Ellipse-like trajectory based airborne laser radar zeroing angle error correction method

Abstract

The application discloses a zero-return angle error correction method based on an elliptical trajectory, belongs to the technical field of laser radar measurement, and is used for correcting the zero-return angle error of an airborne laser radar, and comprises the following steps: after fixing and installing the airborne laser radar, scanning a vertical plane wall, collecting original point clouds, and establishing a coordinate system; the original point clouds are denoised, fitted, clustered, and the centroid is calculated to determine an error-free symmetry reference; the influence of the zero-return angle error on the symmetry of the trajectory is analyzed, the geometric relationship of the scanning light is deduced, and the correlation model of the error and the centroid deviation is deduced; the centroid X-axis coordinate is taken as an observation quantity, the final optimal zero-return angle estimation value is obtained through target function optimization, cost function optimization and multi-frame weighted average; and the point cloud coordinates are corrected based on the final optimal zero-return angle estimation value, and the accuracy is guaranteed through iteration verification. The application corrects the error based on the elliptical trajectory, has high accuracy and strong robustness, effectively improves the reliability of the airborne laser radar point cloud data, and is suitable for high-precision calibration scenes.

Description

TECHNICAL FIELD

[0001] The application discloses a zero angle error correction method based on an elliptical trajectory for an airborne laser radar and belongs to the technical field of airborne laser radar measurement. BACKGROUND

[0002] In recent years, airborne laser depth sounding radars (LiDARs) have been widely used in the fields of marine topographic mapping, environmental monitoring, underwater target detection and the like. Many airborne laser radars use a wedge-shaped rotating mirror as a scanning mechanism, and a light beam emitted by a laser is deflected by the rotation of the wedge-shaped prism to scan a target area. The laser footprint trajectory formed by this scanning mode is an elliptical trajectory, that is, an elliptical scanning path is presented on a target plane (such as the ground or a wall surface). However, in the elliptical scanning mode, the incidence angles of the laser at different positions on the trajectory are different, and it is necessary to calibrate the incidence angle of each scanning footpoint in advance. This not only increases the workload of the airborne laser radar in specific applications (such as underwater target detection) in use, but also increases the complexity of the back-end data calculation. In actual applications, optical and installation errors of the airborne laser radar will cause systematic deviations in the point cloud measurement data. One important error source is the zero angle error of the laser radar, that is, the deviation of the scanning optical zero point (reference zero angle) from the ideal reference. Due to manufacturing and assembly errors or improper installation adjustment, the starting angle of the laser radar scanning may be deviated, causing the scanning trajectory to be deviated or offset relative to the device coordinate system. If the zero angle error is not corrected, systematic errors will occur between the laser point cloud coordinates and the actual coordinates, thereby reducing the accuracy of point cloud positioning and ranging, and being not conducive to subsequent data fusion with an inertial navigation system and the like.

[0003] At present, the correction of the installation angle error of the laser radar mainly depends on artificial experience adjustment and repeated trial and error. The usual practice is to manually adjust the installation angle of the laser radar several times to align the collected point cloud with the reference true value. A correction method for calculating the installation angle error of a vehicle-mounted laser radar is proposed in the prior art, which uses a calibration object to assist in calculating the angle error, places two targets that are symmetrical left and right on the same plane and at the same height, and calculates the three-axis angle error of the laser radar by comparing the point cloud coordinates with the actual coordinates. This method uses the known actual positions of the symmetrical targets to solve the deviation once through a geometric model, thereby reducing the repeatability of pure manual adjustment. However, this method needs additional target arrangement and accurate measurement equipment, and it is relatively complex to implement for an airborne laser radar, and still needs manual participation in data collection and calculation. SUMMARY

[0004] The purpose of the present application is to provide an airborne laser radar zero angle error correction method based on an elliptical trajectory, to solve the problem that the existing airborne laser radar angle (zero angle) correction method depends on manual precision, has a complicated process, and the calibration accuracy is limited by human factors or requires additional equipment.

