CN113468706B - Laser point cloud power transmission line lead fitting method for distribution network live working robot - Google Patents

Laser point cloud power transmission line lead fitting method for distribution network live working robot Download PDF

Info

Publication number
CN113468706B
CN113468706B CN202110845733.5A CN202110845733A CN113468706B CN 113468706 B CN113468706 B CN 113468706B CN 202110845733 A CN202110845733 A CN 202110845733A CN 113468706 B CN113468706 B CN 113468706B
Authority
CN
China
Prior art keywords
point cloud
curve
fitting
point
lead
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.)
Active
Application number
CN202110845733.5A
Other languages
Chinese (zh)
Other versions
CN113468706A (en
Inventor
胡子琦
苑晶
公岩松
俞诗卓
张雪波
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.)
Nankai University
Original Assignee
Nankai University
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 Nankai University filed Critical Nankai University
Priority to CN202110845733.5A priority Critical patent/CN113468706B/en
Publication of CN113468706A publication Critical patent/CN113468706A/en
Application granted granted Critical
Publication of CN113468706B publication Critical patent/CN113468706B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/18—Network design, e.g. design based on topological or interconnect aspects of utility systems, piping, heating ventilation air conditioning [HVAC] or cabling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/25—Design optimisation, verification or simulation using particle-based methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/08—Probabilistic or stochastic CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/04—Power grid distribution networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Biophysics (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Numerical Control (AREA)
  • Laser Beam Processing (AREA)

Abstract

The invention belongs to the field of electric power robots, and particularly relates to a method for fitting a laser point cloud power transmission line lead of a distribution network live working robot. S1, acquiring three-dimensional point cloud of a lead of a power transmission line, and processing the three-dimensional point cloud by adopting an Euclidean distance filter; s2, performing point cloud down-sampling on the filtered data by adopting a self-organizing layered particle swarm optimization algorithm; s3, inputting the point cloud after down-sampling, and fitting a B spline curve based on iterative optimization of control points; and S5, fusing the control points and outputting the final B spline curve.

