CN115641406A - Partial overlapping segmentation and clustering method for power line laser point cloud data - Google Patents
Partial overlapping segmentation and clustering method for power line laser point cloud data Download PDFInfo
- Publication number
- CN115641406A CN115641406A CN202211324256.9A CN202211324256A CN115641406A CN 115641406 A CN115641406 A CN 115641406A CN 202211324256 A CN202211324256 A CN 202211324256A CN 115641406 A CN115641406 A CN 115641406A
- Authority
- CN
- China
- Prior art keywords
- power line
- points
- model
- catenary
- segment
- 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
Links
Images
Landscapes
- Length Measuring Devices By Optical Means (AREA)
Abstract
The invention discloses a partial overlapping segmentation and clustering method for power line laser point cloud data, which comprises the following steps: s1, determining a mathematical expression of a three-dimensional reconstruction model of a single-gear single power line; the model comprises a straight line segment model and a catenary line segment model; s2, data preprocessing and partial superposition data segmentation; and S3, model fitting, line segment labeling and synthesizing. The method solves the problem that the existing method cannot meet the power line laser point cloud clustering of various configurations, solves the problem that the existing method cannot meet the identification of split conductor laser points, realizes the correct segmentation of power line point clouds with any length, any tower height difference and irregular fracture and thickness, and improves the accuracy, the automation degree, the robustness and the universality of single power line laser radar point identification.
Description
Technical Field
The invention relates to the technical field of three-dimensional reconstruction of power lines, in particular to a partial overlapping segmentation and clustering method for laser point cloud data of a power line.
Background
In power line patrol, three-dimensional reconstruction of a power line plays an important role, which is the basis of important applications such as dangerous point detection, conductor phase-to-phase height difference measurement, inter-conductor distance measurement, conductor-to-ground distance measurement, three-dimensional visualization, conductor sag analysis, conductor icing analysis, conductor windage yaw analysis and the like, and is one of the focuses of research. In the existing research, the three-dimensional reconstruction of the power line comprises 4 important components: the method comprises the steps of power line laser point detection, single-gear power line laser point identification, single-gear power line laser point cloud clustering (namely determining the total root/number of single-gear power lines and determining the laser point of a single power line) and single power line three-dimensional modeling. The single-gear power line laser point cloud clustering is the most critical and complex link and is also the focus of research. Currently, the related clustering methods are: the method comprises a Hough transformation method, a 3D connected component analysis method and a power line model growing and merging method. However, the following problems still exist in the existing research.
1) The Hough transform method cannot be applied to the situations of vertical arrangement, mixed arrangement and staggered arrangement of power lines. In addition, for a plurality of vertically arranged power lines, the severe sag phenomenon can cause that laser points of different power lines cannot be separated through an elevation value characteristic. The power line in the first level of the high-voltage line in the real world can have various configuration arrangement modes such as triangle, horizontal, vertical, mixed arrangement, staggered arrangement and the like, and the existing method cannot meet the power line laser point cloud clustering of various configurations.
2) The model growing and merging method cannot meet the identification of split conductor laser points; while the 3D connected component analysis method cannot satisfy reconstruction of any split of the split conductor. The types of power lines closely related to power line inspection can be divided into a conducting line and a lightning conductor, wherein the conducting line can be further divided into a single conducting line and a split conducting line (also called a conducting wire bundle, and can be further divided into 2, 4, 6, 8 splits and the like). In the laser point cloud, the morphological characteristics of any split (beam) of a single conductor, a lightning conductor and a split conductor are extremely similar; overall, the split conductor is different from the single conductor and the lightning conductor in morphological characteristics. And the model growing and merging method cannot be applied to the split conductor. In addition, for the split conductor, the spacer connects the splits, and the spacer point and the power line point cannot be distinguished in detail in the laser point cloud, which causes the 3D connected component analysis method to group the splits into one type, and cannot identify the split conductors.
3) The existing method is easily influenced by irregular fracture and gross error of laser point cloud.
Disclosure of Invention
The invention aims to provide a power line laser point cloud data partial overlapping segmentation and clustering method, which solves the problem that the existing method cannot meet the power line laser point cloud clustering of various configurations, solves the problem that the existing method cannot meet the identification of split conductor laser points, realizes the correct segmentation of power line point clouds with any length, any tower height difference and irregular fracture and gross error, and improves the identification precision, the automation degree, the robustness and the universality of a single power line laser radar point.