[0005] The airborne laser radar zero angle error correction method based on an elliptical trajectory comprises:

[0006] S1, install the airborne laser radar at a fixed position, scan and collect original point clouds of a vertical plane wall target, and establish a coordinate system;

[0007] S2, denoise, plane fitting, screening, protruding point cloud clustering, and centroid coordinate calculation are performed on the original point clouds to determine the symmetry reference under the condition of no zero angle error;

[0008] S3, based on the symmetry reference, analyze the influence of the zero angle error on the symmetry of the trajectory, obtain the geometric relationship of the scanning light and the correlation model of the error and the centroid deviation;

[0009] S4, taking the X-axis value of the centroid coordinate as the observation, the final optimal zero angle estimation value is obtained through target function optimization, cost function optimization, and multi-frame weighted average;

[0010] S5, based on the final optimal zero angle estimation value, the laser radar point cloud is corrected, and the calibration accuracy is ensured to meet the requirements through iteration verification.

[0011] S1 comprises installing the airborne laser radar at a fixed position, selecting a vertical plane wall parallel to the laser radar scanning plane as the scanning target, vertically fixing a rod-shaped object on the vertical plane wall as a calibration mark, controlling the laser radar to scan the wall and the rod-shaped object, and collecting original point clouds containing wall plane point clouds and rod-shaped object point clouds; a coordinate system O-XYZ is established with the optical center of the laser radar as the origin, wherein the X-axis is vertically upward, the Y-axis is along the extension direction of the wall surface, and the Z-axis points to the inside of the wall surface.

[0012] S2 comprises S2.1, filtering the original point clouds collected in S1 using a statistical outlier removal filtering algorithm to filter out outliers in the original point clouds, improve the quality of the point cloud data, and the original point clouds are:

[0013] ;

[0014] wherein, is the original point cloud, is the kth protruding point cloud the coordinate in the X-axis direction, the kth protruding point cloud the coordinate in the Y-axis direction, For the k-th protrusion point cloud Coordinates in the Z-axis direction;

[0015] S2.2, establish the wall plane equation:

[0016] ;

[0017] ;

[0018] wherein, is the normal vector, is the point cloud coordinate vector, is the three-dimensional point cloud coordinate, , is the nominal distance of the laser to the wall surface;

[0019] S2.3, using the random sampling consistency plane fitting algorithm to the point cloud data filtered in S2.1 for plane fitting, detecting the plane where the wall surface is located, setting the plane point determination threshold , the points satisfying are identified as wall plane points, and the rest are non-plane points;

[0020] S2.4, calculate the normal distance of the non-plane point to the wall surface plane:

[0021] ;

[0022] Set the distance threshold , filter out the deviation points , cluster the deviation points by clustering algorithm, and get two clusters of protrusion point clouds:

[0023] ;

[0024] ;

[0025] In the formula, and are two clusters of protrusion point clouds, is the th element in , is the th element in , the two clusters of protrusion point clouds correspond to the point clouds formed when the laser class elliptical trajectory sweeps the rod on both sides respectively;

[0026] S2.5, respectively calculate the centroid coordinate vectors and of the two clusters of protrusion point clouds:

[0027] ;

[0028] ;

[0029] wherein, is a three-dimensional coordinate of, is a three-dimensional coordinate of, and are respectively and the number of midpoint clouds, and are respectively variables starting from the initial value 1, and when the error is determined to be zero, the centroid satisfies and the symmetry reference of in the XOY plane.

[0030] The scanning light ray geometric relationship includes:

[0031] S3.1, when the zero-return angle error is clear, the ellipsoidal-like scanning trajectory of the laser radar and the rod-shaped protruding point cloud are both symmetrical about the Y axis in the XOY plane; the zero-return angle error is equivalent to a small rotation deviation angle of the scanning plane of the laser radar around the Z axis, the wall surface at a distance of from the radar causes a trajectory center offset of

[0032] S3.2, the direction angle of the horizontal plane projection of the scanning light ray of the laser radar is defined as , and the unit direction vector of a bundle of outgoing light in the ellipsoidal-like scanning trajectory is:

[0033] ;

[0034] Let the zero-return angle error of the laser radar around the Z axis be , and the actual outgoing direction is:

[0035] ;

[0036] ;

[0037] wherein, is the rotation matrix introduced around the Z axis:

[0038] When the wall surface distance is , the intersection of the light ray and the wall surface plane satisfies:

[0039] ;

[0040] ;

[0041] wherein, is the parameter of the intersection of the light ray and the wall surface, is the projection component of the exit direction vector in the Y-axis direction; and exists , the projection component of the exit direction vector in the X-axis direction.