Description

Laser point cloud power transmission line lead fitting method for distribution network live working robot
Technical Field
The invention belongs to the field of electric power robots, and particularly relates to a method for fitting a laser point cloud power transmission line lead of a distribution network live working robot.
Background
A distribution network live working robot is an intelligent device applied to assembly and maintenance of a distribution network live working line. In recent years, with the development of live working technology, distribution network live working robots are receiving more and more extensive attention. Under the assistance of sensors such as laser and the like, the distribution network live working robot can realize distribution network environment perception under the condition of low illumination, realize live working and effectively improve the working safety. In the process of live working, the motion planning of the mechanical arm is one of many basic tasks. In a complex operating environment, motion planning and operation control requires sensing spatial location information of various power line leads in the environment. Typically, the robot will fit the spatial parametric representation of the wire through sensor information such as a lidar or camera.
The existing three-dimensional reconstruction method of the transmission line mainly aims at the transmission line, and the transmission line is more regular in shape and can be approximately considered to be in accordance with a quadratic polynomial function. And the estimation of the power transmission line curve model is realized through least square fitting based on quadratic polynomial curve hypothesis. However, due to the assumption of a quadratic polynomial form on the curve model to be fitted, accurate curve fitting cannot be achieved on power line leads having a more strongly nonlinear form. Specifically, according to the characteristics that two ends of a power transmission line are suspended and the middle of the power transmission line naturally droops, the method solves the optimal geometric model of line fitting; then, by establishing an optimal plane coordinate system for power line fitting, a least square method based on quadratic polynomial curve hypothesis is adopted to fit the row lines. Since the power line leads have a more complex morphology than the row lines, this method is not suitable for fitting the lead parameter equations.
Disclosure of Invention
Therefore, a laser point cloud lead fitting method based on hierarchical self-organizing particle swarm optimization and control point iterative optimization is needed to be provided, and accurate three-dimensional reconstruction of the lead of the power transmission line is achieved.
In order to achieve the purpose, the invention adopts the following technical scheme:
a method for fitting a laser point cloud power transmission line lead of a distribution network live working robot comprises the following steps,
s1, acquiring three-dimensional point cloud of the lead of the power transmission line, and processing by adopting an Euclidean distance filter;
s2, performing point cloud down-sampling on the filtered data by adopting a self-organizing layered particle swarm optimization algorithm;
s3, inputting the point cloud after down-sampling, and fitting a B spline curve based on iterative optimization of control points;
and S5, fusing the control points and outputting the final B spline curve.
In the step S2, heuristically calculating the size of the voxel in the downsampling by using SOH-PSO, fitting a new curve equation by using a B-spline after updating the downsampling voxel side length each time in an iteration manner, and evaluating the new downsampling voxel side length, namely the particle state, by using an evaluation function; the merit function is defined as:
Figure BDA0003180707390000021
Figure BDA0003180707390000022
wherein p and f are the laser points and the fitting curve, phi is the set formed by all the laser points after the outer points are removed, and kappamaxIs the maximum curvature value, ω, on the fitted curve1And ω2Is a weighting coefficient, N is the number of three-dimensional points of the original point cloud, d (p, f) is the Euclidean distance from the point p to the closest point of the fitting curve, dmax(p, f) is the maximum of the distance of p to the closest point of the fitted curve, DpMeaning the average distance of the point cloud to the fitted curve.
In the further optimization of the technical scheme, the step 2 of point cloud down-sampling comprises the following steps,
s21, inputting initial point cloud data, and initializing the position and speed of a particle swarm;
s22, traversing each particle, determining the side length of a down-sampled voxel according to the state of the particle, then down-sampling the point cloud, fitting a curve equation to the down-sampled point cloud by using a B spline, and updating p according to a formula (1)ikAnd pgThe two parameters respectively represent the optimal state of the current particle and the optimal states of all the particles;
s23, calculating the speed of the current particle:
vik＝c1×rand1(·)×(pik-xik)+c2×rand2(·)×(pg-xik) (3)
wherein, c1，c2Is a constant parameter, rand1(. o) and rand2(. is) is [0, 1)]Random number of interval, xikIs the state of the ith particle at the moment k, the voxel side length;