In order to achieve the purpose, the invention adopts the technical scheme that: a power line laser point cloud data partial overlapping segmentation and clustering method is characterized by comprising the following steps:
s1, determining a mathematical expression of a three-dimensional reconstruction model of a single-gear single power line; the adopted three-dimensional reconstruction model of the single-gear single power line comprises two parts: the first part is a straight line segment model, and the straight line segment is generated by least square fitting of projection points of the power line laser point cloud on an XOY plane; the second part is a catenary line segment which is positioned on the vertical plane of the straight line segment and is generated by the information of the power line laser point cloud; the straight line segment and the catenary segment are coplanar, and two end points of the straight line segment and two end points of the catenary segment have a vertical projection relation;
s2, data preprocessing and partial superposition data segmentation; carrying out preprocessing and sectional organization of power line laser point cloud data; the method comprises the following steps: processing input point cloud data and calculating a statistical value; least squares linear fitting of the straight line segments; finding length and scale factors; and a step of partially overlapping the data segments;
s3, model fitting and line segment labeling and synthesizing; model fitting and line segment labeling and synthesis are an iterative process, random sampling is adopted in each iteration to extract partial points from unlabeled laser points, whether the fitting models of the points simultaneously accord with the straight line segment model and the catenary line segment model proposed in S1 is checked, and meanwhile, the negative influence of rough difference points in laser point cloud is reduced in the iterative process;
in the step of segment labeling and synthesis, 7 key parameters are set, including: a coincidence coefficient theta; maximum number of iterations I MAX (ii) a The number of non-clustered points accounts for the total number N of laser points all Minimum ratio of R N_TH (ii) a Maximum distance D from projection point of laser point on XOY plane to fitting straight line L_MAx (ii) a The length of the fitting straight line segment of the power line to be extracted accounts for the length L of the power line sp Minimum ratio of R L_TH (ii) a Maximum distance D of laser spot to fitted catenary C_MAx (ii) a Power line to be extractedThe breaking length of the power line occupies the length L of the power line sp Minimum ratio of R B_TH 。
Preferably, the step S1 further includes:
s11, establishing a straight-line segment model: the straight-line segment model in the XOY plane includes two parts: a straight line model and two end points; the two end points are determined by the straight line model and the two extreme value scale factor points;
s111, determining a straight line model, wherein the straight line model adopts a normal line type shown in formula (1):
l=x·cosα+ysinα (1)
in the formula: alpha is a vertical line segment from the original point to the straight line, the inclination angle of the straight line where the vertical line segment is located, l is the length of the straight line segment, and two end points of the straight line segment are respectively M and N; meanwhile, the intersection point of the vertical line segment and the fitting straight line is set as P (x) fp ,y fp ) I.e. drop foot P, with scale factor s =0;
s112, determining coordinates of the two end points M and N, and solving a scale factor of each power line laser point, wherein the specific process is as follows as shown in a formula (2) or a formula (3):
let the horizontal coordinate of any power line laser point be Q (x) 0 ,y 0 ),Q(x 0 ,y 0 ) The coordinates of the projected points to the fitted straight line are Q '(x' 0 ,y′ 0 ) Calculating the scale factor s of the vertical projection point according to the formula (2) or the formula (3):
when fabs (sin α) ≧ fabs (cos α):
when fabs (sin α) < fabs (cos α):
the fabs function is a function for solving an absolute value, and after a scale factor of each power line laser point is solved, the maximum scale factor S can be obtained max And a minimum scale factorS min If the two corresponding vertical projection points are M, N respectively, M, N is two end points of the straight-line segment model;
s12, establishing a catenary line segment model: making a vertical plane perpendicular to the XOY plane by the straight line segment in the step S11, wherein a catenary line segment model in the vertical plane comprises two parts, namely a catenary and two end points; the catenary model is generated by fitting derived data of power line laser point clouds, and each point in the derived data comprises two parts: the z value of each laser point and a corresponding scale factor s; two end points are determined by projecting the two end points of the straight-line segment model obtained in the step S11 to the catenary model; catenary model:
in the formula: k. h is 1 And h 2 For the coefficient to be solved, Q and more than Q laser point cloud data can be fitted to obtain the parameters of the catenary model, wherein Q is more than or equal to 6;
calculate the catenary with parameters using theorem 1: for the catenary with parameters, setting any point on the catenary as a tangent vector, wherein the minimum distance to the point is equal to the catenary parameters;
using theorem 1, a right triangle is constructed for each tangent vector, where one side has a length of k 1 A and theta 1 The following equation:
k=(y 1 -h 2 )sin(θ 1 ) (5)
likewise, there are:
k=(y 2 -h 2 )sin(θ 2 ) (6)
equating (4) and (5), resulting in a value:
bringing it into (4), one can calculate:
when modeling the catenary, the algorithm estimates the parameter h using the first and last local models identified over the span 1 、h 2 、k , This estimate is then refined using a numerical method to fit a catenary to the LIDAR data points that contain all local models currently identified over the span, with the estimate derived in the above equation being used as an initial estimate of the parameter values required for such numerical method.