[0042] Further, the X coordinate of the projection in the XOY plane is obtained as :

[0043] ;

[0044] wherein, is the projection component of the exit direction vector in the X-axis direction. and exists , the projection component of the exit direction vector in the X-axis direction. The correlation model of the error and the centroid deviation includes establishing

[0045] a quantitative relationship with the protrusion point cloud deviation;

[0046] S3.3, when there is no , the parameters of the scanning light ray corresponding to the vertical rod scanned out satisfy the following equation: and

[0047] ;

[0048] wherein, is the projection value of the scanning light ray X-axis when there is no , the corresponding scanning angle ; is the projection value of the scanning light ray X-axis when there is no , the corresponding scanning angle ;

[0049] When there is , the horizontal coordinates of the two clusters of protrusion point cloud centroids are:

[0050] ;

[0051] ;

[0052] wherein, is the projection value of the scanning light ray X-axis when there is , the corresponding scanning angle ; is the projection value of the scanning light ray X-axis when there is , the corresponding scanning angle ;​​​

[0053] The centroids of the two clusters of protrusion points satisfy the following equation:

[0054] ;

[0055] wherein, and are the X-axis projection values of the two clusters of protrusion points respectively when

[0056] The solution of the above equation is transformed into finding the rotation angle that makes approximate to zero, and through first-order linearization, we get:

[0057] ;

[0058] ;

[0059] wherein, and are the X-axis coordinates of the centroids of the two clusters of protrusion points respectively when and are constant coefficients determined by the scanning geometry.

[0060] The S4 includes S4.1, one-dimensional optimization of the angle error with the verticality of all the rod protrusion point clouds after rotation as the optimization criterion, and the set of all rod protrusion point clouds is defined as:

[0061] ;

[0062] wherein, is a single protrusion point cloud in the set of protrusion point clouds is the X-axis coordinate of the kth protrusion point cloud is the Y-axis coordinate of the kth protrusion point cloud is the Z-axis coordinate of the kth protrusion point cloud When using the two centroid vectors and , for a given zeroing angle compensation value , the rotation gives the centroid X coordinate as:

[0063] ;

[0064] ;

[0065] ;

[0066] wherein,​​​​​ For given X coordinate after rotation, For given X coordinate after rotation;

[0067] Define the objective function :

[0068] ;

[0069] Take as the initial value, bring into to replace , and get the optimal zeroing angle estimate value through one-dimensional search .

[0070] The S4 includes S4.2, based on all protrusion point cloud sets , calculate the X coordinate after rotation of each protrusion point cloud under the given optimal zeroing angle estimate value for correction calculation:

[0071] ;

[0072] Wherein, is the corrected coordinate in the X-axis direction of the protrusion point cloud set after given rotation;

[0073] Calculate the average value of the corrected protrusion point cloud in the X direction:

[0074] ;

[0075] Wherein, is the average value of the corrected coordinate in the X-axis direction of the protrusion point cloud set after given rotation, is the number of points;

[0076] The cost function is:

[0077] ;

[0078] Solve the optimal zeroing angle error that makes the minimum, and the variance of the protrusion point cloud in the X direction is the minimum:

[0079] .

[0080] The S4 comprises: S4.3, calculating the optimal zero-return angle error of each frame in the multi-frame point cloud respectively , setting the weight related to the number of rod points or fitting residual of each frame , obtaining the final optimal zero-return angle error estimation value by weighted average through the following formula :

[0081] ;

[0082] wherein, is the final optimal zero-return angle error estimation value.

[0083] The S5 comprises: S5.1, writing the optimal zero-return angle error of each frame obtained by the S4 into the point cloud fusion program , and correcting the coordinates of the protrusion point cloud fused by the laser radar, and the correction formula is:

[0084] ;

[0085] wherein, is the protrusion point cloud after correction of the kth protrusion point cloud;

[0086] The projection of the corrected elliptical-like scanning track on the wall surface is symmetrical about the Y axis, and the two clusters of protrusion point clouds are located on the same straight line perpendicular to the YOZ plane in the X direction.

[0087] The S5 comprises: S5.2, after the compensation correction is completed, the same wall and rod are scanned again, and the horizontal difference of the two clusters of protrusion point cloud centroids is recalculated according to S2 to S4:

[0088] ;

[0089] The calibration termination threshold is set , if , it is determined that the zero-return angle calibration is completed; if it is not satisfied, the horizontal difference is taken as a new initial error, and the S4 to S5.1 is repeated until the accuracy requirement of is met.

