target point arranged at a target point in a tunnel site;

[0009] Collecting tunnel point cloud data in a vehicle driving process by a laser radar;

[0010] Extracting a target point cloud from the complete tunnel point cloud data, and identifying a coordinate of a target center point;

[0011] Corresponding the coordinate true value of the target point with the identified coordinate of the target center point, and selecting a set number of points by a method of random sampling and then traversing combination to perform registration operation, to obtain a coordinate conversion matrix;

[0012] Calculating a coordinate of a target center point after correction based on the coordinate conversion matrix, calculating a standard deviation of each target center point coordinate after correction and a corresponding coordinate true value, selecting a coordinate conversion matrix corresponding to the minimum standard deviation, and performing laser radar positioning correction.

[0013] As an optional solution, collecting tunnel point cloud data in a vehicle driving process by a laser radar, specifically comprising:

[0014] Collecting point cloud data in a driving process by a laser radar, and calculating the collected point cloud data according to path information, to obtain tunnel point cloud data according to position, pose information and reflected laser point position information.

[0015] As an optional solution, extracting a target point cloud from complete point cloud data, specifically comprising:

[0016] Pretreating point cloud data to eliminate abnormal point cloud data;

[0017] Classifying point cloud data according to normal and curvature of the point cloud data, screening out vertical plane point cloud data whose normal vector is perpendicular to the ground normal vector, to obtain point cloud data containing road side stones and target points;

[0018] Dividing the point cloud data containing road side stones and target points into two groups according to point cloud reflection intensity, and extracting the group with less point cloud quantity as target point cloud.

[0019] As an optional solution, pretreating point cloud data to eliminate abnormal point cloud data, specifically comprising:

[0020] Traversing all point cloud data, calculating an average distance L i between each point and its nearest k neighboring points;

[0021] Calculating a mean value and a standard deviation of all average distances, and obtaining a distance threshold d max based on the mean value and the standard deviation;

[0022] Again traverse the point cloud, eliminate the points whose average distance with k neighbor points is greater than d. max

[0023] As an optional solution, the point cloud data is pre-processed to eliminate abnormal point cloud data, and further comprising:

[0024] Eliminate the point cloud data outside the distance range relative to the ground point cloud based on the road surface point cloud.

[0025] As an optional solution, the coordinates of the target center point are identified, and the boundary barycenter method is proposed, comprising:

[0026] Boundary extraction is performed on the target to obtain a fine target contour;

[0027] The average value of the four edges of the target is calculated to obtain four point coordinates;

[0028] The average value of the four point coordinates is calculated to obtain the center point coordinates.

[0029] As an optional solution, a method of first random sampling and then iterative traversal is used to select a certain number of points for registration operation to obtain the optimal coordinate transformation matrix, comprising:

[0030] At least three pairs of points are randomly selected to iteratively traverse the point cloud data for registration, and the point cloud data includes the coordinate true value data set of the target point and the coordinate data set of the target center point; each data set establishes a spatial coordinate system, and the relative positional relationship between the two coordinate systems is calculated to obtain the rigid transformation matrix of the measured point cloud and the coordinate true value registration;

[0031] All measured point cloud target center point coordinates are corrected using all rigid transformation matrices, and the standard deviation of the corrected target center point coordinates and the coordinate true value data is calculated, and the rigid transformation matrix corresponding to the minimum standard deviation is the optimal coordinate transformation matrix.

[0032] In some other embodiments, the following technical solutions are adopted:

[0033] A tunnel vehicle-mounted laser radar positioning system, comprising:

[0034] A module for obtaining coordinate true values of target points arranged at a tunnel site;

[0035] A module for collecting tunnel point cloud data during vehicle driving through the tunnel by laser radar;

[0036] A module for extracting target point cloud from complete tunnel point cloud data and identifying the coordinates of the target center point;

[0037] ​A module for one-to-one correspondence between the coordinate true value of the target point and the identified target center point coordinate, and selecting a set number of points for registration operation through the method of random sampling and then traversing combination, to obtain the coordinate conversion matrix;

[0038] A module for calculating the target center point coordinate after correction based on the coordinate conversion matrix, calculating the standard deviation of each target center point coordinate after correction and the corresponding coordinate true value, selecting the coordinate conversion matrix corresponding to the minimum standard deviation, and performing laser radar positioning correction.