s24, if vikIf the calculation is zero, then [0,1 ] is generated]Random number rand of interval3(. o) and determining whether the random number is greater than 0.5, if so, vik＝-rand5(. times v), otherwise, vik＝rand4(. x v, v is a constant parameter representing the reinitialization speed, rand4(. o) and rand5(. is) is [0, 1)]Random number of interval, vik＝sign(vik)×min(vik,vmax)，vmaxIs a constant parameter representing the maximum velocity of the particle;
s25, updating the state of the particles, if the search iteration times are large enough or the evaluation function value of the optimal particles is small enough, terminating the iteration, otherwise, returning to the step 2;
and S26, outputting the optimal state of all particles, namely the side length of the voxel.
In a further optimization of the present technical solution, in step S3, the point cloud after downsampling is input, the geometric center of the point cloud in each voxel is used as an initial value of a B-spline interpolation control point, a smooth and continuous curve is fitted to a given ordered sequence of control points through a B-spline interpolation function, and an equation of the fitted curve can be expressed as:
Figure BDA0003180707390000031
wherein B (-) is a basis function, ckAt the kth control point, the basis function of an n-th order B-spline curve can be expressed as:
Figure BDA0003180707390000032
Figure BDA0003180707390000033
Figure BDA0003180707390000034
wherein pi ═ xi (ξ)0,ξ1,ξ2,...,ξn+1)TIs a node vector.
In the further optimization of the technical scheme, the iterative algorithm of the control points comprises the following steps,
s31, calculating local extreme points of the curve on the fitting curve by using a non-maximum suppression algorithm, where κ is { κ ═ κ {0,κ1,...,κv}；
S32, finding the nearest control point c in the vicinity of kk，
S33, using c in the original point cloud datakAs the center point, c is calculated by K nearest neighbor algorithmkThe geometric center of the point cloud in the neighborhood is used as a new control point ck'；
S34, control point set C ═ C0,c1,...,cmRepeat this process until the maximum value of curvature in the fitted curve is less than a set threshold lambdaγ＝5.40
And S35, and finally outputting the fitted curve equation shown in the formula (5).
In a further optimization of the technical scheme, in the step S4, the control points are fused, and all the control points are added into a point set Q to be fused, and the control points p are sequentially taken out from the Qm(ii) a Searching for p using KNNmAll points in the neighborhood q ═ pm,p1 m,p2 m,...,pn m}; adding the geometric center of Q to the set QnewAnd Q is deleted from Q; this loop is performed until Q is an empty set; to QnewThe new control points in (1) recalculate the B-spline curve as the final fitted lead model.
And further optimizing the technical scheme, wherein in the step S1, the three-dimensional point cloud of the power transmission line lead is obtained by scanning with a laser radar.
According to the further optimization of the technical scheme, the B spline curve is a 3-order B spline curve.
Different from the prior art, the technical scheme has the following beneficial effects:
1. the problems that the distribution network live working robot senses and reconstructs a power transmission line lead in a distribution network live environment by using the laser Lidar are solved.
2. The improved B spline interpolation function based on SOH-PSO point cloud down-sampling and control point iterative optimization is utilized to improve the fitting precision of the lead of the power transmission line under the complex form and the reduction degree of the lead posture.
3. By utilizing the novel particle swarm optimization algorithm and the control point iterative optimization algorithm, the problem that the local curvature is too large easily occurs in the process of fitting the B spline interpolation value to the discrete laser point cloud is solved.
Drawings
FIG. 1 is a flow chart of a laser point cloud power line lead fitting method;
FIG. 2 is a flow chart of a control point iterative optimization algorithm;
FIG. 3 is a graph of the fit error for two evaluation indices for different methods;
FIG. 4 is a diagram of the fitting error of four algorithms under different point cloud variances;
FIG. 5 is a schematic view of an envelope of a fitted curve;
FIG. 6 is a schematic representation of a lead line representation obtained by real laser point cloud fitting;
fig. 7 is a graph of the ratio of inner points of four curve fitting methods at different envelope surface radii.
Detailed Description
To explain technical contents, structural features, and objects and effects of the technical solutions in detail, the following detailed description is given with reference to the accompanying drawings in conjunction with the embodiments.
Referring to fig. 1, a flow chart of a laser point cloud power line lead fitting method is shown. The invention provides a method for fitting a laser point cloud power transmission line lead of a distribution network live working robot, which comprises the following steps,
s1, acquiring three-dimensional point cloud of the lead of the power transmission line, and processing by adopting an Euclidean distance filter;
s2, performing point cloud down-sampling on the filtered data by adopting a self-organizing layered particle swarm optimization algorithm;
s3, inputting the point cloud after down-sampling, and fitting a B spline curve based on iteration of control points;