[0090] Compared with the prior art, the beneficial effects of the present application are: the present application makes full use of the geometric characteristics contained in the laser radar itself elliptical-like scanning track, and through feature extraction, geometric modeling and parameter optimization of the point cloud of the static calibration scene, the zero-return angle error of the laser radar can be estimated and compensated without complex external precision devices, thereby significantly improving the positioning accuracy and consistency of the airborne point cloud. BRIEF DESCRIPTION OF DRAWINGS

[0091] Figure 1 is the technical flowchart of the present application; ​

[0092] Figure 2 Fig. 1 is a diagram of a laser scanning platform coordinate system;

[0093] Figure 3 Fig. 2 is a diagram of an initial scanning trajectory point cloud collected;

[0094] Figure 4 Fig. 3 is a diagram of a protrusion point cloud after denoising and wall surface fitting;

[0095] Figure 5 Fig. 4 is a diagram of a corrected point cloud trajectory. DETAILED DESCRIPTION

[0096] The application will be further described below in conjunction with the embodiments. The embodiments are only used to illustrate the application, and are not intended to limit the application in any way. It should be clear that the described embodiments are only some of the embodiments of the present application, and are not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0097] The zero angle error correction method based on the elliptical trajectory of the airborne laser radar comprises:

[0098] S1, install the airborne laser radar at a fixed position, scan and collect the original point cloud of the vertical plane wall target, and establish a coordinate system;

[0099] S2, denoise, plane fitting and screening, protrusion point cloud clustering and centroid coordinate calculation are performed on the original point cloud to determine the symmetry reference under the condition of no zero angle error;

[0100] S3, based on the symmetry reference, analyze the influence of the zero angle error on the symmetry of the trajectory, obtain the geometric relationship of the scanning light and the correlation model of the error and the centroid deviation;

[0101] S4, taking the centroid coordinate X-axis value as the observation, the final optimal zero angle estimation value is obtained through target function optimization, cost function optimization and multi-frame weighted average;

[0102] S5, based on the final optimal zero angle estimation value, the laser radar point cloud is corrected, and the calibration accuracy is ensured to meet the requirements through iteration verification.

[0103] The S1 comprises installing an airborne laser radar at a fixed position, selecting a vertical plane wall parallel to the scanning plane of the laser radar as a scanning target, vertically fixing a rod-shaped object on the vertical plane wall as a calibration mark, controlling the laser radar to scan the wall and the rod-shaped object, and collecting original point clouds containing wall plane point clouds and rod-shaped object point clouds.

[0104] The S2 comprises: S2.1, filtering the original point clouds collected in the S1 by using a statistical outlier removal filtering algorithm to filter out outlier noise points in the original point clouds, and improving the quality of the point cloud data, wherein the original point clouds are:

[0105] ;

[0106] wherein, is the original point cloud, is the kth protrusion point cloud the coordinate in the X-axis direction, the kth protrusion starting point cloud the coordinate in the Y-axis direction, is the kth protrusion point cloud the coordinate in the Z-axis direction;

[0107] S2.2, establishing a wall plane equation:

[0108] ;

[0109] ;

[0110] wherein, is the normal vector, is the point cloud coordinate vector, is the three-dimensional point cloud coordinate, , is the nominal distance of the laser to the wall;

[0111] S2.3, performing plane fitting on the point cloud data filtered in the S2.1 by using a random sample consensus plane fitting algorithm, detecting the plane where the wall is located, setting a plane point determination threshold , and identifying the points satisfying as wall plane points, and the rest as non-plane points;

[0112] S2.4, calculating the normal distance of the non-plane points to the wall plane:

[0113] ;

[0114] setting a distance threshold Filter out The deviation points were clustered using a clustering algorithm to obtain two clusters of protruding point clouds:

[0115] ;

[0116] ;

[0117] In the formula, and They are two clusters of clouds that suddenly appear. yes The first in One element, yes The first in Each element, the two clusters of protruding point clouds respectively correspond to the point clouds formed when the laser-like elliptical trajectory sweeps across the rod-shaped object on both sides;

[0118] S2.5 Calculate the centroid coordinate vectors of the two clusters of originating clouds respectively. and :

[0119] ;

[0120] ;

[0121] in, yes The three-dimensional coordinates yes The three-dimensional coordinates and They are respectively and The number of midpoint clouds, and Let the variables be variables that increase from an initial value of 1. When there is no error, the centroid in the XOY plane satisfies the following condition: and The symmetric reference.

