CN115272075A - Bridge infrastructure laser scanning and inclined live-action point cloud data splicing method - Google Patents

Bridge infrastructure laser scanning and inclined live-action point cloud data splicing method Download PDF

Info

Publication number
CN115272075A
CN115272075A CN202210846607.6A CN202210846607A CN115272075A CN 115272075 A CN115272075 A CN 115272075A CN 202210846607 A CN202210846607 A CN 202210846607A CN 115272075 A CN115272075 A CN 115272075A
Authority
CN
China
Prior art keywords
plane
point cloud
point
points
cloud data
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.)
Pending
Application number
CN202210846607.6A
Other languages
Chinese (zh)
Inventor
丛波日
辛公锋
李庆营
刘如飞
王飞
陈铮
刘宪明
龙关旭
李想
李子豪
齐麟
刘凯峰
张彤辉
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.)
Innovation Research Institute Of Shandong Expressway Group Co ltd
Shandong University of Science and Technology
Shandong Hi Speed Engineering Inspection and Testing Co Ltd
Original Assignee
Innovation Research Institute Of Shandong Expressway Group Co ltd
Shandong University of Science and Technology
Shandong Hi Speed Engineering Inspection and Testing Co Ltd
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 Innovation Research Institute Of Shandong Expressway Group Co ltd, Shandong University of Science and Technology, Shandong Hi Speed Engineering Inspection and Testing Co Ltd filed Critical Innovation Research Institute Of Shandong Expressway Group Co ltd
Priority to CN202210846607.6A priority Critical patent/CN115272075A/en
Publication of CN115272075A publication Critical patent/CN115272075A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformation in the plane of the image
    • G06T3/40—Scaling the whole image or part thereof
    • G06T3/4038—Scaling the whole image or part thereof for image mosaicing, i.e. plane images composed of plane sub-images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformation in the plane of the image
    • G06T3/60—Rotation of a whole image or part thereof
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/50—Image enhancement or restoration by the use of more than one image, e.g. averaging, subtraction
    • G06T5/70—
    • 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
    • G06T2200/00—Indexing scheme for image data processing or generation, in general
    • G06T2200/32—Indexing scheme for image data processing or generation, in general involving image mosaicing
    • 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
    • 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
    • G06T2207/20—Special algorithmic details
    • G06T2207/20212—Image combination
    • G06T2207/20221—Image fusion; Image merging

Abstract

The invention discloses a bridge infrastructure laser scanning and oblique live-action point cloud data splicing method, which belongs to the technical field of photogrammetry and comprises the following steps: carrying out multi-partition least square plane fitting denoising; manually and interactively extracting a common plane; fitting a multi-partition least square plane; performing a minimum median value method to fit a plane on all the plane point clouds after the outliers are deleted, and obtaining a plane model coefficient; calculating conversion parameters between the laser scanning point cloud and the oblique live-action point cloud data by using an indirect adjustment method based on the plane model coefficients of the homonymous plane point cloud, and realizing high-precision fusion between the laser scanning point cloud and the oblique live-action scanning point cloud according to the conversion parameters; and carrying out iterative operation by taking the root mean square error of the distance from the plane to the plane and the iteration times as constraint conditions to complete the point cloud data splicing. The method uses the minimum median square algorithm, can adaptively eliminate the influence of the abnormal points on the plane fitting process without setting excessive parameters, and has higher engineering application value.