and S5, fusing the control points and outputting the final B spline curve.
1. Collecting data
And (4) scanning the three-dimensional point cloud of the lead of the power transmission line by using a laser radar.
2. Point cloud down-sampling
Before down-sampling, outliers are first removed using a euclidean distance filter. In the process of fitting the lead using the B-spline function, a certain number of control points need to be screened from the point cloud of the lead. If a part of control points are screened out without down-sampling lead point clouds, and B-spline fitting is directly used on the original point clouds, the local curvature of a fitting curve is too large easily due to the dense point cloud distribution, the overall attitude expression is influenced, and huge calculation cost can be increased. The strategy for screening control points in this embodiment is to divide the three-dimensional space evenly into voxel squares of equal size by point cloud downsampling, and downsample all the point clouds falling in each voxel to their geometric center points. And substituting the control point obtained by down-sampling as an initial value into a subsequent iterative algorithm for optimization. It is noted that the choice of voxel size in the down-sampling has a significant influence on the number and location of control points. When the voxel is too large, the deviation between the geometric center of the lead point cloud in the voxel and the real position of the lead is too large, and the position precision of the control point is further influenced; when the voxels are too small, the control points are distributed too densely, and the local curvature is easily too large.
In the optimization of the lead point cloud downsampling voxel side length, the value range of the initial value of the particle state is wide. The traditional particle swarm optimization algorithm is sensitive to a set strategy of a voxel side length initial interval. The particle swarm optimization algorithm based on self-organizing layering enables the speed of the particles to be updated and concentrated on the optimal state of each interval by removing the speed item at the previous moment; meanwhile, the speed is reinitialized with a certain probability, and the situation of falling into a local optimal state is improved. In this way, the SOH-PSO can converge to an appropriate downsampled voxel edge length value with less calculation cost and faster speed under the condition of a larger initial particle distribution interval. Thus, this embodiment uses the SOH-PSO to heuristically calculate the size of the voxels in the downsampled. After the side length of the down-sampled voxel is updated in each iteration, a B spline is reused for fitting a new curve equation, and an evaluation function is adopted for evaluating the side length of the new down-sampled voxel, namely the particle state. The merit function is defined as:
Figure BDA0003180707390000051
Figure BDA0003180707390000052
wherein p and f are the laser points and the fitting curve respectively, and phi is the set formed by all the laser points after the outer points are removed. KappamaxIs the maximum curvature value, ω, on the fitted curve1And ω2Is a weighting coefficient, N is the number of three-dimensional points of the original point cloud, d (p, f) is the Euclidean distance from the point p to the closest point of the fitting curve, dmax(p, f) is the maximum of the distance of p to the closest point of the fitted curve, DpMeaning the average distance of the point cloud to the fitted curve. The smaller the evaluation function is, the smoother the representative curve is, and the smaller the euclidean distance from the point cloud is.
Point cloud down-sampling detailed steps:
and S21, inputting initial point cloud data, and initializing the position and the speed of the particle swarm.
S222, traversing each particle, determining the side length of the down-sampled voxel according to the state of the particle, and then down-sampling the point cloud. The point cloud after downsampling is fitted to a curve equation using B-splines. Updating p according to equation (1)ikAnd pgThe two parameters respectively representing the current particleOptimal state and optimal state of all particles.
S23, calculating the speed of the current particle:
vik＝c1×rand1(·)×(pik-xik)+c2×rand2(·)×(pg-xik) (3)
wherein, c1，c2Is a constant parameter, rand1(. o) and rand2(. is) is [0, 1)]Random number of interval, xikIs the state (voxel side length) at the instant of the ith particle k.
S24, if vikIf the calculation is zero, then [0,1 ] is generated]Random number rand of interval3(. cndot.) and determines whether the random number is greater than 0.5. If greater than, vik＝-rand5(. times.v; otherwise, vik＝rand4(. times.v). v is a constant parameter representing the reinitialization speed. rand4(. o) and rand5(. is) is [0, 1)]Random number of intervals. v. ofik＝sign(vik)×min(vik,vmax)，vmaxIs a constant parameter representing the maximum velocity of the particle.
And S25, updating the state of the particles. The number of search iterations is sufficiently large, or the evaluation function value of the optimum particle is sufficiently small, and the iteration is terminated. Otherwise, returning to the step 2.
And S26, outputting the optimal state of all particles, namely the side length of the voxel.
3. Iterative optimization of control points