[0122] The geometric relationships of the scanning rays include:

[0123] S3.1. When there is no zero-return angle error, the elliptical scanning trajectory of the lidar and the cloud of rod-shaped protrusions are symmetrical about the Y-axis in the XOY plane; the zero-return angle error is equivalent to a small rotational deviation angle of the lidar scanning plane about the Z-axis. Range radar The wall at that location produce The trajectory center offset;

[0124] S3.2, Define the direction angle of the laser radar scanning light beam projected onto the horizontal plane as . The unit direction vector of an outgoing beam in an elliptical scanning trajectory for:

[0125] ;

[0126] Let the zero-homing angle error of the lidar around the Z-axis be... Actual launch direction for:

[0127] ;

[0128] ;

[0129] in, It introduces a rotation matrix around the Z-axis:

[0130] The distance between the walls is The point where the light rays intersect the wall surface. satisfy:

[0131] ;

[0132] ;

[0133] In the formula, These are the parameters of the point where the light intersects the wall. The direction angle is And exist At that time, the outgoing direction vector Projection component in the Y-axis direction;

[0134] This allows us to obtain the X coordinate projected onto the XOY plane. :

[0135] ;

[0136] in, The direction angle is And exist At that time, the outgoing direction vector Projection component in the X-axis direction.

[0137] Obtaining a correlation model between error and centroid deviation includes establishing Quantitative relationship with the cloud deviation at the point of origin;

[0138] S3.3, Clearly none At that time, the parameters of the scanning light corresponding to the vertical bar being scanned. and Satisfy the following equation:

[0139] ;

[0140] where, is no , the corresponding scanning angle is the projection value of the scanning light ray X-axis; is no , the corresponding scanning angle is the projection value of the scanning light ray X-axis;

[0141] When there are , the horizontal coordinates of the two clusters of protrusion point cloud centers are:

[0142] ;

[0143] ;

[0144] where, is no , the corresponding scanning angle is the projection value of the scanning light ray X-axis; is no , the corresponding scanning angle is the projection value of the scanning light ray X-axis;

[0145] The centers of the two clusters of protrusion point clouds satisfy the following equation:

[0146] ;

[0147] where, and is no , the projection value of the X-axis corresponding to the two clusters of protrusion point clouds respectively;

[0148] The solution of is transformed into finding the rotation angle that makes approximate zero, and through first-order linearization, we get:

[0149] ;

[0150] ;

[0151] where, and are the X-axis coordinates of the centers of the two clusters of protrusion point clouds when there are no , and and are constant coefficients determined by the scanning geometry.

[0152] The S4 includes S4.1, one-dimensional optimization of the angle error with the verticality of all the rod protrusion point clouds after rotation as the optimization criterion, and the set of all rod protrusion point clouds is defined as:

[0153] ;

[0154] wherein, is a single protrusion point cloud in the set of protrusion point clouds ; is the kth protrusion point cloud coordinate in the X-axis direction, is the kth protrusion point cloud coordinate in the Y-axis direction, is the kth protrusion point cloud coordinate in the Z-axis direction;

[0155] When using two centroid vectors and , for a given zeroing angle compensation value , the rotation results in a centroid X coordinate of:

[0156] ;

[0157] ;

[0158] wherein, is the coordinate in the X-axis direction after rotation with a given ; is the coordinate in the X-axis direction after rotation with a given ;

[0159] The objective function is defined as:

[0160] ;

[0161] With as the initial value, bring into to replace , and through one-dimensional search, the optimal zeroing angle estimation value is obtained.

[0162] The S4 includes, S4.2, based on the set of all protrusion point clouds , calculate the X coordinate after rotation of each protrusion point cloud under a given optimal zeroing angle estimation value for correction calculation:

[0163] ;

[0164] wherein, is the corrected coordinate in the X-axis direction after rotation of the set of protrusion point clouds with a given ;

[0165] The average value of the corrected protrusion point cloud in the X direction is calculated:

[0166] ;

[0167] Wherein, is the set of protrusion point clouds Given the average value of the corrected coordinates in the X-axis direction, is the number of points;

[0168] The cost function is:

[0169] ;

[0170] Solve the optimal zero-angle error that makes the variance of the protrusion point cloud in the X direction minimum:

[0171] .