Preferably, the step S2 of the step of organizing into segments further includes:
s21, processing the input point cloud data and calculating a statistical value: the statistics obtains the total number N of the laser points all ", the unit is one; and marking the clustering state of all laser points as 'not clustered', and meanwhile, calculating the average level 'space sampling interval D' of the power line laser point cloud s ", in m;
s22, least square linear fitting of the straight line segment: in an XOY plane, performing integral least square linear fitting by using horizontal coordinate information of all power line laser radar points of a certain grade, wherein a linear equation adopts a normal line type;
s23, solving length and scale factors: based on the linear equation obtained in the previous step, calculating a scale factor s corresponding to each laser point of the power line according to the linear segment model; maximum scale factor s 'obtained' max Minimum scale factor s' min The corresponding vertical projection points are respectively M 'and N', and the Euclidean distance between the points M 'and N' is recorded as the initial length L of the power line of the gear sp In the unit of m;
s24, partially overlapping data segmentation: the power line laser point clouds are sorted according to the scale factor s, and are divided into m sections according to the scale factor s, wherein the scale factor range of the power line laser point in the ith section is as follows:
i=0,1,2,...,m-1。
preferably, the segmentation method in step S2 is an optimized segmentation method, the laser point cloud data corresponding to the scale factor ranges of the i-th segment and the 1+1 have a certain overlap ratio, the overlap coefficient is defined as θ, where θ is in a range from 0 to 1, and the specific parameters are variable, so that the scale factor range of the i-th segment after optimization is:
preferably, the model fitting and the line segment labeling and synthesizing in step S3 specifically include:
s31, determining the values of the 7 parameters, and initializing the parameters as follows: number of iterations I i =0; number of points N satisfying power line model condition p =0; beginning label P of cluster lab =0;
S32, judging whether the iteration times exceed a set range; if I i >I MAX If so, ending the operation of the algorithm and jumping out of the loop; if I i ≤I MAX The algorithm continues to execute downward;
s33, randomly sampling partial superposition data segments, wherein each segment of m power line laser points defined by the formula (9) must be randomly extracted to form an 'unclustered' point, m points are cumulatively extracted, and the number N of points meeting the power line model condition at present p Recording as m;
s34, fitting an initial three-dimensional reconstruction model of the power line to be extracted and calculating the distance; if the following two conditions are met simultaneously, the next step is carried out, wherein the two conditions are as follows: (1) the distance D from the projection point of the m points on the XOY plane extracted in the step S33 to the fitted straight line i_L Are all less than D L_MAX (ii) a (2) Distance D from m points extracted in step S33 to the catenary model i_C Are all less than D C_MAX ;
S35, searching for a to-be-mentioned objectTaking laser points near the initial three-dimensional reconstruction model of the power line, calculating the distance from the laser points to a fitting straight line and a catenary in the initial three-dimensional reconstruction model for all the points which are not clustered, and if the two conditions in the step S34 are met simultaneously, counting the number N of points which meet the conditions of the power line model currently p Increasing by 1, i.e. N p =N p +1;
S36, calculating the proportion value of the 'non-clustered' points meeting the power line model condition and the condition meeting the condition, if the number of the points currently meeting the power line model condition is N p Is greater than or equal to N all ×R N_TH I.e. N p ≥N all ×R N_TH Entering the next step; otherwise, N p Number of iterations I, noted 0 i Self-increasing 1, i.e. I i =I i +1, return to step S32;
s37: refining the three-dimensional reconstruction model of the power line to be extracted, and after all algorithm execution flows of the step S3 are executed, utilizing the finally obtained N p And fitting the three-dimensional model again by using the information of the laser points meeting the power line model condition. At the same time, N p Reset to 0;
s38, calculating whether the length of the power line to be extracted meets a set condition threshold value or not, and calculating the length of a straight line segment in the current three-dimensional reconstruction model to be L' sp If L' sp Greater than or equal to R L_TH ×L sp I.e. L' sp ≥R L_TH ×L sp If the conditions are met, the next step is carried out; otherwise, the number of iterations I i Self-increasing 1, i.e. I i =I i +1, return to step S32;
s39, calculating the continuity degree of the power line points to be extracted, and uniformly dividing the straight line segment into G = L' sp /D s Taking an integer segment, wherein the spatial sampling interval D s In step S21, the number of points satisfying the power line model condition falling in each segment is calculated as N j Wherein j =1,2, …, G, the calculation formula of the degree of continuity F is
Wherein, the value range of F is 0 to 100, and the larger the value is, the better the continuity is;
s310, calculating to obtain an optimal three-dimensional reconstruction model of the power line to be extracted, and marking all points meeting the power line model conditions as clustered points;
s311, calculating and judging the number N of the current 'unclustered' points g In proportion to the total number of laser spots, if (N) g ÷N all )≥R N_TH Number of iterations I i Self-increasing 1, i.e. I i =I i +1, returning to the step S32; if (N) g ÷N all )<R N_TH The algorithm flow ends.
Compared with the prior art, the invention has the beneficial effects that: the power line laser point cloud data partial overlapping segmentation and clustering method has the following advantages:
(1) The method mainly adopts a random sampling technology based on partial coincidence data and power line three-dimensional reconstruction mathematical model constraint, and perfects the single-file power line laser point cloud clustering method. The power line state in a real three-dimensional scene is described by using a power line three-dimensional mathematical model combining a straight line segment and a catenary line segment, and the accuracy of partially-overlapped laser point cloud obtained by random sampling is restrained.
(2) Experiments show that the clustering method can realize the correct segmentation of power line point clouds with any number of power lines, various power line types (single wires, split wires, lightning conductors and the like), various power line spatial configurations (triangular, horizontal, vertical, mixed arrangement, staggered arrangement and the like), any length, any tower height difference and irregular fracture and gross error, and improves the accuracy, the automation degree, the robustness and the universality of single power line laser radar point identification.
Drawings
The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
Fig. 1 is a schematic diagram of a three-dimensional reconstruction model of a power line.
Figure 2 is a schematic diagram of a straight line segment model.
FIG. 3 is an overall flow diagram of the power line laser point cloud data partial superposition segmentation and clustering method of the present invention.
FIG. 4 is a diagram of the results of example 1 of the present invention, in which: the method comprises the steps of (a) power line LiDAR point cloud, (b) straight line segment fitting and (c) clustering results.
FIG. 5 is a diagram illustrating the results of example 2 of the present invention.
FIG. 6 is a diagram of the results of embodiment 3 of the present invention, in which: the method comprises the following steps of (a) power line LiDAR point cloud and (b) clustering results.
FIG. 7 shows the results of example 4 of the present invention.
FIG. 8 is a graph showing the results of example 5 of the present invention.
Detailed Description
The following embodiments of the present invention are provided by way of specific examples, and other advantages and effects of the present invention will be readily apparent to those skilled in the art from the disclosure herein. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention.
