阅读说明：本技术 一种激光LiDAR电力线点云自动补点方法 (Automatic point cloud point supplementing method for laser LiDAR power line ) 是由 虢韬 徐梁刚 杨恒 杨刘贵 时磊 杨渊 刘勇 龙新 赵健 王迪 于 2020-07-06 设计创作，主要内容包括：本发明公开了一种激光LiDAR电力线点云自动补点方法,它包括：步骤1、电力线点云分档以及坐标转换,得到电力线点云集合；步骤2、单档电力线点云平面分割以及网格化；步骤3、补点计算,对平面分割子集内的电力线进行补点；解决激光LiDAR电力线点云密度低、点云缺失严重的技术问题,降低人工数据处理工作量,提高补点的点位精度,为后续电力线重建、危险点检测等功能应用提供数据支持。(The invention discloses a point cloud automatic point supplementing method for a laser LiDAR power line, which comprises the following steps: step 1, power line point cloud grading and coordinate conversion are carried out to obtain a power line point cloud set; step 2, single-gear power line point cloud plane segmentation and meshing; step 3, point compensation calculation, namely performing point compensation on the power lines in the plane segmentation subsets; the technical problems of low point cloud density and serious point cloud loss of the laser LiDAR power line are solved, the workload of manual data processing is reduced, the point location precision of point compensation is improved, and data support is provided for functional applications such as subsequent power line reconstruction and dangerous point detection.)

1. A laser LiDAR power line point cloud automatic point compensating method comprises the following steps:

step 1, power line point cloud grading and coordinate conversion are carried out to obtain a power line point cloud set;

step 2, single-gear power line point cloud plane segmentation and meshing;

and 3, performing point supplementing calculation, namely performing point supplementing on the power lines in the plane segmentation subsets.

2. The laser LiDAR power line point cloud automatic point-filling method of claim 1, wherein: the method for grading the power line point cloud and converting the coordinates comprises the following steps: obtaining the center coordinates of the towers, sequentially constructing rectangles which take the connecting lines of the towers as the center and are expanded by 25 meters from left to right, and taking points in the rectangles as the same gear to realize power line point cloud grading; and meanwhile, calculating the vertical distance offset d from the point to the pole tower connecting line, and calculating the cumulative distance l of the projection distance of the point and the small-size pole tower connecting line on the pole tower connecting line to realize coordinate conversion.

3. The laser LiDAR power line point cloud automatic point-filling method of claim 2, wherein: the specific method for realizing coordinate conversion comprises the following steps of calculating the vertical distance offset distance d from a point to a pole tower connecting line, and calculating the cumulative distance l of the projection distance of the point and a small-size pole tower connecting line on the pole tower connecting line:

let the coordinate of the small tower A of any gear be (x)_A,y_A) And the B coordinate of the large tower is (x)_B,y_B) The coordinate of any point P on the power line point cloud is (x)_P,y_P) Constructing a vectorAndwherein the content of the first and second substances,

X_AB＝x_B-x_A，Y_AB＝y_B-y_A

X_AP＝x_P-x_A，Y_AP＝y_P-y_A

calculate the area S of Δ ABP_ABP＝(X_AB×Y_AP-X_AP×Y_AB) And/2, obtaining the offset distance:

accumulated distance

Wherein the offset distance d_PThe right side of a tower connecting line is negative, and the left side of the tower connecting line is positive; if the condition is satisfied

4. The laser LiDAR power line point cloud automatic point-filling method of claim 1, wherein: the single-gear power line point cloud plane segmentation method in the step 2 comprises the following steps: the method comprises the steps of realizing segmentation calculation of power lines with non-coincident plane positions by detecting offset jump, designing a one-dimensional grid parallel to a connecting line of center points of a tower, and meshing a segmented point set.

5. The laser LiDAR power line point cloud automatic point-filling method of claim 4, wherein: the method for realizing the segmentation calculation of the power line with non-coincident plane positions by detecting the offset jump, designing a one-dimensional grid parallel to a connecting line of the center points of the tower and meshing the segmented point set comprises the following steps:

step 2.1, sorting any one-gear power Line point cloud set converted in the step 1 according to an ascending order of offset distance to obtain an ascending order set Line { P }₁,P₂,…,P_n}; according to d_i-d_i-1When the value is more than 2, n is more than or equal to i and is used as a constraint condition, searching offset catastrophe points, wherein n is the number of ascending set points; suppose t mutation points, P respectively, are retrieved_m1,P_m2,…,P_mtThen the subset is divided into

[line₁{P₁,P₂,…P_m1-1},line₂{P_m1,P_m1+1,…P_m2-1},…,line_t+1{P_mt,P_mt+1,…P_n}]。

Step 2.2, setting the accumulative distance step length to be 1m and the number of grids to beThe one-dimensional grid of (1) dividing any point P in the point cloud subset into [ l ]_P]And putting the grid indexes into corresponding grids, and arranging the points in the sub-grids according to the ascending order of the elevation.