[0172] S4 includes, S4.3, respectively calculating the optimal zero-angle error of each frame in the multi-frame point cloud , setting the weight related to the number of rod points or fitting residual of each frame , weighted average by the following formula to get the final optimal zero-angle error estimate value :

[0173] ;

[0174] Wherein, is the final optimal zero-angle error estimate value.

[0175] S5 includes, S5.1, writing the obtained by S4 into the point cloud fusion program to correct the coordinates of the protrusion point cloud fused by the laser radar, and the correction formula is:

[0176] ;

[0177] Wherein, is the kth protrusion point cloud after correction of the protrusion point cloud;

[0178] The projection of the corrected elliptical-like scanning trajectory on the wall surface is symmetrical about the Y axis, and the two clusters of protrusion point clouds are located on the same straight line perpendicular to the YOZ plane in the X direction.

[0179] The S5 includes, S5.2, after compensation correction is completed, the same wall and rod are scanned again, and the horizontal difference of two clusters of protrusion point cloud is recalculated according to S2 to S4:

[0180] ;

[0181] Set the calibration termination threshold If , it is determined that the zero return angle calibration is completed; if not, the horizontal difference is taken as a new initial error, and S4 to S5.1 are repeated until the accuracy requirement of is met.

[0182] The technical flowchart of the present application is shown in Figure 1 , including scanning track point cloud acquisition, scanning track geometric feature analysis, scanning track geometric feature analysis including original point cloud denoising, fitting, clustering, and centroid calculation, and scanning track point cloud track geometric modeling, then establishing the relationship between the zero return angle and the coordinates, constructing the objective function, and optimizing the solution , geometric modeling is solved, the optimal is obtained, error compensation is performed, and it is determined whether the track point cloud is qualified, if not, the optimization solution is returned , if yes, the correction is completed. To verify the zero return angle correction performance of the present application, a self-developed airborne laser radar is used for experimental test, the test site is comprehensive test, first, the airborne laser radar is installed at a fixed position, a vertical plane wall target is scanned to acquire original point cloud, and the wall and the scanning plane of the laser radar are parallel, and a rod-shaped object is vertically fixed on the wall surface as a calibration mark, the original point cloud obtained includes the plane point cloud of the wall and the point cloud of the rod-shaped object, as shown in Figure 2 , the coordinate system O-XYZ established with the optical center of the laser radar as the origin, the X axis is vertically upward, the Y axis is along the extension direction of the wall surface, and the Z axis points to the inside of the wall surface. It can be found from Figure 3 that the original point cloud without filtering, denoising, and wall fitting is not easy to observe the protrusion, and it can be seen that the two protrusion point clouds are not on the vertical straight line perpendicular to the YOZ plane.

[0183] Figure 4 is the protrusion point cloud diagram after denoising and wall fitting, it is found that the protrusion point cloud in the point cloud data after filtering, denoising, and wall fitting is more easily observed, and then, part of the protrusion point cloud data of the two clusters of protrusion point clouds is extracted as shown in Table 1 and Table 2.

[0184] Table 1 is part of the point cloud data of the point cloud cluster 1;

[0185] .

[0186] Table 2 is part of the point cloud data of the point cloud cluster 2;

[0187] .

[0188] Point cloud cluster 1 and point cloud cluster 2 are respectively selected 20 representative point cloud data (x, y, z) as shown in Table 1 and Table 2, and the centroid coordinates of the protrusion point cloud 1 are calculated according to the centroid coordinate calculation formula: (-1.343526, 1.699998, -9.851170), and the centroid coordinates of the protrusion point cloud 2 are calculated according to the centroid coordinate calculation formula: (0.697132, 2.370394, -9.716789). , .

[0189] Next, the centroid zero angle of point cloud cluster 1 and point cloud cluster 2 is calculated, and the centroid zero angle is 152.65°. The optimal zero angle is obtained by iteratively calculating the zero angle of a single point cloud, and the centroid zero angle is 152.62°. There is a small deviation between the calculated centroid zero angle and the zero angle obtained by iterative calculation, and the deviation is corrected by Figure 5 The corrected point cloud trajectory diagram shows that after the zero angle error correction, the point cloud trajectory returns to the ideal scanning mode, and the correction effect meets the requirements of verticality and trajectory regularity.

[0190] Of course, the above description is not a limitation of the present application, and the present application is not limited to the above examples. Changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present application should also be within the scope of the present application.