The invention provides a partial overlapping segmentation and clustering method for power line laser point cloud data, wherein one of key links of three-dimensional reconstruction of a power line is a mathematical expression for determining a three-dimensional reconstruction model of a single-file single power line. Three-dimensional power line models typically contain two parts: (1) XOY horizontal in-plane projection model; (2) a model is projected in a certain vertical projection plane.
The three-dimensional reconstruction model of the single-gear single power line adopted by the invention also comprises two mutually related parts, as shown in figure 1.
The first part is a straight line segment which is generated by least square fitting of projection points of the power line laser point cloud on an XOY plane;
the second part is a catenary line segment which is positioned on the vertical plane of the straight line segment and is generated by the information of the power line laser point cloud.
In addition, the straight line segment and the catenary line segment are not only coplanar, but also two end points of the straight line segment and two end points of the catenary line segment have a vertical projection relationship.
It should be noted that the three-dimensional reconstruction model provided by the invention is suitable for three-dimensional modeling of any type of power line (including single conductor, lightning conductor, any split (bundle) of split conductor and the whole split conductor).
1. Straight line segment model
A straight-line segment model in the XOY plane includes two parts: a straight line and two end points.
The linear model is generated by least square fitting of a power line laser point on an XOY plane projection point; the two end points are determined by the straight line model and the two extreme scale factor points.
Step 1, determining a straight line model, wherein the straight line model adopts a normal line type shown in formula (1).
l=x·cosα+ysinα (1)
In the formula: α is an inclination angle of a straight line passing through the origin to the straight line, and the straight line is the vertical line, as shown in fig. 1; l is the length of the straight line segment, and the two end points of the straight line are M and N, respectively, as shown in FIG. 1.
Meanwhile, the intersection point of the vertical line segment and the fitting straight line is set as P (x) fp ,y fp ) As shown in fig. 1, i.e. the foot P, with a scaling factor s =0.
And step 2, determining the coordinates of the two endpoints M and N. Calculating a scale factor of each power line laser point, as shown in formula (2) or formula (3), the specific process is as follows:
let the horizontal coordinate of any power line laser point be Q (x) 0 ,y 0 ),Q(x 0 ,y 0 ) The coordinates of the projected points to the fitted straight line are Q '(x' 0 ,y′ 0 ) As shown in fig. 2, the scale factor s of the vertical projection point is calculated according to equation (2) or equation (3).
When fabs (sin α) ≧ fabs (cos α):
when fabs (sin α) < fabs (cos α):
the fabs function is an absolute function. After the scale factor of each power line laser point is obtained, the maximum scale factor S can be obtained max And a minimum scale factor S min . The two corresponding vertical projection points are M, N, respectively, and M, N is the two end points of the required straight-line segment model.
2. Catenary section model
The straight line (segment) passing through the 1 part is taken as a vertical plane perpendicular to the XOY plane, and a catenary line segment model in the vertical plane also comprises two parts, namely a catenary line and two end points.
The catenary model is generated by fitting derived data of power line laser point clouds, and each point in the derived data comprises two parts:
the z value of each laser point and a corresponding scale factor s;
and two end points are determined by projecting the two end points of the straight-line segment model obtained in the step 2 to the catenary model.
Catenary model:
in the formula: k. h is 1 And h 2 Is the coefficient to be found. In addition, experiments find that Q or more laser point cloud data can be fitted to the parameters of the catenary model, wherein Q is more than or equal to 6.
This catenary will be embedded in a plane P spanned by the vector and the Z-axis. To calculate the parameters of the catenary with respect to P, we used the following theorem.
Theorem 1: for a catenary with a parameter, let be any point on the catenary, let be a tangent vector.
Using theorem 1, a right triangle can be constructed for each tangent vector, where one side has a length of k 1 A and theta 1 The following equation:
k=(y 1 -h 2 )sin(θ 1 ) (5)
likewise, there are:
k=(y 2 -h 2 )sin(θ 2 ) (6)
equating (4) and (5), resulting in a value:
bringing it into (4), one can calculate:
in modeling the catenary, the algorithm estimates the parameter h using the first and last local models identified over the span 1 、h 2 K, then refine the estimate using numerical methods to fit a catenary to the LIDAR data points that contain all local models currently identified on the span. The estimated value obtained in the above equation is used as an initial estimate of the parameter value required for such numerical methods.
The overhead transmission line takes 'gear' as a basic composition unit, only focuses on the clustering of single-gear power line laser point clouds, and the three-dimensional coordinates of the point clouds are original information without any additional processing. The clustering method of the invention comprises two important parts:
(1) data preprocessing and partially overlapping data segmentation;
(2) model fitting and line segment labeling and synthesis.
The overall technical flow is shown in fig. 3.
Step 1: in order to use the layered random sampling, the preprocessing and the segmentation organization of the point cloud data are required to be carried out, and the point cloud segmentation organization reflects the layering of the sampling; the stage of the segmented organization mainly comprises the following 4 steps:
step 1.1, processing the input point cloud data and calculating a statistical value. Obtaining the total number N of laser points through statistics all "(unit: number); and the cluster status of all laser spots is marked as "not clustered". At the same time, the average level 'space sampling interval D' of the power line laser point cloud is calculated s "(unit: m).
Step 1.2, least squares linear fitting of the straight line segments. And in the XOY plane, performing integral least square linear fitting by using the horizontal coordinate information of all power line laser radar points of a certain level, wherein a linear equation also adopts a normal line type.