Description

Bridge infrastructure laser scanning and inclined live-action point cloud data splicing method
Technical Field
The invention discloses a bridge infrastructure laser scanning and inclined live-action point cloud data splicing method, and belongs to the technical field of photogrammetry.
Background
With the rapid development of the social economy in China, the infrastructure construction is more and more extensive. Under the large background of live-action three-dimensional Chinese construction, the requirement for a three-dimensional live-action model reaches unprecedented height. The bridge is used as an important component of traffic infrastructure, and due to the complex structure, the scene full coverage is difficult to realize by a single data acquisition mode, so that the integrity of the subsequent three-dimensional live-action model construction is influenced. Therefore, the research of the multi-source point cloud data fusion technical method has important significance and application value.
The essence of the fusion of the bridge laser scanning and the inclined live-action point cloud is that the laser scanning point cloud and the inclined live-action point cloud which are positioned at different spatial positions are converted into the same position through rotation and translation operations. In the existing method, an ICP algorithm is mostly adopted for registration, and the basic idea for realizing the ICP algorithm is as follows: for each point in the laser scanning point cloud, finding a point closest to the oblique real scene point cloud in an Euclidean distance to form a matching point set; and solving a rotation matrix and a translation matrix by adopting an SVD (singular value decomposition) algorithm according to the matching point set, and performing iterative operation until the sum of squared Euclidean distances between all matching point pairs is minimum. In a road scene, the efficiency of a point-to-point-based registration method is not high due to the complex structure of bridge infrastructure and large data volume; and point cloud data acquired by different platforms have larger difference in scanning precision and density, so that homonymous point pairs in strict sense do not exist, and the registration error is larger.
Disclosure of Invention
The invention discloses a bridge infrastructure laser scanning and inclined live-action point cloud data splicing method, which aims to solve the problems of low efficiency and large registration error of a point-to-point based registration method in the prior art.
A bridge infrastructure laser scanning and oblique live-action point cloud data splicing method comprises the following steps:
s1, fitting and denoising a multi-partition least square plane;
s1.1, manually and interactively extracting a common plane;
s1.2, fitting a multi-partition least square plane;
s2, performing a minimum median value method to fit a plane on all the plane point clouds after the outliers are deleted, and obtaining a plane model coefficient;
s3, calculating a conversion parameter between the laser scanning point cloud and the inclined live-action point cloud data by using an indirect balance method based on a plane model coefficient of the homonymous plane point cloud, and realizing high-precision fusion between the laser scanning point cloud and the inclined live-action scanning point cloud according to the conversion parameter;
and S4, carrying out iterative operation by taking the root mean square error of the distance from the plane to the plane and the iteration times as constraint conditions to complete point cloud data splicing.
Preferably, the S1 includes: performing spatial multi-partition least square plane fitting in each plane point cloud data, accumulating and statistically analyzing the distance from the discrete points to the partition fitting plane, and deleting the points far away from the fitting plane as outliers;
and extracting the same plane characteristic target in the laser scanning point cloud and the inclined live-action point cloud data through the point cloud data displayed by software.
Preferably, said S1.2 comprises:
for any point P in the point cloud datai(xi,yi,zi) With PiTaking r as radius as center to search sphere neighborhood, if P is in the sphere neighborhoodiIf the number of the neighboring points is less than 3, the P is addediPoint marking is outlier deletion; if the number of the adjacent points is more than 3, P is addediThe points in its neighborhood construct the covariance matrix M3×3：
Figure BDA0003729612180000021
Wherein m is a point PiNumber of adjacent points of (1), m>3；
Figure BDA0003729612180000022
Is PiAnd the three-dimensional centroid of m adjacent points
Figure BDA0003729612180000023
When i =0, represents PiThe point itself;
solving the normalized eigenvector corresponding to the minimum eigenvalue of covariance
Figure BDA0003729612180000024
I.e. the unit normal vector of the local plane, and further obtain PiDistance d to local planeiComprises the following steps:
Figure BDA0003729612180000025
and traversing to calculate the mean value mu of the distance from each point to the local plane:
Figure BDA0003729612180000026
standard deviation σ:
Figure BDA0003729612180000027
when a certain point PiThe distance to the local plane is in the range of (μ - σ, μ + σ) to retain the point, and out of this range is defined as outlier deletion.
Preferably, the S2 includes:
s3.1, randomly selecting three points in a plane, calculating model parameters, randomly selecting 3 points from the point cloud of the plane, and solving a plane equation through the 3 points; assume that the 3 points selected are p1(x1,y1,z1)，P2(x2,y2,z2)，P3(x3,y3,z3) The following can be obtained:
Figure BDA0003729612180000028
the unit normal vector n of the plane is:
Figure BDA0003729612180000029
unit normal vector sum of known planesA point P on the plane1The coefficients of the plane equation can be obtained;
s3.2, calculating the distance dis from each point to the plane obtained in S2.1, counting the square median omega of the distance, and performing iterative computation k times;
s3.3, selecting a fitting plane corresponding to the minimum value of omega in the k iterations as a best fitting plane, calculating a weight corresponding to each datum according to the formulas (1) and (2), and adaptively eliminating abnormal points through the weights;
Figure BDA00037296121800000210
wherein σ is the robust standard deviation; n is the number of points; r isiRepresenting the distance between the ith point and the fitting plane; med represents the median of the distances;
Figure BDA00037296121800000211
in the formula, wiRepresenting the weight of the ith point, and considering the ith point as an abnormal point if the weight is 0;
s3.4, in the point set after the abnormal points are removed, calculating the distance between each point and the best fit plane, and taking the point with the distance smaller than a given threshold value as an inner point;
and S3.5, recalculating the parameters of the plane model by using a least square method for the interior points.
Preferably, the S3 includes:
acquiring a coordinate conversion model, constructing an indirect adjustment model based on a plane model coefficient of the point cloud of the same-name plane, and calculating a translation vector between the point cloud of the laser scanning and the point cloud of the inclined live view;
setting the same plane in a coordinate system O1-X1Y1Z1Is represented by P (a)1,b1,c1,d1) In the coordinate system O2-X2Y2Z2Is represented by Q (a)2,b2,c2,d2) Wherein a, b, c and d represent description parameters of the plane; according to affine transformation, the following relationship is satisfied between the two planes:
Figure BDA0003729612180000031
in the formula:
Figure BDA0003729612180000032
wherein α is the angle of rotation about the X axis; beta is the rotation angle around the Y axis; gamma is the rotation angle around the Z axis;
the rotation matrix is linearized, and nine parameters are separated and converted into three parameters:
Figure BDA0003729612180000033
for alpha, beta, gamma respectively with respect to a2,b2,c2The final linear model is obtained by calculating the partial derivatives as follows: v = Bx-l
Wherein:
Figure BDA0003729612180000034
wherein
Figure BDA0003729612180000035
Is a2Is determined by the estimated value of (c),
Figure BDA0003729612180000036
Figure BDA0003729612180000037
Figure BDA0003729612180000038
is b is2Is determined by the estimated value of (c),
Figure BDA0003729612180000039
Figure BDA00037296121800000310
Figure BDA00037296121800000311
is c2Is determined by the estimated value of (c),
Figure BDA00037296121800000312
Figure BDA00037296121800000313
Figure BDA00037296121800000314
is d2Is determined by the estimated value of (c),
Figure BDA00037296121800000315
according to the indirect adjustment principle, the correction value of each parameter in the transformation matrix can be calculated:
Figure BDA00037296121800000316
preferably, the S4 includes:
using the root mean square error of the distance from the plane to the plane and the iteration number as constraint conditions, and aiming at any homonymous plane (P) after registrationi,Qi) Calculating the Euclidean distance d between themPQThen the root mean square error rmse is:
Figure BDA0003729612180000041
when the rmse is smaller than a set threshold, the registration is successful, and meanwhile, the constraint of iteration times is given, so that infinite loop iteration is avoided under the condition that the root mean square error does not meet the set threshold;
and (3) constructing a rotation matrix and a translation vector by using coordinate conversion parameters:
Figure BDA0003729612180000042
wherein R is a rotation matrix, R (1, 1) is a first row and a first column element in the rotation matrix, R (1, 2) is a first row and a second column element in the rotation matrix, R (1, 3) is a first row and a third column element in the rotation matrix, R (2, 1) is a second row and a first column element in the rotation matrix, R (2, 2) is a second row and a second column element in the rotation matrix, R (2, 3) is a second row and a third column element in the rotation matrix, R (3, 1) is a third row and a first column element in the rotation matrix, R (3, 2) is a third row and a second column element in the rotation matrix, and R (3, 3) is a third row and a third column element in the rotation matrix; t is a translation vector, T (1, 1) is a first row and first column element in the translation vector, T (2, 1) is a second row and first column element in the translation vector, and T (3, 1) is a third row and first column element in the translation vector;
the coordinate conversion model is as follows:
Figure BDA0003729612180000043
wherein x istargetIs the x coordinate, y of any point in the target point cloudtargetIs the y coordinate, z, of any point in the target point cloudtargetIs the z coordinate, x, of any point in the target point cloudsourceIs the x coordinate, y of any point in the source point cloudsourceIs the y coordinate, z, of any point in the source point cloudsourceIs the z coordinate of any point in the target point cloud.
