CN109377521B - Point cloud registration method for data acquisition midpoint of ground laser scanner to optimal plane - Google Patents

Point cloud registration method for data acquisition midpoint of ground laser scanner to optimal plane Download PDF

Info

Publication number
CN109377521B
CN109377521B CN201811056661.0A CN201811056661A CN109377521B CN 109377521 B CN109377521 B CN 109377521B CN 201811056661 A CN201811056661 A CN 201811056661A CN 109377521 B CN109377521 B CN 109377521B
Authority
CN
China
Prior art keywords
plane
point
point cloud
points
optimal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
CN201811056661.0A
Other languages
Chinese (zh)
Other versions
CN109377521A (en
Inventor
陶武勇
花向红
续东
吴伟
冯绍权
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Wuhan University WHU
Original Assignee
Wuhan University WHU
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Wuhan University WHU filed Critical Wuhan University WHU
Priority to CN201811056661.0A priority Critical patent/CN109377521B/en
Publication of CN109377521A publication Critical patent/CN109377521A/en
Application granted granted Critical
Publication of CN109377521B publication Critical patent/CN109377521B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
    • G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
    • G06T7/35—Determination of transform parameters for the alignment of images, i.e. image registration using statistical methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G06T2207/10008—Still image; Photographic image from scanner, fax or copier
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10028—Range image; Depth image; 3D point clouds

Abstract

The invention discloses a point cloud registration method from a data acquisition midpoint of a ground laser scanner to an optimal plane, which comprises searching the optimal plane corresponding to each point, and establishing pairing from the point to the optimal plane; calculating plane parameters by adopting constraint weighted total least square; and calculating conversion parameters by adopting constraint weighted total least squares. The invention has better registration precision.