6. The laser LiDAR power line point cloud auto-spotting method of claim 5, wherein: step 3, the supplementary point calculation method comprises the following steps:

step 3.1, setting the ith sub-Grid in the one-dimensional Grid to be Grid in step 2.2_i{P₁,P₂,…,P_n}, according to l_t-l_t-1If the number of the accumulation distance mutation points is more than 4, and t is less than or equal to N, the number of the grid power line elements L is greater than or equal to N_iN + 1; traversing the entire grid, the number of power lines currently partitioning the subset is Max ({ L)₁,L₂,…,L_n})；

Step 3.2, in the initial point collecting stage, starting from the small grid with the accumulation distance of 0, traversing the grid backwards, and when the number L of power lines of the grid is_i＝Max({L₁,L₂,…,L_nAnd) adding the lowest point of the grid into an initial sampling point array; when the number of the initial sampling point groups is equal to 5, stopping traversing, and performing polynomial fitting to obtain an initial local power line equation;

3.3, after obtaining an initial local power line equation, carrying out extrapolation calculation in front and back directions of the accumulated distance;

step 3.4, recording the local power line equation of each section and the corresponding sampling point array, and setting the first point of the sampling point array as P_starThe last point is P_endUsing local power line equations, the accumulation distance is given by l_starStarting with_endFor termination, carrying out interpolation calculation at an interval of 0.1m to obtain a local power line compensation point; putting all interpolation results of the local power line equations to the same point set, and rotating the interpolation results to an original point cloud coordinate system through coordinates to obtain a point supplementing result of the lowest power line under the plane segmentation subset;

3.5, establishing a kd-tree by using a local power line point complement set, and removing the laser points on the power lines extracted by the grid at this time through proximity search; repeating the steps 3.1-3.4, and sequentially supplementing points to the power lines in the plane segmentation subsets.

7. The laser LiDAR power line point cloud automatic point-filling method of claim 6, wherein: the polynomial fitting method adopts a least square theory, takes the accumulated distance l as an independent variable, performs linear fitting with the offset distance d, and performs 3 times of term fitting with the elevation h, and a specific equation is as follows:

in the formula, a₁,a₂And b₁,b₂,b₃,b₄Respectively, linear equation and polynomial equation parameters.

8. The laser LiDAR power line point cloud automatic point-filling method of claim 6, wherein: step 3.3 the extrapolation calculation method comprises: taking the rear as an example, continuously traversing the grid backward from the grid where the last point of the initial sampling point group is located, and setting the lowest point of the grid as P_low(d_low,l_low,h_low) To do so byJudging whether the lowest point of the grid is a new sampling point or not for the constraint condition, wherein

The technical field is as follows:

the invention belongs to the automatic processing technology of laser LiDAR point clouds of a power transmission line, and relates to an automatic point cloud point compensating method of a laser LiDAR power line.

Background art:

the airborne laser LiDAR technology is a technical revolution again after the GNSS positioning technology in the measurement field, and develops an observation mode from traditional point-to-point observation to point-to-surface and line-to-surface observation. The traditional manual line patrol labor intensity is high, the working condition is hard, the efficiency is low, the period of repeated line patrol is long, the accuracy rate of data patrol is not high, and the airborne laser LiDAR technology is widely applied to the electric line patrol work of the overhead transmission line due to the advantages of high measurement precision, high working efficiency and less manpower investment.

At present, the processing of power line inspection laser LiDAR point cloud data is roughly divided into two ideas, one is that a kd-tree or an oc-tree is constructed through dense power line point clouds to perform proximity search so as to solve the distance, and therefore dangerous point detection and wire inter-phase distance measurement are achieved; and secondly, reconstructing the power line discrete point cloud into a mathematical curve equation expression by a power line reconstruction technology, and realizing dangerous point detection and inter-phase conductor measurement by calculating the distance from a point to the curve equation. Above two kinds of thinking all need to regard as the data basis with intensive power line point cloud, when power line point cloud density is low, the point cloud disappearance is serious, can lead to the calculated result to take place the deviation: in the first method, due to the lack of the closest point in the kd-tree or the oc-tree, the distance calculation result is larger, and the accuracy of the dangerous point detection and the wire inter-phase distance measurement result is further influenced; the second method mostly needs to use a clustering method to segment the power line at present, the point cloud density is low, the point cloud loss is serious, segmentation errors are directly caused, and a fitted mathematical equation is not consistent with the actual position of the power line.