And step 1.3, solving length and scale factors. And (3) calculating a scale factor s corresponding to each laser point of the power line according to the linear equation obtained in the step 1 and the formula linear segment model. Maximum scale factor s 'obtained' max Minimum scale factor s' min The corresponding vertical projection points are M 'and N', respectively. The Euclidean distance between points M 'and N' is recorded as the initial length L of the power line of the gear sp (unit: m).
And 1.4, partially overlapping data segments. Sequencing the power line laser point clouds according to the scale factor s and sequencing the power line laser point clouds according to the scale factor s s The size of (2) divides the power line laser point cloud into m sections. Wherein, the range of the scale factor of the power line laser spot in the i (i =0,1,2., m-1) section is as follows:
furthermore, the invention provides an optimized segmentation method, wherein laser point cloud data corresponding to the scale factor ranges of the i-th section and the 1+1 have certain coincidence degree, the coincidence coefficient is defined as theta, the range of the theta is 0-1, and specific parameters are variable. Thus, the scale factor range of the ith segment after optimization is:
step 2: model fitting and line segment labeling and synthesis are an iterative process, random sampling is adopted in each iteration to extract partial points from unlabeled laser points, whether fitting models of the points simultaneously accord with straight line segment models and catenary line segment models proposed by the part 1 is checked, and meanwhile, the negative influence of rough difference points in laser point cloud is reduced in the iterative process.
In the step of segment labeling and synthesis, 7 key parameters are set, including: a coincidence coefficient theta; maximum number of iterations I MAX (ii) a The number of non-clustered points accounts for the total number N of laser points all Minimum ratio of R N_TH (ii) a Maximum distance D from projection point of laser point on XOY plane to fitting straight line L_MAx (ii) a The length of the fitting straight line segment of the power line to be extracted accounts for the length L of the power line sp Minimum ratio of R L_TH (ii) a Maximum distance D of laser spot to fitted catenary C_MAX (ii) a The breaking length of the power line to be extracted accounts for the length L of the power line sp Minimum ratio of R B_TH 。
Step 2 (detail):
step 2.1, determine the values of the 7 parameters as described above. In addition, the initialization parameters are: number of iterations I i =0; number of points N satisfying power line model condition p =0; beginning label P of cluster lab =0。
Step 2.2, judging whether the iteration times exceed a set range; if I i >I MAX If so, ending the operation of the algorithm and jumping out of the loop; if I i ≤I MAX The algorithm continues to execute downward.
And 2.3, randomly sampling the partially overlapped data segments. From m segments of power line laser points defined by the formula (9), each segment must randomly extract one 'non-clustered' point, and cumulatively extract m points. Meanwhile, the number of points N currently satisfying the power line model condition p Is denoted as m.
And 2.4, fitting an initial three-dimensional reconstruction model of the power line to be extracted and calculating the distance. If the following two conditions are satisfied simultaneouslyAnd entering the next step, wherein two conditions are as follows: (1) the distance D from the projection point of the m points on the XOY plane extracted in the step 2.3 to the fitted straight line i_L Are all less than D L_MAX (ii) a (2) Distance D from m points extracted in step 2.3 to the catenary model i_C Are all less than D C_MAx 。
And 2.5, searching a laser point near the initial three-dimensional reconstruction model of the power line to be extracted. Calculating the distance from all 'unclustered' points to a fitting straight line and a catenary in the initial three-dimensional reconstruction model, and if the two conditions in the step 2.4 are met simultaneously, counting the number N of points meeting the power line model condition currently p Increasing by 1, i.e. N p =N p +1。
And 2.6, calculating the proportion value of the non-clustered points meeting the power line model condition and meeting the condition. Number of points N if the power line model condition is currently satisfied p Greater than or equal to N all ×R N_TH I.e. N p ≥N all ×R N_TH Entering the next step; otherwise, N p Number of iterations I, noted 0 i Self-increasing 1, i.e. I i =I i +1, return to step 2.2.
Step 2.7: and refining the three-dimensional reconstruction model of the power line to be extracted. After all algorithm execution flows of the step 2 are completed, the finally obtained N is utilized p And fitting the three-dimensional model again by using the information of the laser points meeting the power line model condition. At the same time, N p Reset to 0.
And 2.8, calculating whether the length of the power line to be extracted meets a set condition threshold value. Calculating the length of a straight line segment in the current three-dimensional reconstruction model to be L' sp . If L' sp Greater than or equal to R L_TH ×L sp I.e. L' sp ≥R L_TH ×L sp If the conditions are met, the next step is carried out; otherwise, the number of iterations I i Self-increasing 1, i.e. I i =I i +1, return to step 2.2.
And 2.9, calculating the continuity degree of the power line points to be extracted. The straight line segment was evenly divided into G = L' sp /D s (integer) segment, which is emptyInter-sampling interval D s In step 1.1, the number of points that fall within each segment and satisfy the power line model conditions is calculated as N j Where j =1,2, …, G. The degree of continuity F is calculated by
Wherein, the value range of F is 0 to 100, and the larger the numerical value is, the better the continuity is.
And 2.10, calculating to obtain an optimal three-dimensional reconstruction model of the power line to be extracted, and marking all points meeting the power line model conditions as clustered points.
Step 2.11, calculating and judging the number N of the current 'uncleaved' points g The proportion of the total number of laser points. If (N) g ÷N all )≥R N_TH Number of iterations I i Self-increasing 1, i.e. I i =I i +1, returning to the step 2.2; if (N) g ÷N all )<R N_TH Then the algorithm flow ends.