Description

Point cloud registration method for data acquisition midpoint of ground laser scanner to optimal plane
Technical Field
The invention belongs to the technical field of point cloud data processing of a ground three-dimensional laser scanner, relates to a point cloud registration method, and particularly relates to a point cloud registration method for acquiring a point to an optimal plane by using a ground laser scanner.
Background
In the data acquisition process of the ground three-dimensional laser scanner, the situation that the foreground blocks the background may exist. Therefore, in order to obtain a complete three-dimensional model of a scanned target, the target needs to be scanned from different viewing angles to obtain point cloud data of the target from different viewing angles. However, the point cloud data directly output by the ground three-dimensional laser scanner is based on the local coordinates of the coordinate system of the station, and therefore, the point cloud data in different coordinate systems need to be converted into a common coordinate system, that is, the point cloud data in different stations are unified into a common reference system, and this process is called point cloud registration.
Point cloud registration mainly comprises two steps:
(1) establishing pairing;
(2) and calculating conversion parameters.
Currently, registration methods are mainly classified into three categories, namely a point-to-point registration method, a point-to-surface registration method and a surface-to-surface registration method. Establishing point-point pairing by a point-to-point registration method, and calculating a conversion parameter by minimizing the distance between a point and a point; establishing point-surface pairing by a point-surface registration method, and calculating conversion parameters by minimizing the distance between points and surfaces; the face-to-face registration method then creates face-to-face pairings, calculating the transformation parameters by minimizing the face-to-face distance.
The most popular point-to-point registration method is the Iterative Closest Point (ICP) algorithm, which assumes that there are exact points in the two point clouds that correspond one to one, however, for the actually measured point clouds, this assumption does not hold, i.e., there is no possibility that there is a point in one-to-one correspondence between the two actually measured point clouds. This results in a slow convergence rate of the point-to-point registration method. Therefore, a point-to-surface registration method has been proposed, and many studies have shown that the point-to-surface registration method is superior to the point-to-point registration method in terms of registration accuracy and convergence rate. However, there are few studies on how to construct a plane, and in the icpatch (iterative closed patch) method, a tin (triangular angular network) fragment is used as a corresponding plane, and in the icpp (iterative closed point) method, a fragment composed of the nearest three points is used as a corresponding plane. However, if the local surface of the object is a plane, more points should be used to fit the plane, so as to minimize the effect of noise. If the local surface is a curved surface, fewer points should be used to fit the plane, which may reduce the effect of the surface on the plane to which it is fitted. Therefore, a method is provided for searching an optimal plane, wherein the optimal plane can best represent a local surface of a target, and a point-to-optimal plane registration method is obtained and named as an iterative closest optimal plane (ICOPlane) algorithm.
Disclosure of Invention
The invention discloses a point cloud registration method from a data acquisition midpoint of a ground laser scanner to an optimal plane, which is suitable for point cloud registration of a free form target and can realize accurate registration of two point clouds.
The technical scheme adopted by the invention is as follows: a point cloud registration method for data acquisition of ground laser scanner from midpoint to optimal plane assumes two point clouds P1And P2With P1As a reference point cloud, P2As point cloud to be registered, i.e. P2Registration to P1(ii) a And assuming that a coarse registration has been performed, a point cloud is obtained
Figure BDA0001795957360000021
Namely, it is
Figure BDA0001795957360000022
Wherein R isCIs the rotation matrix obtained by coarse registration, tCIs the translation vector obtained by coarse registration;
characterized in that the method comprises the following steps:
step 1: obtaining a point cloud P by a ground laser scanner1And P2Calculating the point cloud P1，P2Covariance matrix of each point in the
Figure BDA0001795957360000023
And
Figure BDA0001795957360000024
where i represents the ith point in the point cloud, so
Figure BDA0001795957360000025
Representing a point cloud P1The covariance matrix of the ith point in the time domain,
Figure BDA0001795957360000026
representing a point cloud P2A covariance matrix of the ith point;
step 2: as a point cloud
Figure BDA0001795957360000027
Searching each point in the image to obtain a corresponding optimal plane, and establishing a pairing between the point and the optimal plane;