Claims

1. A method for correcting the zero-point angle error of an airborne lidar based on a quasi-elliptical trajectory, characterized in that, include: S1. Install the airborne lidar in a fixed position to scan and collect the original point cloud of the vertical plane wall target, and establish a coordinate system; S2. Denoise the original point cloud, perform plane fitting and screening, cluster the point cloud and calculate the centroid coordinates to determine the symmetric reference under the condition of no zero-return angle error; S3. Based on the symmetric benchmark, analyze the influence of the zero-angle error on the trajectory symmetry, and obtain the geometric relationship of the scanning ray and the correlation model between the error and the centroid deviation; S4. Using the centroid coordinate X-axis value as the observation, the final optimal zero-point angle estimate is obtained through objective function optimization, cost function optimization, and multi-frame weighted averaging. S4 includes S4.1, which uses the verticality of all rod protrusion point clouds after rotation as the optimization criterion to perform one-dimensional optimization of the angle error, and defines the set of all rod protrusion point cloud as: ； in, and They are two clusters of clouds that suddenly appear. For the sudden start of cloud collection A single protrusion point cloud within; For the kth protrusion point cloud Coordinates in the X-axis direction For the kth protrusion point cloud The coordinates in the Y-axis direction, For the kth protrusion point cloud Coordinates in the Z-axis direction; Using two centroid vectors and At that time, for a given zero-angle compensation value The centroid X coordinate is obtained by rotation: ； ； in, for Given The coordinates in the X-axis direction after rotation. for Given The coordinates in the X-axis direction after rotation; Define the objective function : ； by As the initial value, Bring into China alternative The optimal zero-angle estimate is obtained through a one-dimensional search. ; S4 includes S4.2, based on the set of all protrusion point clouds. Calculate the estimated value of the cloud at each abrupt initiation point, given the optimal zero-return angle. The X-coordinate after downward rotation is corrected and calculated: ； in, For the sudden start of cloud collection Given Corrected coordinates in the X-axis direction after rotation; Calculate the average value of the corrected abrupt inflection point cloud in the X direction: ； in, For the sudden start of cloud collection Given The average value of the corrected coordinates in the X-axis direction. Points; Cost function for: ； Solve Minimum optimal zero-point error The variance of the cloud at the point of protrusion is the smallest in the X direction: ； S4 includes S4.3, calculating the optimal zero-point angle error for each frame in the multi-frame point cloud. Set weights related to the number of poles or the fitted residuals in each frame. The final optimal zero-angle error estimate is obtained by weighting the values ​​using the following formula. : ； in, This is the final optimal zero-angle error estimate; S5. Based on the final optimal zero-point angle estimate, the coordinates of the lidar point cloud are corrected, and the calibration accuracy is ensured to meet the requirements through iterative verification.

2. The method for correcting the zero-point angle error of airborne lidar based on a quasi-elliptical trajectory according to claim 1, characterized in that, S1 includes installing an airborne lidar in a fixed position, selecting a vertical plane wall parallel to the lidar's scanning plane as the scanning target, vertically fixing a rod-like object on the vertical plane wall as a calibration mark, controlling the lidar to scan the wall and the rod-like object, and acquiring an original point cloud containing the point cloud of the wall surface and the point cloud of the rod-like object; establishing a coordinate system O-XYZ with the optical center of the lidar as the origin, where the X-axis is vertically upward, the Y-axis extends along the wall surface, and the Z-axis points inward towards the wall surface.

3. The method for correcting the zero-point angle error of airborne lidar based on a quasi-elliptical trajectory according to claim 1, characterized in that, S2 includes: S2.1, applying a statistical outlier removal filtering algorithm to the original point cloud collected in S1 to filter out outliers and improve the quality of the point cloud data. The original point cloud is: ； in, The original point cloud, For the kth protrusion point cloud Coordinates in the X-axis direction The kth burst origin cloud The coordinates in the Y-axis direction, For the kth protrusion point cloud Coordinates in the Z-axis direction; S2.2 Establish the wall plane equation: ； ； in, It is the normal vector. The point cloud coordinate vector, For 3D point cloud coordinates, , The nominal distance at which the radar reaches the wall; S2.