However, due to the influences of factors such as high flying speed of the carrying platform, low collection density of laser LiDAR equipment, low voltage level of a power transmission line, wire shielding by a ground wire, poor external observation conditions and the like, the situations of low density of power line point cloud and serious point cloud loss frequently occur, and the serious influence is caused on the functional applications such as subsequent power line reconstruction, dangerous point detection, wire inter-phase distance measurement and the like. At present, power line point compensation is basically carried out in a mode of fitting interpolation through manual point collection, operation is complex, workload is large, and situations of nonuniform manual point selection and point selection errors exist.

The invention content is as follows:

the technical problems to be solved by the invention are as follows: the automatic point cloud point supplementing method for the laser LiDAR power line aims to solve the technical problems of low point cloud density and serious point cloud loss of the laser LiDAR power line, reduces the workload of manual data processing, improves the point location precision of point supplement, and provides data support for functional application of subsequent power line reconstruction, dangerous point detection and the like.

The technical scheme of the invention is as follows:

a laser LiDAR power line point cloud automatic point compensating method comprises the following steps:

step 1, power line point cloud grading and coordinate conversion are carried out, and a power line point cloud set of each grade is obtained;

step 2, single-gear power line point cloud plane segmentation and meshing;

and 3, performing point supplementing calculation, namely performing point supplementing on the power lines in the plane segmentation subsets.

The method for grading the power line point cloud and converting the coordinates comprises the following steps: obtaining the center coordinates of the towers, sequentially constructing rectangles which take the connecting lines of the towers as the center and are expanded by 25 meters from left to right, and taking points in the rectangles as the same gear to realize power line point cloud grading; and meanwhile, calculating the vertical distance offset d from the point to the pole tower connecting line, and calculating the cumulative distance l of the projection distance of the point and the small-size pole tower connecting line on the pole tower connecting line to realize coordinate conversion.

The specific method for realizing coordinate conversion comprises the following steps of calculating the vertical distance offset distance d from a point to a pole tower connecting line, and calculating the cumulative distance l of the projection distance of the point and a small-size pole tower connecting line on the pole tower connecting line:

let the coordinate of the small tower A of any gear be (x)_A,y_A) And the B coordinate of the large tower is (x)_B,y_B) The coordinate of any point P on the power line point cloud is (x)_P,y_P) Constructing a vector

Andwherein the content of the first and second substances,

X_AB＝x_B-x_A，Y_AB＝y_B-y_A

X_AP＝x_P-x_A，Y_AP＝y_P-y_A

calculate the area S of Δ ABP_ABP＝(X_AB×Y_AP-X_AP×Y_AB) /2, the offset distance can be obtained

Accumulated distance

Wherein the offset distance d_PThe right side of a tower connecting line is negative, and the left side of the tower connecting line is positive; if the condition is satisfied

If not, the next gear judgment is carried out.

The single-gear power line point cloud plane segmentation method in the step 2 comprises the following steps: the method comprises the steps of realizing segmentation calculation of power lines with non-coincident plane positions by detecting offset jump, designing a one-dimensional grid parallel to a connecting line of center points of a tower, and meshing a segmented point set.

The method for realizing the segmentation calculation of the power line with non-coincident plane positions by detecting the offset jump, designing a one-dimensional grid parallel to a connecting line of the center points of the tower and meshing the segmented point set comprises the following steps:

step 2.1, sorting any one-gear power Line point cloud set converted in the step 1 according to an ascending order of offset distance to obtain an ascending order set Line { P }₁,P₂,…,P_n}; according to d_i-d_i-1When the value is more than 2, n is more than or equal to i and is used as a constraint condition, searching offset catastrophe points, wherein n is the number of ascending set points; suppose t mutation points, P respectively, are retrieved_m1,P_m2,…,P_mtThen the subset is divided into

[line₁{P₁,P₂,…P_m1-1},line₂{P_m1,P_m1+1,…P_m2-1},…,line_t+1{P_mt,P_mt+1,…P_n}]。

Step 2.2, setting the accumulative distance step length to be 1m and the number of grids to be

The one-dimensional grid of (1) dividing any point P in the point cloud subset into [ l ]_P]And putting the grid indexes into corresponding grids, and arranging the points in the sub-grids according to the ascending order of the elevation.