and step 3: according to the pairing of the established points to the optimal plane, the points are subjected to point cloud
Figure BDA0001795957360000028
Registration to the point cloud P1To obtain a point cloud
Figure BDA0001795957360000029
Namely, accurately calculating a rotation matrix R and a translational vector t;
and 4, step 4: computing transformed point clouds
Figure BDA00017959573600000210
And updating the point cloud
Figure BDA00017959573600000211
And 5: and (4) repeating the steps 2 to 4 until the variation of the root mean square error RMS is less than a given threshold value.
Compared with the prior art, the invention has the characteristics that:
(1) in point cloud registration, it is usually assumed that the target local region is a plane, and in the past point-to-plane registration method, how to fit the plane is rarely discussed. If the local area is a plane, more points should be used to fit the plane, so that the effect of noise is reduced as much as possible. If the local region is a curved surface, fewer points should be used to fit the plane, since the assumption that the local region is a plane does not hold. Therefore, the invention provides a method for searching the optimal plane, so that the corresponding plane (namely the optimal plane) searched by each point can best approximate the local shape of the target;
(2) and the invention adopts constrained weighted total least square calculation when calculating the plane parameters and the conversion parameters. In previous researches, least square calculation is adopted, which means that errors of some variables are ignored. The constraint weighted total least square considers the errors of all variables, the theory is more rigorous, and more accurate parameter estimation can be obtained;
(3) the experimental analysis shows that: compared with the ICPP algorithm, the method can obtain more accurate conversion parameters and requires fewer iterations. The experiment also designs that plane parameters and conversion parameters are calculated by using constraint weighted least squares, and the calculation results are compared with the calculation results of the method provided by the invention, so that the registration precision can be improved by using constraint weighted total least squares.
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention;
fig. 2 and 3 are experimental data used in a simulation experiment in the embodiment of the present invention, which are two point clouds of a mechanical part model and two point clouds of a human head model, respectively.
FIGS. 4 and 5 are diagrams illustrating the end of the calculation of the mechanical part model and the human head model, respectively, according to an embodiment of the present invention; norm in the figure represents the Norm of the difference between the conversion parameter estimate and the true value of the conversion parameter; ICPP is the existing registration method, and ICOPlane represents the registration method provided by the invention; the ICOPlane-WLS is a method obtained by modifying ICOPlane, namely plane parameters and conversion parameters are calculated by adopting constraint weighted least square, and other parts are kept the same as the ICOPlane;
fig. 6 and 7 are experimental data used in a real experiment in the embodiment of the present invention, which are point clouds obtained by scanning a statue and a stake, respectively.
Detailed Description
The following detailed description of the various steps of the proposed method and the experimental procedures are provided to facilitate the understanding and practice of the invention by those skilled in the art, it being understood that the implementation examples described herein are only for the purpose of illustration and explanation and are not intended to limit the invention.
Referring to fig. 1, the point cloud registration method for acquiring midpoint to optimal plane of ground laser scanner provided by the invention assumes that there are two point clouds P1And P2With P1As a reference point cloud, P2As a cloud of points to be registered, i.e. P2Registration to P1(ii) a And assuming that a coarse registration has been performed, a point cloud is obtained
Figure BDA0001795957360000041
Namely, it is
Figure BDA0001795957360000042
Wherein R isCIs the rotation matrix obtained by coarse registration, tCIs the translation vector obtained by coarse registration;
characterized in that the method comprises the following steps:
step 1: by ground laser scanningThe instrument obtains a point cloud P1And P2Calculating the point cloud P1，P2Covariance matrix of each point in the
Figure BDA0001795957360000043
And
Figure BDA0001795957360000044
where i represents the ith point in the point cloud, so
Figure BDA0001795957360000045
Representing a point cloud P1The covariance matrix of the ith point in the time domain,
Figure BDA0001795957360000046
representing a point cloud P2A covariance matrix of the ith point;
according to the measuring principle of ground laser scanner, one point p in point cloudiThe coordinates of (a) are:
Figure BDA0001795957360000047
where ρ isiDenotes the distance measurement, θiThe vertical angle is indicated as the angle of the vertical,
Figure BDA0001795957360000048
represents a horizontal angle;
linearizing the above equation:
Figure BDA0001795957360000049
the covariance matrix of the point cloud is obtained as:
Figure BDA00017959573600000410
wherein the content of the first and second substances,
Figure BDA00017959573600000411
σρrepresenting the range error, σθThe error in the vertical angle is expressed as,
Figure BDA00017959573600000412
indicating the horizontal angle error, αiIndicating the angle of incidence and delta the angle of divergence of the light pillar.
Step 2: as a point cloud
Figure BDA00017959573600000413
Searching each point in the image to obtain a corresponding optimal plane, and establishing a pairing between the point and the optimal plane;
this embodiment uses the Nearest Neighbor (KNN) algorithm at P1Searching in a point cloud
Figure BDA00017959573600000414
Middle point
Figure BDA00017959573600000415
Nearest k points, then take the nearest v points from the k points to fit a plane, the value of v being from v0Increasing to k with an interval of 1, fitting a plane to the different values of v (in the present embodiment, k is 20, v is05). Thus, different v values correspond to one plane, and a plurality of planes can be obtained. The fitting accuracy σ of each plane is then calculatedPAnd taking the plane corresponding to the minimum fitting precision as an optimal plane, and establishing the pairing from the point to the optimal plane. The same method is used for point clouds
Figure BDA0001795957360000051
The other points in the list search for the corresponding best plane.
In this process, this embodiment uses a constrained weighted total least squares fit plane, assuming the equation for one plane is:
Figure BDA0001795957360000052
wherein a, b, c, d represent plane parameters, and x, y, z represent point coordinates for fitting a plane;
considering the error in the three directions of x, y and z:
Figure BDA0001795957360000053
wherein e isx,ey,ezRepresenting errors in three directions of x, y and z;
according to the first equation of the above formula, at an initial value (e)0,ξ0) Linearization is carried out to obtain:
ψ(e,ξ)＝ψ(e0,ξ0)+A0(ξ-ξ0)+B0(e-e0)＝0
wherein
Figure BDA0001795957360000054
v represents the number of points used to fit the plane;
ξ＝[a b c d]T，ξ0is an initial value of xi, e0Is the initial value of e, coefficient matrix A0And B0Comprises the following steps:
Figure BDA0001795957360000055
Figure BDA0001795957360000056
wherein, a0,b0,c0Is xi0The elements (A) and (B) in (B),
Figure BDA0001795957360000057
is e0The elements of (1);
by also linearizing the second equation, one can obtain:
Figure BDA0001795957360000061
wherein delta xi is xi-xi0，L＝B0e0-ψ(ξ0,e0)，C＝[2a0 2b0 2c0 0]，
Figure BDA0001795957360000062
The solution of the above formula is then obtained:
Figure BDA0001795957360000063
wherein
Figure BDA0001795957360000064
D is the covariance matrix of the v point coordinates.
The estimate of the residual vector is:
Figure BDA0001795957360000065
wherein the initial value xi0Can be obtained by Principal Component Analysis (PCA) method, and has initial value e0Set to a zero vector. And (4) until a convergence condition is reached through iterative calculation.
The covariance matrix of the unit weight standard deviation and the plane parameter is:
Figure BDA0001795957360000066
Figure BDA0001795957360000067
the fitting accuracy of the final plane is calculated as:
Figure BDA0001795957360000068
wherein d isiThe distance of the points to the fitted plane is indicated.
And step 3: according to the pairing of the established points to the optimal plane, the points are subjected to point cloud
Figure BDA0001795957360000069
Registration to the point cloud P1To obtain a point cloud
Figure BDA00017959573600000610
Namely, accurately calculating a rotation matrix R and a translational vector t;
assuming that a point has been registered to the corresponding optimal plane, the distance of this point to the corresponding optimal plane is theoretically zero:
Figure BDA00017959573600000611
wherein the content of the first and second substances,
Figure BDA00017959573600000612
points representing transitions
Figure BDA00017959573600000613
The coordinates of (a);
point of conversion
Figure BDA00017959573600000614
The coordinates of (a) are:
Figure BDA0001795957360000071
wherein
Figure BDA0001795957360000072
Indicating points
Figure BDA0001795957360000073
The coordinates of (a). The rotation matrix and translation vector are represented as:
Figure BDA0001795957360000074
wherein r is1～r99 elements, t, representing a rotation matrix1～t 33 elements representing translation vectors;
then there are:
Figure BDA0001795957360000075
considering the errors of the plane parameters and point coordinates, the above equation becomes:
Figure BDA0001795957360000076
wherein the content of the first and second substances,
Figure BDA0001795957360000077
an error of a parameter of the plane is represented,
Figure BDA0001795957360000078
error representing point coordinates;
at an initial value (ε)0,β0) And (3) carrying out linearization to obtain:
ψ(ε,β)＝ψ(ε0,β0)+X0(β-β0)+G0(ε-ε0)＝0
β＝[r1 r2 r3 r4 r5 r6 r7 r8 r9 t1 t2 t3]T，
Figure BDA0001795957360000079
u represents the number of pairings;
coefficient matrix X0Comprises the following steps:
Figure BDA00017959573600000710
coefficient matrix G0Comprises the following steps:
Figure BDA0001795957360000081