Compared with the prior art, the invention has the beneficial effects that: performing multi-partition least square fitting on the plane point cloud data, performing statistical analysis on the distance from the point to a partition fitting plane, and eliminating noise points in the point cloud data, so that the influence of flying points, noise and the like in the point cloud data on the subsequent plane fitting precision is overcome; by using a minimum median square algorithm, the influence of abnormal points on the plane fitting process can be removed in a self-adaptive manner without setting excessive parameters, and the method has high engineering application value;
the corresponding relation is established according to the parameters of the homonymy plane, and the conversion parameters are calculated by using the indirect adjustment model, so that the method is simple, the calculation speed is high, and the problem of large-angle rotation can be effectively solved; the registration method based on the plane features is used for realizing the fusion of laser scanning and oblique live-action point cloud data, has high efficiency and high precision, and overcomes the influence of mass point cloud data on the registration efficiency and the influence of different scanning instruments on the registration precision.
Drawings
FIG. 1 is a technical flow diagram of the present invention;
FIG. 2 is a schematic diagram of planar interactive extraction 1;
FIG. 3 is a schematic diagram of planar interactive extraction 2;
FIG. 4 is a schematic diagram of multi-partition least squares fitting denoising in FIG. 1;
FIG. 5 is a schematic diagram of multi-partition least squares fitting denoising in FIG. 2;
FIG. 6 is a diagram of a least median squares plane fitting process;
FIG. 7 is a diagram of the initial position of the point cloud 1;
FIG. 8 is a diagram of the initial position of the point cloud 2;
fig. 9 is a registration result fig. 1;
fig. 10 is a registration result fig. 2.
Detailed Description
The present invention will be described in further detail with reference to specific embodiments below:
a bridge infrastructure laser scanning and oblique live-action point cloud data splicing method is shown in figure 1 and comprises the following steps:
s1, carrying out multi-partition least square plane fitting denoising, as shown in a graph 4 and a graph 5;
s1.1, manually and interactively extracting a common plane, as shown in a figure 2 and a figure 3;
s1.2, fitting a multi-partition least square plane, as shown in a figure 6;
s2, performing plane fitting by a minimum median plane method on all the plane point clouds of which the outliers are deleted to obtain a plane model coefficient;
s3, calculating a conversion parameter between the laser scanning point cloud and the inclined live-action point cloud data by using an indirect balance method based on a plane model coefficient of the homonymous plane point cloud, and realizing high-precision fusion between the laser scanning point cloud and the inclined live-action scanning point cloud according to the conversion parameter;
and S4, carrying out iterative operation by taking the root mean square error of the distance from the plane to the plane and the iteration times as constraint conditions to complete point cloud data splicing.
Taking a high-speed simply supported beam bridge as an example, the initial positions of the laser scanning and inclined live-action point clouds are shown in fig. 7 and 8, wherein fig. 7 is a top view, and fig. 8 is a front view; the fusion result is shown in fig. 9 and 10, in which fig. 9 is a plan view and fig. 10 is a front view.
The S1 comprises: performing spatial multi-partition least square plane fitting in each plane point cloud data, accumulating and statistically analyzing the distance from the discrete points to the partition fitting plane, and deleting the points far away from the fitting plane as outliers;
and extracting the same plane characteristic target in the laser scanning point cloud and the inclined live-action point cloud data through the point cloud data displayed by software.
The S1.2 comprises:
for any point P in the point cloud datai(xi,yi,zi) With PiTaking r as radius to search the sphere neighborhood if P is in the sphere neighborhoodiIf the number of the neighboring points is less than 3, the P is addediPoint marking is outlier deletion; if the number of the adjacent points is more than 3, P is addediThe points in its neighborhood construct the covariance matrix M3×3：
Figure BDA0003729612180000061
Wherein m is a point PiNumber of adjacent points of (c), m>3；
Figure BDA0003729612180000062
Is PiAnd the three-dimensional centroid of m adjacent points
Figure BDA0003729612180000063
When i =0, represents PiThe point itself;
solving the normalized eigenvector corresponding to the minimum eigenvalue of the covariance
Figure BDA0003729612180000064
I.e. the unit normal vector of the local plane, and further obtain PiDistance d to local planeiComprises the following steps:
Figure BDA0003729612180000065
and traversing and calculating the mean value mu of the distance from each point to the local plane:
Figure BDA0003729612180000066
standard deviation σ:
Figure BDA0003729612180000067
when a certain point PiThe distance to the local plane is in the range (μ - σ, μ + σ) where the point is retained and not defined as outlier deletion.
The S2 comprises the following steps:
s3.1, randomly selecting three points in a plane, calculating model parameters, randomly selecting 3 points from the point cloud of the plane, and solving a plane equation through the 3 points; assume that the selected 3 points are P1(x1,y1,z1)，P2(x2,y2,z2)，P3(x3,y3,z3) The following can be obtained:
Figure BDA0003729612180000068
the unit normal vector n of the plane is:
Figure BDA0003729612180000069