The input is a point cloud after downsampling. And taking the geometric center of the point cloud in each voxel as an initial value of the B-spline interpolation control point. A smooth continuous curve can be fitted to a given ordered sequence of control points by means of a B-spline interpolation function. The equation for fitting the curve can be expressed as:
Figure BDA0003180707390000061
wherein B (-) is a basis function. c. CkIs the kth control point. Basis function of n-th order B-spline curveThe number may be expressed as:
Figure BDA0003180707390000062
Figure BDA0003180707390000063
Figure BDA0003180707390000064
wherein pi ═ xi (ξ)0,ξ1,ξ2,...,ξn+1)TIs a node vector (knock vector). The embodiment adopts a 3-order B-spline curve, and minimizes the calculation cost while ensuring the smoothness of the curve.
Because the positions of the control points after the down-sampling cannot be strictly distributed on the lead, the situation that the local curvature is too large and the shape of the lead is seriously deviated from the actual lead shape easily occurs on a fitting curve after the traditional B-spline interpolation. In order to optimize the position of the control point, the embodiment proposes an algorithm for iterative optimization of the control point, which is a flowchart of the iterative optimization algorithm of the control point, as shown in fig. 2.
The control point iterative optimization algorithm comprises the following steps:
s31, calculating local extreme points of the curve on the fitting curve by using Non-Maximum Suppression (NMS) algorithm, and marking as κ ═ κ0,κ1,...,κv}。
S32, finding the nearest control point c in the vicinity of kk。
S33, using c in the original point cloud datakC is calculated by K-Nearest Neighbor (KNN) algorithm as the center pointkThe geometric center of the point cloud in the neighborhood is used as a new control point ck'。
S34, control point set C ═ C0,c1,...,cmRepeat this process until the maximum value of curvature in the fitted curve is less than a set threshold lambdaγ＝5.40。
And S35, and finally outputting the fitted curve equation shown in the formula (5).
4. Control point fusion
After the iterative algorithm is converged, the distance between the control points is less than lambdacontrolAnd (5) fusing the control points of 0.12 to optimize the number of the control points. In the fusion process, all control points are added into a point set Q to be fused, and control points p are taken out from Q in sequencem(ii) a Searching for p using KNNmNeighborhood (distance p)mLess than 0.1m) of the points q ═ pm,p1 m,p2 m,...,pn m}; adding the geometric center of Q to the set QnewAnd Q is deleted from Q; this loop is performed until Q is an empty set; to QnewThe new control points in (3) recalculate the B-spline curve as the final fitted lead model, as shown in equation (5).
5. Simulation experiment
In order to verify the fitting effect of the method of the embodiment on the three-dimensional point cloud curve, the method is compared with a quadratic polynomial-based least square fitting method which is proposed in B-Spline Interpolation, Bessel curve (Hou H.S and Andrews H.C. C. customer Spline for Image interaction and Digital filtration [ J ]. IEEE Transactions on optics, Speech and Signal Processing,1978,26(6):508 + 517 ]) and literature (Reinhaus, Zhang Jixian. overhead transmission line airborne laser radar point cloud power line three-dimensional reconstruction [ J ] surveying and mapping report, 2016,45(03):347 + 353 ]) on simulation data. The three simulation curve equations used in this patent are respectively:
Figure BDA0003180707390000071
Figure BDA0003180707390000081
Figure BDA0003180707390000082
firstly, random three-dimensional point clouds with different quantities and variances are generated around three simulation curves through normal distribution and are used for simulating laser point clouds obtained by a robot in a real scene. And then, fitting the simulated laser point cloud by using the method, the B spline curve, the Bezier curve and the least square algorithm provided by the patent. The method selects Dynamic Time Warping Distance (Dynamic Time Warping Distance) and Hausdorff Distance (Hausdorff Distance) as evaluation indexes of curve fitting precision, and calculates the error between a real curve and a fitting curve. Specifically, the calculation formula of the dynamic time warping distance is as follows:
Figure BDA0003180707390000083
where Q and C are two discrete curve sequences, ωkIs the kth mapping obtained using dynamic time warping in both sequences. The formula for computing the hausdorff distance is:
Figure BDA0003180707390000084
where d (-) denotes the euclidean distance between the two points a, b. The Hausdorff distance defines the maximum distance from one point cloud set to the closest point of another point cloud set, so that the similarity of two discrete point cloud tracks can be indirectly evaluated.