since the rotation matrix is an orthogonal matrix, the elements in the rotation matrix obey the following constraints:
Figure BDA0001795957360000082
then the following constraint equation linearizes the above equation:
Kδβ＝k0
wherein
Figure BDA0001795957360000083
Figure BDA0001795957360000084
The following equation can be obtained:
Figure BDA0001795957360000085
wherein δ β ═ β - β0，y＝G0ε0-ψ(β0,ε0)。
The resulting equation is solved as:
Figure BDA0001795957360000086
wherein
Figure BDA0001795957360000091
Figure BDA0001795957360000092
The estimate of the residual vector is:
Figure BDA0001795957360000093
initial value epsilon0Set as zero vector, initial value beta0＝[1 0 0 0 1 0 0 0 1 0 0 0]T(ii) a And (4) reaching a convergence condition through iterative calculation.
And 4, step 4: computing transformed point clouds
Figure BDA0001795957360000094
And updating the point cloud
Figure BDA0001795957360000095
And 5: and (4) repeating the steps 2 to 4 until the variation of the root mean square error RMS is less than a given threshold value.
In the experimental process of this embodiment, a mechanical component model and a point cloud of a human head model are used in a simulation experiment, and the point clouds of the two models are converted by setting a true value of a conversion parameter, so that each model has two point clouds with different viewing angles, as shown in fig. 2 and 3. Then, in the embodiment, Gaussian noise with standard deviation of 1,2,3,4,5cm is added to the coordinates of each point in the four point clouds, and conversion parameters are calculated by adopting three methods, namely ICPP, ICOPlane-WLS and ICOPlane. The difference norm of the calculated conversion parameter estimate and the conversion parameter true value is shown in fig. 4 and 5. And the number of iterations required for each method are shown in tables 1 and 2.
TABLE 1 mechanical part model number of iterations required for various methods
Figure BDA0001795957360000096
TABLE 2 number of iterations required for various methods of the human head model
Figure BDA0001795957360000097
As can be seen from fig. 4 and 5, the difference norm calculated by ICOPlane is smaller than the difference norm calculated by ICPP, regardless of the calculation result of the mechanical part model or the calculation result of the human head model. This shows that the conversion parameters of the ICOPlane calculation are closer to the true values, and the calculation result is more accurate. Comparing the results of the calculations for ICOPlane and ICOPlane-WLS, it can be seen that the results for the two methods are substantially the same for the machine component model, with the result for ICOPlane being slightly closer to the true value. For the human head model, the calculation results for ICOPlane are significantly closer to the true value than those for ICOPlane-WLS. Therefore, computing the plane parameters and the transformation parameters by using constrained global minimum two-times does benefit to improve the registration accuracy, and may sometimes not improve significantly.
As can be seen from tables 1 and 2, the number of iterations required for the calculations ICOPlane and ICOPlane-WLS is always less than the number of iterations required for the calculations ICPP at different noise levels. This indicates that point-to-best plane pairing is preferred over point-to-segment pairing (segment formed by the nearest three points) because the number of iterations is mainly determined by the capabilities of the pairing method, e.g., point-to-plane pairing is preferred over point-to-point pairing.
In a real experiment, a Riegl VZ-400 ground laser scanner was used to collect data, scans the statue and stump from two perspectives, respectively, and laid 6 spherical targets around the target, a RiSCAN PRO was used to obtain the center coordinates of each target, and then conversion parameters were calculated using the target center coordinates, which served as reference or "true values". Thus, each target obtains two point clouds with different view angles, a part of each point cloud is extracted for registration, the extracted point clouds are subjected to coarse registration, and the point clouds subjected to the coarse registration are shown in fig. 6 and 7. The results of the calculations for the three methods are shown in tables 3 and 4.
TABLE 3 calculation of three methods for figurine data
Figure BDA0001795957360000101
Table 4 calculation results of three methods of stake data
Figure BDA0001795957360000102
As can be seen from tables 3 and 4, ICOPlane calculates the smallest difference norm, calculates the closest transition parameter to the "true" value, and requires the least number of iterations. The difference norm calculated by ICPP is the largest, and the calculated conversion parameter is far away from the true value.
It should be understood that parts of the specification not set forth in detail are well within the prior art.
It should be understood that the above description of the preferred embodiments is given for clearness of understanding and no unnecessary limitations are to be understood therefrom, for those skilled in the art may make modifications and alterations without departing from the scope of the invention as defined by the appended claims.