knowing the unit normal vector of the plane and a point P on the plane1The coefficients of the plane equation can be obtained;
s3.2, calculating the distance dis from each point to the plane obtained in S2.1, counting the square median omega of the distance, and performing iterative computation k times;
s3.3, selecting a fitting plane corresponding to the minimum value of omega in the k iterations as a best fitting plane, calculating a weight corresponding to each datum according to the formulas (1) and (2), and adaptively eliminating abnormal points through the weights;
Figure BDA00037296121800000610
wherein σ is the robust standard deviation; n is the number of points; r is a radical of hydrogeniRepresenting the distance between the ith point and the fitting plane; med represents the median of the distances;
Figure BDA00037296121800000611
in the formula, wiRepresenting the weight of the ith point, and considering the ith point as an abnormal point if the weight is 0;
s3.4, in the point set after the abnormal points are removed, calculating the distance between each point and the best fit plane, and taking the point with the distance smaller than a given threshold value as an inner point;
and S3.5, recalculating the parameters of the plane model by using a least square method for the interior points.
The S3 comprises the following steps:
acquiring a coordinate conversion model, constructing an indirect adjustment model based on a plane model coefficient of homonymous plane point cloud, and calculating a translation vector between the laser scanning point cloud and the oblique real scene scanning point cloud;
setting the same plane in a coordinate system O1-X1Y1Z1Is represented by P (a)1,b1,c1,d1) In the coordinate system O2-X2Y2Z2In (b) is represented by Q (a)2,b2,c2,d2) Wherein a, b, c and d represent description parameters of the plane; according to affine transformation, the following relationship is satisfied between the two planes:
Figure BDA0003729612180000071
in the formula:
Figure BDA0003729612180000072
wherein α is the angle of rotation about the X axis; β is the angle of rotation about the Y axis; gamma is the rotation angle around the Z axis;
the rotation matrix is linearized, and nine parameters are separated and converted into three parameters:
Figure BDA0003729612180000073
for alpha, beta, gamma respectively with respect to a2,b2,c2The final linear model is obtained by calculating the partial derivatives as follows: v = Bx-l
Wherein:
Figure BDA0003729612180000074
wherein
Figure BDA0003729612180000075
Is a2Is determined by the estimated value of (c),
Figure BDA0003729612180000076
Figure BDA0003729612180000077
Figure BDA0003729612180000078
is b is2Is determined by the estimated value of (c),
Figure BDA0003729612180000079
Figure BDA00037296121800000710
Figure BDA00037296121800000711
is c2Is determined by the estimated value of (c),
Figure BDA00037296121800000712
Figure BDA00037296121800000713
Figure BDA00037296121800000714
is d2Is determined by the estimated value of (c),
Figure BDA00037296121800000715
according to the indirect adjustment principle, the correction value of each parameter in the transformation matrix can be calculated:
Figure BDA00037296121800000716
the S4 comprises the following steps:
using the root mean square error of the distance from the plane to the plane and the iteration number as constraint conditions, and aiming at any homonymous plane (P) after registrationi,Qi) Calculating the Euclidean distance d between themPQThen the root mean square error rmse is:
Figure BDA00037296121800000717
when the rmse is smaller than a set threshold, the registration is successful, and meanwhile, the constraint of iteration times is given, so that infinite loop iteration is avoided under the condition that the root mean square error does not meet the set threshold;
and (3) constructing a rotation matrix and a translation vector by using coordinate conversion parameters:
Figure BDA0003729612180000081
wherein R is a rotation matrix, R (1, 1) is a first row and a first column element in the rotation matrix, R (1, 2) is a first row and a second column element in the rotation matrix, R (1, 3) is a first row and a third column element in the rotation matrix, R (2, 1) is a second row and a first column element in the rotation matrix, R (2, 2) is a second row and a second column element in the rotation matrix, R (2, 3) is a second row and a third column element in the rotation matrix, R (3, 1) is a third row and a first column element in the rotation matrix, R (3, 2) is a third row and a second column element in the rotation matrix, and R (3, 3) is a third row and a third column element in the rotation matrix; t is a translation vector, T (1, 1) is a first row and first column element in the translation vector, T (2, 1) is a second row and first column element in the translation vector, and T (3, 1) is a third row and first column element in the translation vector;
the coordinate conversion model is as follows:
Figure BDA0003729612180000082
wherein xtargetIs the x coordinate, y, of any point in the target point cloudtargetIs the y coordinate, z, of any point in the target point cloudtargetIs the z coordinate, x, of any point in the target point cloudsourceIs the x coordinate, y, of any point in the source point cloudsourceIs the y coordinate, z, of any point in the source point cloudsourceIs the z coordinate of any point in the target point cloud.
It is to be understood that the above description is not intended to limit the present invention, and the present invention is not limited to the above examples, and those skilled in the art may make modifications, alterations, additions or substitutions within the spirit and scope of the present invention.