[0039] In some other embodiments, the following technical solutions are adopted:

[0040] A terminal device comprising a processor and a memory, the processor being configured to implement instructions, and the memory being configured to store a plurality of instructions adapted to be loaded and executed by the processor to implement the tunnel-mounted laser radar positioning correction method described above.

[0041] In some other embodiments, the following technical solutions are adopted:

[0042] A computer-readable storage medium having a plurality of instructions stored therein, the instructions being adapted to be loaded and executed by the processor of a terminal device to implement the tunnel-mounted laser radar positioning correction method described above.

[0043] Compared with the prior art, the present application has the following advantages:

[0044] (1) The ground target orientation recognition algorithm proposed by the present application automatically recognizes the target point features of the tunnel point cloud data, which is simple and efficient, high-precision, stable, and highly targeted.

[0045] Firstly, the present method first excludes isolated noise points, and if the local data is segmented and denoised, the lack of local sample data may lead to excessive denoising or insufficient local denoising, resulting in errors, and the precision is high.

[0046] Secondly, the preliminary segmentation method of the target point cloud of the present application adopts the method of taking the relative height range of each cross-sectional point cloud data relative to the ground, which is not affected by the vehicle driving track and the tunnel size compared with the method of screening and segmenting according to the radar emission angle, and can also exclude the influence of the change of elevation in the length direction of the tunnel, and has better accuracy and stability.

[0047] Then, the further segmentation of the target point cloud in the method is based on the spatial distribution characteristics of the segmented tunnel structure. The normal and curvature of each part are calculated, and the ground, curbstone and target point cloud are extracted through curvature screening; taking the ground point cloud features with the largest data amount as the benchmark, the point cloud data with the normal perpendicular to the ground normal direction and the curvature same as the ground curvature are screened, and the curbstone and target point cloud are extracted. This method can realize the directional and automatic acquisition of target data, and the calculation efficiency of the segmented tunnel for directly extracting target data can be improved by 2 times.

[0048] Finally, the method classifies the data according to the different laser reflection intensities of the curbstone and target with different materials; and according to the difference between the distribution areas of the two, the point cloud with small data amount is selected as the target point cloud. The above-mentioned method realizes the directional separation and extraction of the continuous point cloud of the target and the curbstone in a simple and clear way.

[0049] (2) The tunnel target center point automatic extraction algorithm is proposed to realize the automatic extraction of the target point center point coordinates, which has the advantages of automation, high precision and directional adaptation to vehicle-mounted scanning point cloud data.

[0050] According to the feature-recognized point cloud data, the target point center point coordinates are automatically extracted, the manual coordinate extraction operation is reduced, and human error is avoided.

[0051] (3) Under the premise of rectangular target recognition, the boundary barycenter method is proposed by improving the simple barycenter method. The outermost point cloud is selected to calculate the barycenter coordinates, i.e. the center point coordinates of the target, by extracting multi-layer boundary point cloud in a given angle range. The existing geometric coordinate extraction precision is poor; the weighted barycenter method is closely related to the point cloud intensity, and is more suitable for fixed data scanning. Mobile scanning will pass through the target, and the front and rear scanning results will affect each other, resulting in errors. The vehicle-mounted laser radar scanning point cloud data is mostly linear array point cloud, and the extracted target point cloud will have point cloud data missing and internal data inclined distribution at local positions such as corner points. The simple barycenter method for solving the center point coordinates will produce errors. Based on the above shortcomings, the boundary barycenter method is proposed. The outermost point cloud is selected as the boundary to calculate the center point coordinates by extracting multi-layer boundary point cloud in a given angle range. In this way, the uneven extraction of each boundary point cloud or the more missing data of a certain boundary can be avoided, and the high-precision coordinate data can be solved.

[0052] (4) The tunnel point cloud iterative correction algorithm is proposed, the overall target standard deviation is proposed as a registration algorithm effect measurement index, and three pairs of corresponding points with the best registration effect are selected as the basis by calculating and screening the target point standard deviation, so as to perform overall tunnel point cloud registration and correction. The above method quantifies the tunnel correction effect compared with the traditional single registration and correction method, has higher precision (the point cloud data positioning average precision is improved from 6-7 cm to 2 cm, and the precision is improved by 2 times), and has better stability; the uniformly distributed target points are used as sample data for traversal registration, which is more representative and efficient.

[0053] Other features and advantages of the present application will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 The tunnel vehicle-mounted laser radar positioning correction method in the embodiment of the present application is shown in the schematic diagram.

[0055] Figure 2 The target point cloud feature recognition process in the embodiment of the present application is shown in the schematic diagram.