Claims (4)

1. A point cloud registration method for data acquisition of ground laser scanner from midpoint to optimal plane assumes two point clouds P1And P2With P1As a reference point cloud, P2As point clouds to be registered, i.e. P2Registration to P1(ii) a And assuming that a coarse registration has been performed, a point cloud is obtained
Figure FDA0002986516420000011
Namely, it is
Figure FDA0002986516420000012
Wherein R isCIs the rotation matrix obtained by coarse registration, tCIs the translation vector obtained by coarse registration;
characterized in that the method comprises the following steps:
step 1: obtaining a point cloud P by a ground laser scanner1And P2Calculating the point cloud P1，P2Covariance matrix of each point in the
Figure FDA0002986516420000013
And
Figure FDA0002986516420000014
where i represents the ith point in the point cloud, so
Figure FDA0002986516420000015
Representing a point cloud P1The covariance matrix of the ith point in the time domain,
Figure FDA0002986516420000016
representing a point cloud P2A covariance matrix of the ith point;
step 2: as a point cloud
Figure FDA0002986516420000017
Searching each point in the image to obtain a corresponding optimal plane, and establishing pairing between the point and the optimal plane;
wherein, a nearest neighbor KNN algorithm is adopted in P1Searching in a point cloud
Figure FDA0002986516420000018
Middle point
Figure FDA0002986516420000019
Nearest k points, then take the nearest v points from the k points to fit a plane, the value of v being from v0Increasing to k, fitting a plane to different v values with an interval of 1, so that different v values correspond to one plane, and obtaining a plurality of planes; the fitting accuracy σ of each plane is then calculatedPTaking a plane corresponding to the minimum fitting precision as an optimal plane, and establishing pairing from points to the optimal plane; in the same way, as a point cloud
Figure FDA00029865164200000110
Searching the corresponding optimal plane at other points;
and step 3: according to the pairing of the established points to the optimal plane, the points are subjected to point cloud
Figure FDA00029865164200000111
Registration to the point cloud P1To obtain a point cloud
Figure FDA00029865164200000112
Namely, accurately calculating a rotation matrix R and a translational vector t;
and 4, step 4: computing transformed point clouds
Figure FDA00029865164200000113
And updating the point cloud
Figure FDA00029865164200000114
And 5: and (4) repeating the steps 2 to 4 until the variation of the root mean square error RMS is less than a given threshold value.
2. The ground laser scanner data acquisition midpoint-to-optimal plane point cloud registration method of claim 1, characterized in that: in step 1, according to the measuring principle of the ground laser scanner, a point p in the point cloudiThe coordinates of (a) are:
Figure FDA0002986516420000021
where ρ isiDenotes the distance measurement, θiThe vertical angle is indicated as the angle of the vertical,
Figure FDA0002986516420000022
represents a horizontal angle;
linearizing the above equation:
Figure FDA0002986516420000023
the covariance matrix of the point cloud is obtained as:
Figure FDA0002986516420000024
wherein the content of the first and second substances,
Figure FDA0002986516420000025
σρrepresenting the range error, σθThe error in the vertical angle is expressed as,
Figure FDA0002986516420000026
indicating the horizontal angle error, αiIndicating the angle of incidence and delta the angle of divergence of the light pillar.
3. The ground laser scanner data acquisition midpoint-to-optimal plane point cloud registration method of claim 1, characterized in that: in step 2, a plane is fitted by using constraint weighted total least squares, and the equation of one plane is assumed as follows:
Figure FDA0002986516420000027
wherein a, b, c and d represent plane parameters; x, y, z represent point coordinates for fitting a plane;
considering the error in the three directions of x, y and z:
Figure FDA0002986516420000028
wherein e isx,ey,ezRepresenting errors in three directions of x, y and z;
according to the first equation of the above formula, at an initial value (e)0,ξ0) Linearization is carried out to obtain:
ψ(e,ξ)＝ψ(e0,ξ0)+A0(ξ-ξ0)+B0(e-e0)＝0；