Claims (6)

1. A bridge infrastructure laser scanning and oblique live-action point cloud data splicing method is characterized by comprising the following steps:
s1, fitting and denoising a multi-partition least square plane;
s1.1, manually and interactively extracting a common plane;
s1.2, fitting a multi-partition least square plane;
s2, performing a minimum median value method to fit a plane on all the plane point clouds after the outliers are deleted, and obtaining a plane model coefficient;
s3, calculating a conversion parameter between the laser scanning point cloud and the inclined live-action point cloud data by using an indirect balance method based on a plane model coefficient of the homonymous plane point cloud, and realizing high-precision fusion between the laser scanning point cloud and the inclined live-action scanning point cloud according to the conversion parameter;
and S4, carrying out iterative operation by taking the root mean square error of the distance from the plane to the plane and the iteration times as constraint conditions to complete point cloud data splicing.
2. The bridge infrastructure laser scanning and oblique live-action point cloud data splicing method according to claim 1, wherein the S1 comprises: performing spatial multi-partition least square plane fitting in each plane point cloud data, accumulating and statistically analyzing the distance from the discrete points to the partition fitting plane, and deleting the points far away from the fitting plane as outliers;
and extracting the same plane characteristic target in the laser scanning point cloud and the inclined live-action point cloud data through the point cloud data displayed by software.
3. The bridge infrastructure laser scanning and oblique live-action point cloud data splicing method according to claim 2, wherein the S1.2 comprises:
for any point P in the point cloud datai(xi，yi，zi) With PiTaking r as radius to search the sphere neighborhood if P is in the sphere neighborhoodiIf the number of the neighboring points is less than 3, the P is addediPoint markers are outlier deletions; if the number of the adjacent points is more than 3, P is addediThe points in its neighborhood construct the covariance matrix M3×3：
Figure FDA0003729612170000011
Wherein m is a point PiNumber of adjacent points of (1), m>3；
Figure FDA0003729612170000012
Is PiAnd the three-dimensional centroids of m adjacent points
Figure FDA0003729612170000013
When i =0, represents PiThe point itself;
solving the normalized eigenvector corresponding to the minimum eigenvalue of the covariance
Figure FDA0003729612170000014
I.e. the unit normal vector of the local plane, and further obtain PiDistance d to local planeiComprises the following steps:
Figure FDA0003729612170000015
traversing and calculating the mean value of the distance from each point to the local plane
Figure FDA0003729612170000016
Standard deviation σ:
Figure FDA0003729612170000017
when a certain point PiThe distance to the local plane is in the range (μ - σ, μ + σ) where the point is retained and not defined as outlier deletion.
4. The bridge infrastructure laser scanning and oblique live-action point cloud data splicing method according to claim 3, wherein the S2 comprises:
s3.1, randomly selecting three points in a plane, calculating model parameters, randomly selecting 3 points from the point cloud of the plane, and solving a plane equation through the 3 points; assume that the selected 3 points are P1(x1，y1，z1)，P2(x2，y2，z2)，P3(x3，y3，z3) The following can be obtained:
Figure FDA0003729612170000021
the unit normal vector n of the plane is:
Figure FDA0003729612170000022
knowing the unit normal vector of the plane and a point P on the plane1The coefficients of the plane equation can be obtained;
s3.2, calculating the distance dis from each point to the plane obtained in the S2.1, counting the square median omega of the distance, and performing iterative computation for k times;
s3.3, selecting a fitting plane corresponding to the minimum value of omega in the k iterations as a best fitting plane, calculating a weight corresponding to each datum according to the formulas (1) and (2), and removing abnormal points in a weight self-adaptive manner;
Figure FDA0003729612170000023
in the formula, sigma is a steady standard deviation; n is the number of points; r isiRepresenting the distance between the ith point and the fitting plane; med represents the median of the distances;
Figure FDA0003729612170000024
in the formula, wiRepresenting the weight of the ith point, and considering the ith point as an abnormal point if the weight is 0;
s3.4, in the point set after the abnormal points are removed, calculating the distance between each point and the best fit plane, and taking the point with the distance smaller than a given threshold value as an inner point;
and S3.5, recalculating the parameters of the plane model by using a least square method for the interior points.