[0056] Figure 3 The target point coordinate automatic extraction process in the embodiment of the present application is shown in the schematic diagram.

[0057] Figure 4 The target point coordinate iterative correction process in the embodiment of the present application is shown in the schematic diagram.

[0058] Figure 5 The target point coordinate iterative correction process in the embodiment of the present application is shown in the schematic diagram. Figure 2 The ground target point directional recognition result in the embodiment of the present application is shown in the schematic diagram.

[0059] Figure 6 The target boundary extraction result in the embodiment of the present application is shown in the schematic diagram. Figure 3 The target boundary extraction result in the embodiment of the present application is shown in the schematic diagram.

[0060] Figure 7 The error comparison analysis diagram before and after correction in the embodiment of the present application is shown in the schematic diagram. Figure 4 The error comparison analysis diagram before and after correction in the embodiment of the present application is shown in the schematic diagram. DETAILED DESCRIPTION

[0061] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in the present application have the same meaning as generally understood by those skilled in the art to which the present application belongs.

[0062] It is to be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, elements, components, and / or groups thereof, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof.

[0063] Embodiment One

[0064] In one or more embodiments, a tunnel vehicle-mounted laser radar positioning correction method is disclosed, which is combined with Figure 1 , including the following processes:

[0065] (1) Pre-processing, used to obtain the layout of the target points and the total station measurement target point coordinates;

[0066] (2) Data acquisition, used to obtain the point cloud data around the driving path;

[0067] (3) Target recognition, used to extract the target points in the tunnel point cloud;

[0068] (4) Target coordinate calculation, used to automatically calculate the point coordinates of the target center point;

[0069] (5) Iterative registration, randomly selecting three one-to-one point positions to register the point coordinates measured by the total station with the automatically extracted point coordinates, calculating the standard deviation of each point and filtering out the registration mode with the smallest standard deviation through the iterative algorithm, which is used as the correction basis to obtain the corrected point cloud data.

[0070] In combination with Figure 1 , the specific implementation process of the tunnel vehicle-mounted laser radar positioning correction method of the present embodiment is as follows:

[0071] S101: Obtain the coordinate true value of the target points laid in the tunnel site;

[0072] In the present embodiment, the target points are laid in the tunnel in advance according to the tunnel site conditions, so that the target points are uniformly distributed in the length direction of the tunnel. In the present embodiment, a square hard sheet (thickness <5mm, vehicle-mounted scanning point cloud point spacing >5mm) with a size of 60*60cm is selected as the target, which serves as a reference. The square grid is uniformly distributed on the hard sheet, and the greater the difference between the reflection intensity of the hard sheet and the reflection intensity of the road curb, the better. In order to improve the accuracy, the target point spacing is controlled to be about 100m, and the fixed mode is to be grounded and attached to the road curb. In order to accurately measure the target point coordinates, the total station station measurement return method is used to measure the target point center point, and through post-processing, the error is ensured to be within the allowable range, and the true and reliable target point coordinate true value is obtained.

[0073] S102: Collect tunnel point cloud data in the vehicle driving process by laser radar;

[0074] In this embodiment, point cloud data in the driving process is collected by laser radar. The point cloud data is collected and then calculated according to the path information. The tunnel point cloud is calculated according to the position, pose information and reflected laser point position information.

[0075] S103: Extract target point cloud from complete point cloud data and identify the coordinates of the target center point;

[0076] In this embodiment, extracting target point cloud from complete point cloud data means extracting the laid target point cloud from complete tunnel point cloud, combining Figure 2 , the specific process is as follows:

[0077] S1031: During the driving process, the device will cause disturbance in different directions of the measuring vehicle due to acceleration, deceleration and road undulation, resulting in isolated points, out-of-town points and local position mutations in the obtained point cloud data. Therefore, first, traverse the point cloud, calculate the average distance L i between each point and its nearest k neighbors; second, calculate the mean μ and standard deviation σ of all average distances, then the distance threshold d max can be represented as d max = μ + α × σ, α is a proportionality coefficient, which is determined according to the project requirements; finally, traverse the point cloud again, and remove the points whose average distance with k neighbors is greater than d max .

[0078] S1032: According to the previous processing work, the target point is laid close to the ground, so the ground point cloud is taken as the reference, and the point cloud data outside the range of 1m from the ground point cloud height of each tunnel section is removed, and the point cloud data including the ground, partial tunnel wall, target, curbstone and partial vehicle is retained.