The clustering method for the single-gear power line laser point cloud of the overhead power transmission line is realized. And totally adopting the laser point cloud data of 4-gear overhead power transmission lines to carry out 5 clustering experiments. The basic case of experimental data is as follows:
in embodiment 1, the power line laser point cloud shown in fig. 4 includes 6 power lines, the types of the power lines are single wires, and a horizontal and vertical mixed arrangement structure is adopted, so that the power line points have an irregular fracture phenomenon and a very small number of rough difference points. The length of the gear power line is about 411.0m.
In embodiment 2, the power line laser point cloud shown in fig. 5 includes 6 power lines, the types of the power lines are all single wires, and a staggered arrangement structure is adopted, so that a large number of irregular fractures exist in the point cloud, but coarse difference points do not exist. The length of the gear power line is about 90.5m.
In example 3, the power line laser point cloud shown in fig. 6 includes 14 power lines, the types of the power lines are all single wires, a horizontal and vertical mixed arrangement structure is adopted, irregular breakage occurs in the point cloud, and a small number of coarse difference points exist. The length of the gear power line is about 170.0m.
Example 4, the power line laser point cloud shown in fig. 7 includes 2 lightning conductors and 2 split conductors (each having 4 splits), and adopts a horizontal and vertical mixed arrangement structure, so that the point cloud has an irregular fracture phenomenon and a large number of rough difference points. The length of the gear power line is about 390.0. In addition, it can be seen that, in the 4 embodiments, except for embodiment 3, there is a significant difference in the tower heights at the two ends of each power line in other data.
Example 5 as shown in fig. 8, a clustering experiment for extracting 4 power lines (except 2 lightning conductor lines, each split of 2 split conductor lines is treated as one power line) was performed on the basis of the first three experimental data and named "experiment five".
For the quantitative evaluation of the clustering effect, 3 indexes of accuracy, completeness and quality are adopted. The basic unit of the 3 index statistics is an 'object', and the rough difference point is not considered in the statistics. By using the clustering method of the invention, the accuracy and the integrity of 5 experiments are 100%, and the quality is 1. This shows that the clustering method of the present invention obtains a completely correct power line extraction result without error.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical spirit of the present invention be covered by the claims of the present invention.
Claims (5)
1. A power line laser point cloud data partial overlapping segmentation and clustering method is characterized by comprising the following steps:
s1, determining a mathematical expression of a three-dimensional reconstruction model of a single-gear single power line; the adopted three-dimensional reconstruction model of the single-gear single power line comprises two parts: the first part is a straight line segment model, and the straight line segment is generated by least square fitting of projection points of the power line laser point cloud on an XOY plane; the second part is a catenary line segment which is positioned on the vertical plane of the straight-through line segment and is generated by the information of the power line laser point cloud; the straight line segment and the catenary line segment are coplanar, and two end points of the straight line segment and two end points of the catenary line segment have a vertical projection relation;
s2, data preprocessing and partial superposition data segmentation; carrying out preprocessing and sectional organization of power line laser point cloud data; the method comprises the following steps: processing input point cloud data and calculating a statistical value; least squares linear fitting of the straight line segments; finding length and scale factors; and partially overlapping the data segments;
s3, model fitting and line segment labeling and synthesizing; model fitting and line segment labeling and synthesis are an iterative process, random sampling is adopted in each iteration to extract partial points from unlabeled laser points, whether the fitting models of the points simultaneously conform to the straight line segment model and the catenary line segment model proposed in S1 is checked, and meanwhile, the negative influence of rough difference points in laser point cloud is reduced in the iterative process;
in the step of segment labeling and synthesis, 7 key parameters are set, including: a coincidence coefficient theta; maximum number of iterations I MAX (ii) a The number of non-clustered points accounts for the total number N of laser points all Minimum ratio of R N_TH (ii) a Maximum distance D from projection point of laser point on XOY plane to fitting straight line L_MAX (ii) a The length of the fitting straight line segment of the power line to be extracted accounts for the length L of the power line sp Minimum ratio of R L_TH (ii) a Maximum distance D of laser spot to fitted catenary C_MAX (ii) a The breaking length of the power line to be extracted accounts for the length L of the power line sp Minimum ratio of R B_TH 。
2. The method for partially overlapping, segmenting and clustering power line laser point cloud data according to claim 1, wherein the step S1 further comprises:
s11, establishing a straight-line segment model: the straight-line segment model in the XOY plane includes two parts: a straight line model and two end points; the two end points are determined by the straight line model and the two extreme value scale factor points;
s111, determining a straight line model, wherein the straight line model adopts a normal line type shown in formula (1):
l=x·cosα+ysinα (1)
in the formula: alpha is a vertical line segment from the original point to the straight line, the inclination angle of the straight line where the vertical line segment is located, l is the length of the straight line segment, and two end points of the straight line segment are respectively M and N; meanwhile, the intersection point of the vertical line segment and the fitting straight line is set as P (x) fp ,y fp ) I.e. drop foot P, with scale factor s =0;
s112, determining coordinates of the two end points M and N, and solving a scale factor of each power line laser point, wherein the specific process is as follows as shown in a formula (2) or a formula (3):