In simulation, parameter c1＝c 22, v is 0.05. According to the comparison of multiple sets of simulation, when the parameter lambda isγ＝5.40，λcontrolWhen the lead wire fitting ratio is 0.12, a good lead wire fitting result can be obtained. The parameters of the three sets of test point clouds are shown in table 1. And uniformly extracting 30 samples from the three groups of test point clouds from small to large according to the distribution interval of the number of the point clouds. Evaluation indexes of the four methods for the fitting result of the test point cloud are shown in fig. 3, and are fitting error graphs of the two evaluation indexes of different methods. FIG. 3 shows only the algorithm simulations for Housdov distances with an error within 10.0mAnd (6) synthesizing the result. When the fitting error exceeds 10.0m, the fitting is considered to fail and is not shown in fig. 3. Table 2 and table 3 show the mean fit error for the four algorithms under two evaluation indices. It can be seen that the algorithm provided by the patent has more robust and accurate fitting results on different point cloud numbers, measurement indexes and variances. Meanwhile, the method provided by the patent is insensitive to the change of the point cloud number and the distribution variance, and is more suitable for fitting and rebuilding the lead of the power transmission line in the actual power distribution operation environment.
TABLE 1 three sets of simulated point cloud parameters
Figure BDA0003180707390000085
Figure BDA0003180707390000091
Table 2 mean error of four algorithms at dynamic time planning distance (.' indicates failure of fit)
Figure BDA0003180707390000092
Table 3 mean error of the four algorithms at the hausdorff distance (.' indicates failure of fit)
Figure BDA0003180707390000093
In order to further verify the robustness of the algorithm to the point cloud distribution variance, under the condition that the number of the point clouds is fixed, 16 groups of point cloud samples with the variance uniformly distributed between [0.01 and 0.08] from small to large are generated by the curve I. The test results are shown in fig. 4, which is a fitting error graph of four algorithms under different point cloud variances. As can be seen from the results, the algorithm disclosed by the patent has a relatively stable fitting result for point clouds with different variance distributions.
6. Actual testing
In order to verify the processing effect of the algorithm on the real lead point cloud, a comparison experiment is carried out on four curve fitting algorithms by using the lead point cloud obtained by scanning of the laser sensor. Because there is no real wire equation, the fitting error between the fitted wire and the real wire cannot be strictly calculated. Instead, by generating an envelope surface as shown in fig. 5 with different radii around the fitting lead, fig. 5 is a schematic diagram of the envelope surface of the fitting curve, and counting the proportion of the number of laser scanning points in the envelope surface to the total number of scanning points, i.e., the proportion of inner points, the effect of lead fitting is indirectly evaluated. In actual testing, parameter c1＝c 22, v is 0.05. According to comparison of multiple groups of actual tests, when the parameter lambda isγ＝6.20，λcontrolWhen the lead wire fitting ratio is 0.15, a good lead wire fitting result can be obtained. Because the power transmission line lead has certain structural toughness for the local curvature of lead can not appear too big the condition. Referring to fig. 6, a schematic diagram of a lead representation obtained by real laser point cloud fitting is shown in fig. 6, which shows lead models respectively obtained by four algorithm fitting. Compared with the other two classical algorithms, the method does not cause the situation of overlarge local curvature, and the overall shape of the lead is more accurately described. Referring to fig. 7, an inner point scale diagram of four curve fitting methods at different envelope surface radii is shown. Fig. 7 shows the inner point ratios of the four algorithms at different envelope radii. It can be seen that the algorithm of the patent has a larger interior point proportion under different radiuses, which shows that the lead point cloud can be better fitted by utilizing the algorithm of the patent.
The method provided by the invention has the following advantages:
1. the method reduces the manual participation degree in the distribution network live-line task and improves the operation safety.
2. The method is suitable for fitting the laser point cloud lead wires in various forms and has a wide application range.
3. The lead is reconstructed through automatic sensing, so that a foundation can be laid for the subsequent mechanical arm obstacle avoidance planning, the labor of manually operating the mechanical arm is saved, and the time and the labor are saved.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrases "comprising … …" or "comprising … …" does not exclude the presence of additional elements in a process, method, article, or terminal that comprises the element. Further, herein, "greater than," "less than," "more than," and the like are understood to exclude the present numbers; the terms "above", "below", "within" and the like are to be understood as including the number.
Although the embodiments have been described, once the basic inventive concept is obtained, other variations and modifications of these embodiments can be made by those skilled in the art, so that the above embodiments are only examples of the present invention, and not intended to limit the scope of the present invention, and all equivalent structures or equivalent processes using the contents of the present specification and drawings, or any other related technical fields, which are directly or indirectly applied thereto, are included in the scope of the present invention.