[0079] S1033: Traverse the point cloud, calculate the normal and curvature of the point cloud obtained in S1032, and classify the point cloud data according to the two, and select the vertical surface point cloud data whose normal vector is perpendicular to the ground normal vector and whose curvature is the same as the ground curvature, to obtain the point cloud data containing the curbstone and target point.

[0080] S1034: Because the curbstone and the target have different materials, traverse the point cloud, and divide the point cloud data in S1033 into two groups of data according to the point cloud reflection intensity. The area of the curbstone on the tunnel side is much larger than the area of the target, so the point number is small, which is used as the target point cloud, to realize the feature recognition of the target. Figure 5 The ground target point directional recognition result is given.

[0081] In this embodiment,Figure 3 The coordinates of the center point of the identified target refer to the coordinates of the center point of the feature-identified target calculated by a built-in algorithm. A boundary barycenter method is proposed. First, an angle threshold range needs to be set to extract the boundary of the target point cloud, and after screening the multi-layer boundary point cloud, the outermost point cloud is obtained to obtain a fine target contour. Second, the average value of the four edges of the target is calculated to obtain four point coordinates. Finally, the center point coordinates are calculated by averaging the four points. Figure 6 The target boundary extraction result is given.

[0082] Among them, the boundary extraction of the target point is based on the spatial distribution relationship of the boundary points. Because the target point cloud obtained in the feature recognition process appears as a square facade in space, through experimental testing, the vector angle between adjacent boundary points is set to 90°-120° for better effect, and considering the discontinuity of point distribution, the distance threshold of adjacent points is controlled at twice the average point distance. According to the above conditions, the boundary point cloud is obtained by traversing the search.

[0083] The average value of the four edges of the target is calculated by selecting the outermost point cloud for coordinate average value calculation. During the boundary extraction process, there are multiple layers of point clouds and local concave-convex situations of point clouds, so the outer layer point cloud is selected for calculation. The missing place of the outermost point cloud is filled by the corresponding point cloud of the next outer layer to improve the calculation accuracy. The center point coordinates of each edge are calculated as follows:

[0084]

[0085] Among them, B i is a certain edge, d i is the coordinates of the outermost points of this edge, and N is the number of outermost points of this edge.

[0086] The calculation of the center point coordinates is to average the three-dimensional coordinates of the four points calculated above to obtain the three-dimensional coordinates of the center point.

[0087] S104: The coordinates of the target point are one-to-one corresponding to the coordinates of the center point of the identified target, and a certain number of points are selected for registration operation by random sampling method to obtain a coordinate conversion matrix.

[0088] In this embodiment, the Figure 4The coordinate true value of each target center point is corresponded to the point cloud center point coordinate, and three pairs of points are selected by random sampling method for registration processing. After this processing, the target point cloud center generates new coordinates, and the standard deviation is calculated between the new coordinates and the coordinate true value to determine the correction effect. It is found that different registration points have different registration correction effects of point cloud in the actual registration correction process, so the above steps are iterated to traverse all registration combination schemes, the optimal registration scheme is obtained by comparing the standard deviation, and then the whole point cloud data is registered and corrected according to the scheme to obtain the corrected data. Figure 7 The error comparison analysis diagram before and after correction is given.

[0089] The coordinate true value of the target point center is determined by the total station point measurement, and the point cloud center point coordinate is extracted from the above center point coordinate.

[0090] Three pairs of corresponding points are selected for registration, and the three pairs of points are divided into true value group and coordinate extraction group, and each group is three non-collinear points. Each group of points can establish a space coordinate system, and the relative position relationship between the two coordinate systems can be used to calculate the rigid transformation matrix of the measured point cloud and the coordinate true value, as follows.

[0091]

[0092] A 3*3 represents a rotation matrix, T 3*1 represents a translation vector (two three-dimensional coordinates are subtracted), V 1*3 represents a perspective transformation vector, and S represents the overall scale factor. Because the point cloud data only exists rotation and translation transformation, there is no deformation, so V is set to zero vector, and the scale factor S = 1.

[0093] Wherein, α, β, γ respectively represent the rotation angles of a space coordinate system relative to another space coordinate system along x, y, z axes, t x , t y , t z respectively represent the translation along x, y, z axes. Because the calculation of the matrix needs to calculate the above six unknown parameters, six linear equations are needed, and at least three groups of corresponding points are needed to solve.