let the horizontal coordinate of any power line laser point be Q (x) 0 ,y 0 ),Q(x 0 ,y 0 ) The coordinates of the projected points to the fitted straight line are Q '(x' 0 ,y′ 0 ) Calculating the scale factor s of the vertical projection point according to the formula (2) or the formula (3):
when) abs (sin α) ≧ fabs (cos α):
when fabs (sin α) < fabs (cos α):
the fabs function is a function for solving an absolute value, and after a scale factor of each power line laser point is solved, the maximum scale factor S can be obtained max And a minimum scale factor S min If the two corresponding vertical projection points are M, N respectively, M, N is two end points of the straight-line segment model;
s12, establishing a catenary line segment model: making a vertical plane perpendicular to the XOY plane by the straight line segment in the step S11, wherein a catenary line segment model in the vertical plane comprises two parts, namely a catenary and two end points; the catenary model is generated by fitting derived data of power line laser point clouds, and each point in the derived data comprises two parts: the z value of each laser point and a corresponding scale factor s; two end points are determined by projecting the two end points of the straight-line segment model obtained in the step S11 to the catenary model; catenary model:
in the formula: k. h is 1 And h 2 For the coefficient to be solved, Q and more than Q laser point cloud data can be fitted to obtain the parameters of the catenary model, wherein Q is more than or equal to 6;
calculate the catenary with parameters using theorem 1: for the catenary with parameters, setting any point on the catenary as a tangent vector, wherein the minimum distance to the point is equal to the catenary parameters;
using theorem 1, a right triangle is constructed for each tangent vector, where one side has a length of k 1 A and E 1 The following equation:
k=(y 1 -h 2 )sin(E 1 ) (5)
likewise, there are:
k=(y 2 -h 2 )sin(θ 2 ) (6)
equating (4) and (5), resulting in a value:
bringing it into (4), one can calculate:
modeling catenaryThe algorithm estimates the parameter h using the first and last local models identified over the span 1 、h 2 K, then refine the estimate using a numerical method to fit a catenary to the LIDAR data points that contain all local models currently identified over the span, the estimated values derived in the above equation serve as initial estimates of the parameter values required for such numerical method.
3. The method for partially overlapping, segmenting and clustering power line laser point cloud data according to claim 2, wherein the segmentation organization stage of the step S2 further comprises:
s21, processing the input point cloud data and calculating a statistical value: the statistics obtains the total number N of the laser points all ", the unit is one; and marking the clustering state of all laser points as 'not clustered', and meanwhile, calculating the average level 'space sampling interval D' of the power line laser point cloud s ", in m;
s22, least square linear fitting of the straight line segment: in an XOY plane, performing integral least square linear fitting by using horizontal coordinate information of all power line laser radar points of a certain grade, wherein a linear equation adopts a normal line type;
s23, solving length and scale factors: based on the linear equation obtained in the previous step, calculating a scale factor s corresponding to each laser point of the power line according to the linear segment model; maximum scale factor s 'obtained' max Minimum scale factor s' min The corresponding vertical projection points are respectively M 'and N', and the Euclidean distance between the points M 'and N' is recorded as the initial length L of the power line of the gear sp In the unit of m;
s24, partially overlapping data segmentation: the power line laser point clouds are sorted according to the scale factor s, and are divided into m sections according to the scale factor s, wherein the scale factor range of the power line laser point in the ith section is as follows:
i=0,1,2,…,m-1。
4. the method for partially overlapping, segmenting and clustering power line laser point cloud data according to claim 3, wherein the segmentation method of step S2 is an optimized segmentation method, the laser point cloud data corresponding to the scale factor ranges of the i-th segment and the 1+1 segment have a certain overlap ratio, and the overlap coefficient is defined as θ, wherein the range of θ is 0-1, and the specific parameters are variable, so that the scale factor range of the i-th segment after optimization is:
5. the method for partially overlapping, segmenting and clustering power line laser point cloud data according to claim 4, wherein the model fitting and segment labeling and synthesizing of the step S3 specifically comprises:
s31, determining the values of the 7 parameters, and initializing the parameters as follows: number of iterations I i =0; number of points N satisfying power line model condition p =0; beginning label P of cluster lab =0;
S32, judging whether the iteration times exceed a set range; if I i >I MAX If so, ending the operation of the algorithm and jumping out of the loop; if I i ≤I MAX The algorithm continues to execute downward;
s33, randomly sampling partial superposition data segments, wherein each segment of m power line laser points defined by the formula (9) must be randomly extracted to form an 'unclustered' point, m points are cumulatively extracted, and the number N of points meeting the power line model condition at present p Recording as m;
s34, fitting an initial three-dimensional reconstruction model of the power line to be extracted and calculating the distance; if the following two conditions are met simultaneously, the next step is carried out, wherein the two conditions are as follows: (1) extracted in step S33The distance D from the projection point of the m points on the XOY plane to the fitted straight line i_L Are all less than D L_MAX (ii) a (2) Distance D from m points extracted in step S33 to the catenary model i_C Are all less than D C_MAX ;
S35, searching laser points near the initial three-dimensional reconstruction model of the power line to be extracted, calculating the distance from the laser points to a fitting straight line and a catenary in the initial three-dimensional reconstruction model for all 'unclustered' points, and if the two conditions in the step S34 are met simultaneously, counting the number N of the points meeting the conditions of the power line model currently p Increasing by 1, i.e. N p =N p +1;
S36, calculating the proportion value of the 'non-clustered' points meeting the power line model condition and the condition meeting the condition, if the number of the points currently meeting the power line model condition is N p Greater than or equal to N all ×R N_TH I.e. N p ≥N all ×R N_TH Entering the next step; otherwise, N p Number of iterations I, noted 0 i Self-increasing 1, i.e. I i =I i +1, return to step S32;