Claims (6)

1. A laser point cloud power transmission line lead fitting method for distribution network live working robots is characterized by comprising the following steps,
s1, acquiring three-dimensional point cloud of the lead of the power transmission line, and processing by adopting an Euclidean distance filter;
s2, performing point cloud down-sampling on the filtered data by adopting a self-organizing layered particle swarm optimization algorithm;
step S2, calculating the size of the voxel in the downsampling in a heuristic manner by adopting an SOH-PSO method, fitting a new curve equation by reusing a B spline after updating the side length of the downsampling voxel in each iteration, and evaluating the side length of the new downsampling voxel, namely the particle state by adopting an evaluation function; the merit function is defined as:
Figure FDA0003501649010000011
Figure FDA0003501649010000012
wherein p and f are the laser points and the fitting curve, phi is the set formed by all the laser points after the outer points are removed, and kappamaxIs the maximum curvature value, ω, on the fitted curve1And ω2Is a weighting coefficient, N is the number of three-dimensional points of the original point cloud, d (p, f) is the Euclidean distance from the point p to the closest point of the fitting curve, dmax(p, f) is the maximum of the distance of p to the closest point of the fitted curve, DpMean distance of the point cloud to the fitted curve;
s3, inputting the point cloud after down-sampling, and fitting a B spline curve based on iterative optimization of control points;
the step S3 inputs the point cloud after downsampling, the geometric center of the point cloud in each voxel is used as the initial value of the B-spline interpolation control point, a smooth and continuous curve is fitted to a given ordered control point sequence through the B-spline interpolation function, and the equation of the fitted curve can be expressed as:
Figure FDA0003501649010000013
wherein B (-) is a basis function, ckAt the kth control point, the basis function of an n-th order B-spline curve can be expressed as:
Figure FDA0003501649010000014
Figure FDA0003501649010000015
Figure FDA0003501649010000016
wherein pi ═ xi (ξ)0,ξ1,ξ2,...,ξn+1)TIs a node vector;
and S4, fusing the control points and outputting the final B spline curve.
2. The laser point cloud power line lead fitting method for distribution network live working robots according to claim 1, characterized in that the step S2 point cloud down-sampling comprises the following steps,
s21, inputting initial point cloud data, and initializing the position and speed of a particle swarm;
s22, traversing each particle, determining the side length of a down-sampled voxel according to the state of the particle, then down-sampling the point cloud, fitting a curve equation to the down-sampled point cloud by using a B spline, and updating p according to a formula (1)ikAnd pgThe two parameters respectively represent the optimal state of the current particle and the optimal states of all the particles;
s23, calculating the speed of the current particle:
vik＝c1×rand1(·)×(pik-xik)+c2×rand2(·)×(pg-xik) (3)
wherein, c1，c2Is a constant parameter, rand1(. o) and rand2(. is) is [0, 1)]Random number of interval, xikIs the state of the ith particle at the moment k, the voxel side length;
s24, if vikIf the calculation is zero, then [0,1 ] is generated]Random number rand of interval3(. o) and determining whether the random number is greater than 0.5, if so, vik＝-rand5(. times v), otherwise, vik＝rand4(. x v, v is a constant parameter representing the reinitialization speed, rand4(. o) and rand5(. is) is [0, 1)]Random number of interval, vik＝sign(vik)×min(vik,vmax)，vmaxIs a constant parameter representing the maximum velocity of the particle;
s25, updating the state of the particles, if the search iteration times are large enough or the evaluation function value of the optimal particles is small enough, terminating the iteration, otherwise, returning to the step 2;
and S26, outputting the optimal state of all particles, namely the side length of the voxel.
3. The method for fitting the laser point cloud power line lead of the distribution network live working robot as claimed in claim 1, wherein the iterative algorithm steps of the control points are as follows,
s31, calculating local extreme points of the curve on the fitting curve by using a non-maximum suppression algorithm, where κ is { κ ═ κ {0,κ1,...,κv}；
S32, finding the nearest control point c in the vicinity of kk，
S33, using c in the original point cloud datakAs the center point, c is calculated by K nearest neighbor algorithmkThe geometric center of the point cloud in the neighborhood is used as a new control point ck'；
S34, control point set C ═ C0,c1,...,cmRepeat this process until the maximum value of curvature in the fitted curve is less than a set threshold lambdaγ＝5.40
And S35, and finally outputting the fitted curve equation shown in the formula (5).
4. The method for fitting the laser point cloud power line lead of the distribution network live working robot as claimed in claim 1, wherein the control points are fused in step S4, all the control points are added into a point set Q to be fused, and the control points p are taken out from Q in sequencem(ii) a Searching for p using KNNmAll points in the neighborhood q ═ pm,p1 m,p2 m,...,pn m}; adding the geometric center of Q to the set QnewAnd Q is deleted from Q; this loop is performed until Q is an empty set; to QnewThe new control points in (1) recalculate the B-spline curve as the final fitted lead model.
5. The method for fitting the laser point cloud powerline lead of the robot for distribution network live working according to claim 1, wherein said step S1 is performed by scanning the three-dimensional point cloud of the powerline lead by a laser radar.
6. The method for fitting the laser point cloud power line lead of the distribution network live working robot as claimed in claim 1, wherein the B-spline curve is a 3-order B-spline curve.
CN202110845733.5A 2021-07-26 2021-07-26 Laser point cloud power transmission line lead fitting method for distribution network live working robot Active CN113468706B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110845733.5A CN113468706B (en) 2021-07-26 2021-07-26 Laser point cloud power transmission line lead fitting method for distribution network live working robot