[0094] The solving method of the above six unknown parameters is as follows:

[0095] M = M' * A 3*3 + T M ; N = N' * A 3*3 + T N ; P = P' * A 3*3 + T P ;

[0096] M, N, P and M', N', P' are respectively a true value group and a coordinate extraction group of three points, T M , T N , T P are respectively the coordinate difference values of the corresponding two points, the six parameters are solved by substituting the corresponding coordinates into the two two combinations respectively and comparing the parameter results, if the error is within the allowable range, the average value is taken as the solution of the six parameters and substituted into the H transformation matrix.

[0097] S105: calculate the center point coordinates of the target after correction based on the coordinate conversion matrix, calculate the standard deviation of each target center point coordinate after correction and the corresponding coordinate true value, select the coordinate conversion matrix corresponding to the minimum standard deviation, and perform laser radar positioning correction.

[0098] In this embodiment, the standard deviation of the coordinate true value and the registered coordinate is taken as the basis for the registration effect, and the calculation formula is as follows:

[0099]

[0100]

[0101] Wherein, S is the variance, σ is the standard deviation, n is the number of target points, X i is the true value of the target center coordinate, X i ' is the registered target center point coordinate.

[0102] In this embodiment, the registration process is to register the n target point center measured point coordinates and true value point coordinates.

[0103] After σ is calculated, the target point number n is arranged and combined, all combinations are traversed, and σ is solved each time. It is compared with the minimum standard deviation before, if it is less than the minimum standard deviation, it is replaced by the minimum standard deviation. After traversal, the coordinate conversion matrix H where the minimum standard deviation is obtained, and then the measured point cloud is corrected according to H.

[0104] In this embodiment, the target point true value and the automatically extracted target center point coordinates are corrected, three pairs of points are randomly extracted on the path for registration, the coordinate rigid transformation matrix is obtained, which is applied to each target point to obtain new target coordinates. The variance of each target point center after registration and the target point true value is calculated, the most accurate registration and correction scheme is selected through the iterative algorithm, and then the rigid transformation matrix under this combination is applied to the whole point cloud. In this way, the cumulative error of the process is minimized and the error is uniformly distributed to each point, avoiding error mutation phenomenon.

[0105] Embodiment two

[0106] In one or more embodiments, a tunnel vehicle-mounted laser radar positioning system is disclosed, specifically comprising:

[0107] a module for obtaining coordinate true values of target points arranged at a tunnel site;

[0108] a module for collecting tunnel point cloud data during vehicle driving by a laser radar;

[0109] a module for extracting target point cloud from the complete tunnel point cloud data and identifying coordinates of target center points;

[0110] a module for one-to-one correspondence of the coordinate true values of the target points and the identified coordinates of the target center points, and selecting a set number of points for registration operation by a method of random sampling first and then traversal combination to obtain a coordinate conversion matrix;

[0111] a module for calculating the coordinates of the target center points after correction based on the coordinate conversion matrix, calculating the standard deviation of each target center point coordinate after correction and the corresponding coordinate true value, selecting the coordinate conversion matrix corresponding to the minimum standard deviation, and performing laser radar positioning correction.

[0112] Embodiment Three

[0113] In one or more embodiments, a terminal device is disclosed, comprising a server, the server comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the tunnel vehicle-mounted laser radar positioning correction method in Embodiment One when executing the program. For brevity, this will not be repeated here.

[0114] It should be understood that in the embodiments, the processor can be a central processing unit CPU, and the processor can also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, ready-to-program gate arrays FPGA or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0115] The memory can include read-only memory and random access memory, and provide instructions and data to the processor, and a portion of the memory can also include non-volatile random access memory. For example, the memory can also store device type information.

[0116] In the implementation process, each step of the above method can be completed by integrated logic circuits of hardware in the processor or instructions in the form of software.

[0117] Embodiment Four

[0118] In one or more embodiments, a computer readable storage medium is disclosed, wherein a plurality of instructions are stored, the instructions being adapted to be loaded and executed by a processor of a terminal device to implement the tunnel vehicle-mounted laser radar positioning correction method described in Embodiment I.

[0119] The above describes the specific embodiments of the application in conjunction with the drawings, but is not a limitation on the scope of protection of the application. Those skilled in the art should understand that various modifications or changes made by those skilled in the art on the basis of the technical solutions of the application without creative labor are still within the scope of protection of the application.