5. The bridge infrastructure laser scanning and inclined live-action point cloud data splicing method according to claim 1, wherein the S3 comprises the following steps:
acquiring a coordinate conversion model, constructing an indirect adjustment model based on a plane model coefficient of the point cloud of the same-name plane, and calculating a translation vector between the point cloud of the laser scanning and the point cloud of the inclined live view;
setting the same plane in a coordinate system O1-X1Y1Z1Is represented by P (a)1，b1，c1，d1) In the coordinate system O2-X2Y2Z2Is represented by Q (a)2，b2，c2，d2) Wherein a, b, c and d represent description parameters of the plane; as known from affine transformation, the following relationship is satisfied between the two planes:
Figure FDA0003729612170000025
in the formula:
Figure FDA0003729612170000026
wherein α is the angle of rotation about the X axis; beta is the rotation angle around the Y axis; gamma is the rotation angle around the Z axis;
the rotation matrix is linearized, and nine parameters are separated and converted into three parameters:
Figure FDA0003729612170000031
for alpha, beta, gamma respectively with respect to a2，b2，c2The final linear model is obtained by calculating the partial derivatives as follows: v = Bx-l
Wherein:
Figure FDA0003729612170000032
wherein
Figure FDA0003729612170000033
Is a2Is determined by the estimated value of (c),
Figure FDA0003729612170000034
Figure FDA0003729612170000035
Figure FDA0003729612170000036
is b is2Is determined by the estimated value of (c),
Figure FDA0003729612170000037
Figure FDA0003729612170000038
Figure FDA0003729612170000039
is c2Is determined by the estimated value of (c),
Figure FDA00037296121700000310
Figure FDA00037296121700000311
Figure FDA00037296121700000312
is d2Is determined by the estimated value of (c),
Figure FDA00037296121700000313
according to the indirect adjustment principle, the correction value of each parameter in the transformation matrix can be calculated:
Figure FDA00037296121700000314
6. the bridge infrastructure laser scanning and oblique live-action point cloud data splicing method according to claim 1, wherein the S4 comprises:
using the root mean square error of the distance from the plane to the plane and the iteration number as constraint conditions, and aiming at any homonymous plane (P) after registrationi，Qi) Calculating the Euclidean distance d between themPQThen the root mean square error rmse is:
Figure FDA00037296121700000315
when the rmse is smaller than a set threshold, the registration is successful, and meanwhile, iteration time constraint is given, so that infinite loop iteration is avoided under the condition that the root mean square error does not meet the set threshold;
and (3) constructing a rotation matrix and a translation vector by using coordinate conversion parameters:
Figure FDA00037296121700000316
wherein R is a rotation matrix, R (1, 1) is a first row and a first column element in the rotation matrix, R (1, 2) is a first row and a second column element in the rotation matrix, R (1, 3) is a first row and a third column element in the rotation matrix, R (2, 1) is a second row and a first column element in the rotation matrix, R (2, 2) is a second row and a second column element in the rotation matrix, R (2, 3) is a second row and a third column element in the rotation matrix, R (3, 1) is a third row and a first column element in the rotation matrix, R (3, 2) is a third row and a second column element in the rotation matrix, and R (3, 3) is a third row and a third column element in the rotation matrix; t is a translation vector, T (1, 1) is a first row and first column element in the translation vector, T (2, 1) is a second row and first column element in the translation vector, and T (3, 1) is a third row and first column element in the translation vector;
the coordinate conversion model is as follows:
Figure FDA0003729612170000041
wherein xtargetIs the x coordinate, y of any point in the target point cloudtargetIs the y coordinate, z, of any point in the target point cloudtargetIs the z coordinate, x, of any point in the target point cloudsourceIs the x coordinate, y of any point in the source point cloudsourceIs the y coordinate, z, of any point in the source point cloudsourceIs the z coordinate of any point in the target point cloud.
CN202210846607.6A 2022-07-05 2022-07-05 Bridge infrastructure laser scanning and inclined live-action point cloud data splicing method Pending CN115272075A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210846607.6A CN115272075A (en) 2022-07-05 2022-07-05 Bridge infrastructure laser scanning and inclined live-action point cloud data splicing method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210846607.6A CN115272075A (en) 2022-07-05 2022-07-05 Bridge infrastructure laser scanning and inclined live-action point cloud data splicing method