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110845733.5A CN113468706B (en) 2021-07-26 2021-07-26 Laser point cloud power transmission line lead fitting method for distribution network live working robot

Publications (2)

Publication Number Publication Date
CN113468706A CN113468706A (en) 2021-10-01
CN113468706B true CN113468706B (en) 2022-03-15

Family

ID=77882508

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110845733.5A Active CN113468706B (en) 2021-07-26 2021-07-26 Laser point cloud power transmission line lead fitting method for distribution network live working robot

Country Status (1)

Country Link
CN (1) CN113468706B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114155168A (en) * 2021-12-08 2022-03-08 武汉中海庭数据技术有限公司 Full-automatic offset correction method and system for point cloud data
CN115131417A (en) * 2022-07-19 2022-09-30 南开大学 Laser point cloud 2D-3D bimodal interaction enhancement irregular wire detection method
CN115480519B (en) * 2022-11-04 2023-02-21 济南邦德激光股份有限公司 B-spline curve fitting method, device and medium based on orthogonal trigonometric decomposition

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105023053A (en) * 2015-04-27 2015-11-04 中国南方电网有限责任公司电网技术研究中心 Method and system for electric quantity distribution

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105023053A (en) * 2015-04-27 2015-11-04 中国南方电网有限责任公司电网技术研究中心 Method and system for electric quantity distribution

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
Self-Organizing Hierarchical Particle Swarm Optimization for Nonconvex Economic Dispatch;K. T. Chaturvedi 等;《IEEE TRANSACTIONS ON POWER SYSTEMS》;20080831;全文 *
Self-Organizing Hierarchical Particle Swarm Optimization with Time-Varying Acceleration Coefficients for Economic Dispatch with Valve Point Effects and Multifuel Options;Jirawadee Polprasert 等;《American Institute of Physics》;20111231;全文 *
基于分层粒子群优化的三维点云配准;黄筱佟 等;《济南大学学报》;20210701;全文 *
基于最少控制点的非均匀有理B样条曲线拟合;周红梅 等;《西安交通大学学报》;20080131;全文 *
激光点云中输电线拟合与杆塔定位方法研究;游安清 等;《计算机科学》;20130430;全文 *

Also Published As

Publication number Publication date
CN113468706A (en) 2021-10-01

Similar Documents

Publication Publication Date Title
CN113468706B (en) Laser point cloud power transmission line lead fitting method for distribution network live working robot
CN107516326B (en) Robot positioning method and system fusing monocular vision and encoder information
Lim et al. Surface reconstruction techniques: a review
Ivrissimtzis et al. Using growing cell structures for surface reconstruction
CN111986115A (en) Accurate elimination method for laser point cloud noise and redundant data
CN112001958B (en) Virtual point cloud three-dimensional target detection method based on supervised monocular depth estimation
CN113139996B (en) Point cloud registration method and system based on three-dimensional point cloud geometric feature learning
CN112212861B (en) Track restoration method based on single inertial sensor
CN112233249A (en) B spline surface fitting method and device based on dense point cloud
CN112444246A (en) Laser fusion positioning method in high-precision digital twin scene
CN114237235B (en) Mobile robot obstacle avoidance method based on deep reinforcement learning
CN112762957A (en) Multi-sensor fusion-based environment modeling and path planning method
CN112288859B (en) Three-dimensional face modeling method based on convolutional neural network
CN114429432B (en) Multi-source information layered fusion method and device and storage medium
CN116310219A (en) Three-dimensional foot shape generation method based on conditional diffusion model
CN111260765A (en) Dynamic three-dimensional reconstruction method for microsurgery operative field
CN111815684B (en) Space multivariate feature registration optimization method and device based on unified residual error model
CN112802089A (en) Point cloud skeleton line extraction method and system based on automatic estimation of bifurcation number
CN117036447A (en) Indoor scene dense three-dimensional reconstruction method and device based on multi-sensor fusion
CN115950414A (en) Adaptive multi-fusion SLAM method for different sensor data
CN108469729B (en) Human body target identification and following method based on RGB-D information
CN113850304A (en) High-accuracy point cloud data classification segmentation improvement algorithm
CN114332156A (en) Real-time three-dimensional motion completion method based on graph convolution neural network
Wang et al. Octree 3D Visualization Mapping based on Camera Information
CN116433493B (en) Workpiece point cloud set splicing method based on metric learning

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