Claims

1. A method for positioning and correcting deviations using a vehicle-mounted lidar in a tunnel, characterized in that, include: Obtain the true coordinates of the target points deployed at the tunnel site; Point cloud data of tunnels during vehicle movement is collected using lidar; Extract the target point cloud from the complete tunnel point cloud data and identify the coordinates of the target center point; The true coordinates of the target points are matched one-to-one with the coordinates of the target center point obtained by identification. A set number of points are selected by random sampling and then traversal combination to perform registration operation and obtain the coordinate transformation matrix. The coordinates of the target center point after correction are calculated based on the coordinate transformation matrix. The standard deviation of the coordinates of each target center point after correction and the corresponding true coordinate value are calculated. The coordinate transformation matrix corresponding to the smallest standard deviation is selected for lidar positioning correction. A boundary centroid method is proposed to extract the coordinates of the target center point, including: The outermost boundary of the target is extracted to obtain a fine target profile; The average value of the four sides of the target is used to obtain the coordinates of four points; The coordinates of the center point are obtained by averaging the coordinates of the four points. The optimal coordinate transformation matrix is ​​obtained by selecting a set number of points through a combination of random sampling and iterative traversal, specifically including: Each time, at least three pairs of points are randomly selected to iteratively traverse the point cloud data for registration. The point cloud data includes the true coordinate data set of the target points and the coordinate data set of the target center point. A spatial coordinate system is established for each set of data. Based on the relative positional relationship between the two coordinate systems, the rigid transformation matrix for registration between the measured point cloud and the true coordinate values ​​is obtained. Correct the coordinates of the center point of all measured point cloud targets using all rigid transformation matrices. Calculate the standard deviation of the corrected target center point coordinates with the true coordinate data. The rigid transformation matrix corresponding to the minimum standard deviation is the optimal coordinate transformation matrix. Extracting the target point cloud from the complete point cloud data specifically includes: Preprocess the point cloud data to remove abnormal point cloud data; The point cloud data is classified according to the normal and curvature of the point cloud data, and the elevation point cloud data with the normal vector perpendicular to the ground normal vector is selected to obtain the point cloud data including the curb stone and the target point. Based on the point cloud reflection intensity, the point cloud data containing the curbstone and the target point are divided into two groups, and the group with fewer point clouds is extracted as the target point cloud. Preprocessing of point cloud data to remove abnormal point cloud data includes: Iterate through all point cloud data and calculate the average distance L between each point and its k nearest neighbors. i ; Calculate the mean and standard deviation of all average distances, and then determine the distance threshold d based on the mean and standard deviation. max ; Iterate through the point cloud again, removing points whose average distance to their k neighboring points is greater than d. max point; Preprocessing of point cloud data to remove outlier point cloud data also includes: Based on the road surface point cloud, point cloud data outside the set distance range relative to the ground point cloud are removed.

2. The tunnel vehicle-mounted lidar positioning and correction method as described in claim 1, characterized in that, The data collected via lidar during vehicle movement includes tunnel point cloud data, specifically: Point cloud data is collected by lidar during the driving process. The collected point cloud data is then calculated based on the path information, the driving position and pose information, and the position information of the reflected laser points to obtain the tunnel point cloud data.

3. A tunnel vehicle-mounted lidar positioning system, based on the tunnel vehicle-mounted lidar positioning and correction method as described in claims 1-2, characterized in that, include: Module used to obtain the true coordinate values ​​of target points deployed at the tunnel site; Module used to collect tunnel point cloud data during vehicle movement using lidar; Used to extract target point cloud from complete tunnel point cloud data and identify the coordinates of the target center point; This module is used to map the true coordinates of target points to the coordinates of the identified target center point, and to select a set number of points for registration by means of random sampling and traversal combination to obtain the coordinate transformation matrix. This module is used to calculate the coordinates of the target center point after correction based on the coordinate transformation matrix, calculate the standard deviation of the coordinates of each target center point after correction and the corresponding true coordinate value, and select the coordinate transformation matrix corresponding to the smallest standard deviation for laser radar positioning correction.

4. A terminal device comprising a processor and a memory, wherein the processor implements various instructions; and the memory stores multiple instructions, characterized in that, The instructions are adapted to be loaded by a processor and executed as described in any one of claims 1-2, for the tunnel vehicle-mounted lidar positioning and correction method.

5. A computer-readable storage medium storing a plurality of instructions, characterized in that, The instructions are adapted to be loaded by the processor of the terminal device and executed according to any one of claims 1-2, the tunnel vehicle-mounted lidar positioning and correction method.

Images