3. The random sampling consistency plane fitting algorithm is used to perform plane fitting on the point cloud data after filtering in S2.1 to detect the plane where the wall is located and set the plane point judgment threshold. , will satisfy The points identified are wall plane points, and the rest are non-plane points; S2.4 Calculate the normal distance from the non-planar point to the wall plane: ； Set distance threshold Filter out The deviation points were clustered using a clustering algorithm to obtain two clusters of protruding point clouds: ； ； In the formula, and They are two clusters of clouds that suddenly appear. yes The first in One element, yes The first in Each element, the two clusters of protruding point clouds respectively correspond to the point clouds formed when the laser-like elliptical trajectory sweeps across the rod-shaped object on both sides; S2.5 Calculate the centroid coordinate vectors of the two clusters of originating clouds respectively. and : ； ； in, yes The three-dimensional coordinates yes The three-dimensional coordinates and They are respectively and The number of midpoint clouds, and Let the variables be variables that increase from an initial value of 1. When there is no error, the centroid in the XOY plane satisfies the following condition: and The symmetric reference.

4. The method for correcting the zero-point angle error of airborne lidar based on a quasi-elliptical trajectory according to claim 1, characterized in that, The geometric relationships of the scanning rays include: S3.

1. When there is no zero-return angle error, the elliptical scanning trajectory of the lidar and the cloud of rod-shaped protrusions are symmetrical about the Y-axis in the XOY plane; the zero-return angle error is equivalent to a small rotational deviation angle of the lidar scanning plane about the Z-axis. Range radar The wall at that location produce The trajectory center offset; S3.2, Define the direction angle of the laser radar scanning light beam projected onto the horizontal plane as . The unit direction vector of an outgoing beam in an elliptical scanning trajectory for: ； Let the zero-homing angle error of the lidar around the Z-axis be... Actual launch direction for: ； ； in, It introduces a rotation matrix around the Z-axis: The distance between the walls is The point where the light rays intersect the wall surface. satisfy: ； ； In the formula, These are the parameters of the point where the light intersects the wall. The direction angle is And exist At that time, the outgoing direction vector Projection component in the Y-axis direction; This allows us to obtain the X coordinate projected onto the XOY plane. : ； in, The direction angle is And exist At that time, the outgoing direction vector Projection component in the X-axis direction.

5. The method for correcting the zero-point angle error of airborne lidar based on a quasi-elliptical trajectory according to claim 1, characterized in that, Obtaining a correlation model between error and centroid deviation includes establishing Quantitative relationship with the cloud deviation at the point of origin; S3.3, Clearly none At that time, the parameters of the scanning light corresponding to the vertical bar being scanned. and Satisfy the following equation: ； in, For none At that time, the corresponding scanning angle The projection value of the scanning ray along the X-axis; For none At that time, the corresponding scanning angle The projection value of the scanning ray along the X-axis; In existence At that time, the x-coordinate of the cloud centroid of the two clusters of protrusions is: ； ； in, For existence At that time, the corresponding scanning angle The projection value of the scanning ray along the X-axis; For existence At that time, the corresponding scanning angle The projection value of the scanning ray along the X-axis; The centroids of the two clusters of bulging cloud satisfy the following equation: ； in, and For existence At that time, the projection values ​​of the X-axis corresponding to the two clusters of protruding cloud points respectively; The solution is transformed into finding the solution that makes The rotation angle approaching zero is obtained through first-order linearization: ； ； in, and They do not exist. At that time, the X-axis coordinates of the cloud centroids of the two clusters of protrusions are... and These are constant coefficients determined by the scanning geometry.

6. The method for correcting the zero-point angle error of airborne lidar based on a quasi-elliptical trajectory according to claim 1, characterized in that, S5 includes, S5.1, obtaining from S4 Write a point cloud fusion program to perform coordinate correction on the point cloud fusion results from the lidar, using the following correction formula: ； in, For the k-th breakthrough point cloud Corrected protrusion point cloud; The projection of the corrected elliptical scanning trajectory onto the wall is symmetrical about the Y-axis, and the two clusters of protruding point clouds are located on the same straight line perpendicular to the YOZ plane in the X direction.

7. The method for correcting the zero-point angle error of airborne lidar based on a quasi-elliptical trajectory according to claim 1, characterized in that, S5 includes, S5.2, after the compensation and correction are completed, scanning the same wall and rod-shaped object again, and recalculating the horizontal difference between the centroids of the two clusters of protruding cloud points according to S2 to S4: ； Set calibration termination threshold ,like If the zero-angle calibration is successful, the horizontal difference will be adjusted accordingly. The corresponding residual is treated as a new initial error, and steps S4 to S5.1 are repeated until the condition is met. The required precision.

Images