wherein
Figure FDA0002986516420000029
v represents the number of points used to fit the plane; xi ═ a b c d]T，ξ0Is an initial value of xi, e0Is the initial value of e, coefficient matrix A0And B0Comprises the following steps:
Figure FDA0002986516420000031
Figure FDA0002986516420000032
wherein, a0,b0,c0Is xi0The elements (A) and (B) in (B),
Figure FDA0002986516420000033
is e0The elements of (1);
by the same token, linearizing the second equation yields:
Figure FDA0002986516420000034
wherein delta xi is xi-xi0，L＝B0e0-ψ(ξ0,e0)，
Figure FDA00029865164200000311
Figure FDA0002986516420000035
The solution of the above formula is then obtained:
Figure FDA0002986516420000036
wherein
Figure FDA0002986516420000037
D is a covariance matrix of the v point coordinates;
the estimate of the residual vector is:
Figure FDA0002986516420000038
wherein the initial value xi0Can be obtained by principal component analysis, initial value e0Set to a zero vector; until reaching the convergence condition through iterative computation;
the covariance matrix of the unit weight standard deviation and the plane parameter is:
Figure FDA0002986516420000039
Figure FDA00029865164200000310
the fitting accuracy of the final plane is calculated as:
Figure FDA0002986516420000041
wherein d isiThe distance of the points to the fitted plane is indicated.
4. The point cloud registration method from the data acquisition midpoint of the ground laser scanner to the optimal plane according to claim 1, wherein the step 3 is realized by the following steps:
assume a point
Figure FDA0002986516420000042
Having been registered to the corresponding best plane, this point has a distance of zero to the corresponding best plane:
Figure FDA0002986516420000043
wherein the content of the first and second substances,
Figure FDA0002986516420000044
points representing transitions
Figure FDA0002986516420000045
The coordinates of (a);
point of conversion
Figure FDA0002986516420000046
The coordinates of (a) are:
Figure FDA0002986516420000047
wherein
Figure FDA0002986516420000048
Indicating points
Figure FDA0002986516420000049
The coordinates of (a);
the rotation matrix R and the translation vector t are represented as:
Figure FDA00029865164200000410
t＝[t1 t2 t3]T
wherein r is1～r99 elements, t, representing a rotation matrix1～t33 elements representing translation vectors;
then there are:
Figure FDA00029865164200000411
considering the errors of the plane parameters and point coordinates, the above equation becomes:
Figure FDA00029865164200000412
wherein the content of the first and second substances,
Figure FDA0002986516420000051
an error of a parameter of the plane is represented,
Figure FDA0002986516420000052
error representing point coordinates;
at an initial value (ε)0,β0) And (3) carrying out linearization to obtain:
ψ(ε,β)＝ψ(ε0,β0)+X0(β-β0)+G0(ε-ε0)＝0
wherein β ═ r1 r2 r3 r4 r5 r6 r7 r8 r9 t1 t2 t3]T，
Figure FDA0002986516420000053
，
u represents the number of pairings;
coefficient matrix X0Comprises the following steps:
Figure FDA0002986516420000054
coefficient matrix G0Comprises the following steps:
Figure FDA0002986516420000055
since the rotation matrix is an orthogonal matrix, the elements in the rotation matrix obey the following constraints:
Figure FDA0002986516420000056
then the following constraint equation linearizes the above equation:
Kδβ＝k0；
wherein
Figure FDA0002986516420000057
Figure FDA0002986516420000061
The following equation is obtained:
Figure FDA0002986516420000062
wherein δ β ═ β - β0，y＝G0ε0-ψ(β0,ε0)；
The resulting equation is solved as:
Figure FDA0002986516420000063
wherein
Figure FDA0002986516420000064
Figure FDA0002986516420000065
The estimate of the residual vector is:
Figure FDA0002986516420000066
initial value epsilon0Set as zero vector, initial value beta0＝[1 0 0 0 1 0 0 0 1 0 0 0]T(ii) a And (4) until a convergence condition is reached through iterative calculation.
CN201811056661.0A 2018-09-11 2018-09-11 Point cloud registration method for data acquisition midpoint of ground laser scanner to optimal plane Expired - Fee Related CN109377521B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811056661.0A CN109377521B (en) 2018-09-11 2018-09-11 Point cloud registration method for data acquisition midpoint of ground laser scanner to optimal plane