s37: refining the three-dimensional reconstruction model of the power line to be extracted, and after all algorithm execution flows of the step S3 are executed, utilizing the finally obtained N p And fitting the three-dimensional model again by using the information of the laser points meeting the power line model condition. At the same time, N p Reset to 0;
s38, calculating whether the length of the power line to be extracted meets a set condition threshold value or not, and calculating the length of a straight line segment in the current three-dimensional reconstruction model to be L' sp If L' sp Greater than or equal to R L_TH ×L sp I.e. L' sp ≥R L_TH ×L sp If the conditions are met, entering the next step; otherwise, the number of iterations I i Self-increasing 1, i.e. I i =I i +1, return to step S32;
s39, calculating the continuity degree of the power line points to be extracted, and uniformly dividing the straight line segment into G = L' sp /D s Taking an integer segment, wherein the spatial sampling interval D s In step S21, the strip satisfying the power line pattern falling in each segment is calculatedThe number of points of the member being N j Wherein j =1,2, …, G, the calculation formula of the degree of continuity F is
Wherein, the value range of F is 0 to 100, and the larger the value is, the better the continuity is;
s310, calculating to obtain an optimal three-dimensional reconstruction model of the power line to be extracted, and marking all points meeting the power line model conditions as clustered points;
s311, calculating and judging the number N of the current 'unclustered' points g In proportion to the total number of laser spots, if (N) g ÷N all )≥R N_TH Number of iterations I i Self-increasing 1, i.e. I i =I 3 +1, return to step S32; if (N) g ÷N all )<R N_TH Then the algorithm flow ends.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211324256.9A CN115641406A (en) | 2022-10-27 | 2022-10-27 | Partial overlapping segmentation and clustering method for power line laser point cloud data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211324256.9A CN115641406A (en) | 2022-10-27 | 2022-10-27 | Partial overlapping segmentation and clustering method for power line laser point cloud data |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115641406A true CN115641406A (en) | 2023-01-24 |
Family
ID=84947039
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211324256.9A Pending CN115641406A (en) | 2022-10-27 | 2022-10-27 | Partial overlapping segmentation and clustering method for power line laser point cloud data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115641406A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117095318A (en) * | 2023-10-20 | 2023-11-21 | 山东科技大学 | High-voltage transmission line extraction method combining transmission trend and tower position |
CN118134985A (en) * | 2024-05-08 | 2024-06-04 | 中国空气动力研究与发展中心低速空气动力研究所 | Complex ice-shaped three-dimensional reconstruction method and medium based on dense time sequence |
-
2022
- 2022-10-27 CN CN202211324256.9A patent/CN115641406A/en active Pending
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117095318A (en) * | 2023-10-20 | 2023-11-21 | 山东科技大学 | High-voltage transmission line extraction method combining transmission trend and tower position |
CN117095318B (en) * | 2023-10-20 | 2024-03-19 | 山东科技大学 | High-voltage transmission line extraction method combining transmission trend and tower position |
CN118134985A (en) * | 2024-05-08 | 2024-06-04 | 中国空气动力研究与发展中心低速空气动力研究所 | Complex ice-shaped three-dimensional reconstruction method and medium based on dense time sequence |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111537990B (en) | Lane calibration method and device and electronic equipment | |
CN115641406A (en) | Partial overlapping segmentation and clustering method for power line laser point cloud data | |
CN113487730B (en) | Urban three-dimensional automatic modeling method based on laser radar point cloud data | |
CN106157361A (en) | A kind of multiple fission conductor full-automatic three-dimensional method for reconstructing based on LiDAR point cloud | |
CN111582380B (en) | Ship track density clustering method and device based on space-time characteristics | |
WO2023103415A1 (en) | Component modeling and parameterization method and system, electronic device, and storage medium | |
CN107610223A (en) | Power tower three-dimensional rebuilding method based on LiDAR point cloud | |
WO2021036550A1 (en) | Visual task-based method and system for evaluating compression quality of point cloud data | |
CN105488847B (en) | The single grade of single power wire three-dimensional rebuilding method based on airborne laser radar point cloud | |
CN103034863A (en) | Remote-sensing image road acquisition method combined with kernel Fisher and multi-scale extraction | |
CN107958486A (en) | A kind of generation method and device of conducting wire vector model | |
CN109102538B (en) | Method for extracting land characteristic points and land characteristic lines by using contour line data and weather map analysis method | |
CN105069395B (en) | Roadmarking automatic identifying method based on Three Dimensional Ground laser scanner technique | |
CN115757369A (en) | Automatic inspection method and system for laser point cloud data | |
CN116704333B (en) | Single tree detection method based on laser point cloud data | |
CN115619972A (en) | Road marking abrasion identification and evaluation method based on point cloud data | |
CN112085843B (en) | Tunnel class target feature real-time extraction and measurement method and device | |
CN113592324A (en) | Cable terminal tower live-line work risk assessment method based on hierarchical analysis | |
CN110647647B (en) | Closed graph similarity searching method based on time sequence complexity difference | |
CN109271466A (en) | A kind of weather data analysis method based on hierarchical clustering Yu K mean algorithm | |
Zhao et al. | A Review of Point Cloud Segmentation of Architectural Cultural Heritage | |
CN115825984A (en) | Power transmission line inspection system and inspection method based on laser point cloud | |
Shen et al. | An automatic framework for pylon detection by a hierarchical coarse-to-fine segmentation of powerline corridors from UAV LiDAR point clouds | |
CN110909792A (en) | Clustering analysis method based on improved K-means algorithm and new clustering effectiveness index | |
CN114419877A (en) | Vehicle track prediction data processing method and device based on road characteristics |
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 |