Publications (1)

Publication Number Publication Date
CN115272075A true CN115272075A (en) 2022-11-01

Family

ID=83768537

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210846607.6A Pending CN115272075A (en) 2022-07-05 2022-07-05 Bridge infrastructure laser scanning and inclined live-action point cloud data splicing method

Country Status (1)

Country Link
CN (1) CN115272075A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117128861A (en) * 2023-10-23 2023-11-28 常州市建筑材料研究所有限公司 Monitoring system and monitoring method for station-removing three-dimensional laser scanning bridge

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117128861A (en) * 2023-10-23 2023-11-28 常州市建筑材料研究所有限公司 Monitoring system and monitoring method for station-removing three-dimensional laser scanning bridge

Similar Documents

Publication Publication Date Title
CN104599272B (en) Towards the airborne LiDAR point cloud and image association method for registering of removable target ball
CN106570905B (en) A kind of noncooperative target point cloud initial attitude verification method
CN111540005B (en) Loop detection method based on two-dimensional grid map
CN106023298A (en) Point cloud rigid registration method based on local Poisson curved surface reconstruction
CN114862932B (en) BIM global positioning-based pose correction method and motion distortion correction method
CN109146990B (en) Building outline calculation method
CN115272075A (en) Bridge infrastructure laser scanning and inclined live-action point cloud data splicing method
CN108764157B (en) Building laser foot point extraction method and system based on normal vector Gaussian distribution
CN111474443A (en) Method and device for measuring power transmission line
CN115564926A (en) Three-dimensional patch model construction method based on image building structure learning
CN113409332B (en) Building plane segmentation method based on three-dimensional point cloud
CN103065295A (en) Aviation and ground lidar data high-precision automatic registering method based on building angular point self-correction
CN113721254A (en) Vehicle positioning method based on road fingerprint space incidence matrix
CN116051540B (en) Method and system for acquiring positioning pose of transformer wiring terminal based on point cloud model
CN107240133A (en) A kind of stereoscopic vision mapping model method for building up
CN108595373B (en) Uncontrolled DEM registration method
CN111598941A (en) Method, device and equipment for measuring gradient of tower and storage medium
CN115512067A (en) Structural component parameter extraction and deformation measurement method based on laser point cloud
Sun et al. Datum feature extraction and deformation analysis method based on normal vector of point cloud
CN114677435A (en) Point cloud panoramic fusion element extraction method and system
CN114820400B (en) Airborne LiDAR point cloud ground point filtering method
Ye et al. Gaussian Mixture Model of Ground Filtering Based on Hierarchical Curvature Constraints for Airborne Lidar Point Clouds
Xu et al. A LiDAR data-based camera self-calibration method
CN114842084B (en) Map construction method and device and mobile detection equipment
CN117058358B (en) Scene boundary detection method and mobile platform

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