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811056661.0A CN109377521B (en) 2018-09-11 2018-09-11 Point cloud registration method for data acquisition midpoint of ground laser scanner to optimal plane

Publications (2)

Publication Number Publication Date
CN109377521A CN109377521A (en) 2019-02-22
CN109377521B true CN109377521B (en) 2021-05-04

Family

ID=65405530

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811056661.0A Expired - Fee Related CN109377521B (en) 2018-09-11 2018-09-11 Point cloud registration method for data acquisition midpoint of ground laser scanner to optimal plane

Country Status (1)

Country Link
CN (1) CN109377521B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109946701B (en) * 2019-03-26 2021-02-02 新石器慧通（北京）科技有限公司 Point cloud coordinate conversion method and device
CN110443836B (en) * 2019-06-24 2022-03-18 中国人民解放军战略支援部队信息工程大学 Point cloud data automatic registration method and device based on plane features
CN110716194B (en) * 2019-09-10 2022-09-16 中国科学院上海技术物理研究所 Calibration method of multi-beam three-dimensional laser imaging system simultaneously referencing sphere and surface
CN111553938A (en) 2020-04-29 2020-08-18 南京航空航天大学 Multi-station scanning point cloud global registration method based on graph optimization
CN112132875B (en) * 2020-08-31 2023-07-28 青岛秀山移动测量有限公司 Multi-platform point cloud matching method based on surface features
CN113587816A (en) * 2021-08-04 2021-11-02 天津微深联创科技有限公司 Array type large-scene structured light three-dimensional scanning measurement method and device

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101839701A (en) * 2010-05-27 2010-09-22 中国科学院自动化研究所 Method for automatically partitioning tree point cloud data
CN101937564A (en) * 2010-08-17 2011-01-05 北京航空航天大学 Method for correcting registration accumulative errors of multi-view point cloud sequence
CN104392488A (en) * 2014-12-11 2015-03-04 福州大学 Automatic point cloud data rectification method aiming at laser scanner and three-coordinate measuring arm

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9760996B2 (en) * 2015-08-11 2017-09-12 Nokia Technologies Oy Non-rigid registration for large-scale space-time 3D point cloud alignment

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101839701A (en) * 2010-05-27 2010-09-22 中国科学院自动化研究所 Method for automatically partitioning tree point cloud data
CN101937564A (en) * 2010-08-17 2011-01-05 北京航空航天大学 Method for correcting registration accumulative errors of multi-view point cloud sequence
CN104392488A (en) * 2014-12-11 2015-03-04 福州大学 Automatic point cloud data rectification method aiming at laser scanner and three-coordinate measuring arm

Also Published As

Publication number Publication date
CN109377521A (en) 2019-02-22

Similar Documents

Publication Publication Date Title
CN109377521B (en) Point cloud registration method for data acquisition midpoint of ground laser scanner to optimal plane
CN110116407B (en) Flexible robot position and posture measuring method and device
KR102257610B1 (en) EXTRINSIC CALIBRATION METHOD OF PLURALITY OF 3D LiDAR SENSORS FOR AUTONOMOUS NAVIGATION SYSTEM
Lv et al. Targetless calibration of lidar-imu system based on continuous-time batch estimation
CN111414798A (en) Head posture detection method and system based on RGB-D image
JP6370038B2 (en) Position and orientation measurement apparatus and method
CN112017220A (en) Point cloud accurate registration method based on robust constraint least square algorithm
CN110335297A (en) A kind of point cloud registration method based on feature extraction
CN110533726B (en) Laser radar scene three-dimensional attitude point normal vector estimation correction method
CN113327275B (en) Point cloud double-view-angle fine registration method based on multi-constraint point to local curved surface projection
CN109523595A (en) A kind of architectural engineering straight line corner angle spacing vision measuring method
CN114862932B (en) BIM global positioning-based pose correction method and motion distortion correction method
JP2012026895A (en) Position attitude measurement device, position attitude measurement method, and program
CN110322492B (en) Space three-dimensional point cloud registration method based on global optimization
CN106570905A (en) Initial attitude verification method of non-cooperative target point cloud
CN112381862B (en) Full-automatic registration method and device for CAD (computer-aided design) model and triangular mesh
Ge et al. Online 3-D modeling of complex workpieces for the robotic spray painting with low-cost RGB-D cameras
CN108876862A (en) A kind of noncooperative target point cloud position and attitude calculation method
Al-Nuaimi et al. Analyzing LiDAR scan skewing and its impact on scan matching
Wei et al. Optimization of 3-D pose measurement method based on binocular vision
CN113313200B (en) Point cloud precision matching method based on normal constraint
CN111768370B (en) Aeroengine blade detection method based on RGB-D camera
CN113406658A (en) Mobile robot positioning method based on point-line characteristic scanning matching
CN110554405B (en) Normal scanning registration method and system based on cluster combination
CN112525106A (en) Three-phase machine cooperative laser-based 3D detection method and device

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20210504

Termination date: 20210911