Step 3, the supplementary point calculation method comprises the following steps:

step 3.1, setting the ith sub-Grid in the one-dimensional Grid to be Grid in step 2.2_i{P₁,P₂,…,P_n}, according to l_t-l_t-1If the number of the accumulation distance mutation points is more than 4, and t is less than or equal to N, the number of the grid power line elements L is greater than or equal to N_iN + 1; the number of power lines partitioning a subset is Max ({ L) across the entire grid₁,L₂,…,L_n})；

Step 3.2, in the initial point collecting stage, starting from the small grid with the accumulation distance of 0, traversing the grid backwards, and when the number L of power lines of the grid is_i＝Max({L₁,L₂,…,L_nAnd) adding the lowest point of the grid into an initial sampling point array; when the number of the initial sampling point groups is equal to 5, stopping traversing, and performing polynomial fitting to obtain initial local partA power line equation;

3.3, after obtaining an initial local power line equation, carrying out extrapolation calculation in front and back directions of the accumulated distance;

step 3.4, recording the local power line equation of each section and the corresponding sampling point array, and setting the first point of the sampling point array as P_starThe last point is P_endUsing local power line equations, the accumulation distance is given by l_starStarting with_endFor termination, carrying out interpolation calculation at an interval of 0.1m to obtain a local power line compensation point; putting all interpolation results of the local power line equations to the same point set, and rotating the interpolation results to an original point cloud coordinate system through coordinates to obtain a point supplementing result of the lowest power line under the plane segmentation subset;

3.5, establishing a kd-tree by using a local power line point complement set, and removing the laser points on the power lines extracted by the grid at this time through proximity search; repeating the steps 3.1-3.4, and sequentially supplementing points to the power lines in the plane segmentation subsets.

The polynomial fitting method adopts a least square theory, takes the accumulated distance l as an independent variable, performs linear fitting with the offset distance d, and performs 3 times of term fitting with the elevation h, and a specific equation is as follows:

in the formula, a₁,a₂And b₁,b₂,b₃,b₄Respectively are linear equation and cubic equation parameters.

Step 3.3 the extrapolation calculation method comprises: taking the rear as an example, the last point P is counted from the initial sampling point_endThe grid is traversed backwards, and the lowest point of the grid is set as P_low(d_low,l_low,h_low) To do so by

Judging whether the lowest point of the grid is a new sampling point or not for the constraint condition, whereinb₁,b₂,b₃,b₄Cubic equation parameters of a local power line equation; when the number of new sampling points is equal to 5, P is utilized_endCarrying out polynomial fitting with the new sampling point again to obtain a new local power line equation; and (4) continuing extrapolation calculation by taking the new local power line equation as a reference, and continuously repeating the process until the grid is traversed.

The invention has the beneficial effects that:

the invention provides a point cloud automatic point compensating method for a laser LiDAR power line, which has the following beneficial effects:

firstly, the method only needs the center coordinates of the tower and the power line point cloud as initial data, automatic point acquisition does not need manual intervention, and the automation degree is high;

secondly, the quality of the newly-collected point is detected by local point fitting polynomial equation extrapolation point positions, and the influence of noise points and non-root power line points on the fitting polynomial equation is avoided. Meanwhile, a piecewise fitting mode is adopted, so that each piecewise fitting equation is most fit with the power line, and the point-supplementing precision is higher.

The method carries out horizontal segmentation through offset distance information, sequentially segments the power line at the lowest part of each horizontal segmentation layer, is suitable for point cloud point compensation of single-loop, same-tower double-loop and same-tower multi-loop overhead line power lines, and has strong algorithm applicability and high robustness.

The technical problems of low point cloud density and serious point cloud loss of the laser LiDAR power line are solved, the workload of manual data processing is reduced, the point location precision of point compensation is improved, and data support is provided for functional applications such as follow-up power line reconstruction and dangerous point detection.

The specific implementation mode is as follows:

step 1, power line point cloud grading and coordinate conversion

The power line point cloud grading and coordinate conversion method in the step 1 comprises the following steps: and (3) acquiring center coordinates of the towers, sequentially constructing rectangles which are extended by 25 meters from left to right and take the connecting lines of the towers as the center, and taking points in the rectangles as the same gear to realize power line point cloud grading. And meanwhile, calculating the vertical distance offset d from the point to the pole tower connecting line, and calculating the cumulative distance l of the projection distance of the point and the small-size pole tower connecting line on the pole tower connecting line to realize coordinate conversion. The specific calculation theory is as follows:

let the coordinate of the small tower A of any gear be (x)_A,y_A) And the B coordinate of the large tower is (x)_B,y_B) The coordinate of any point P on the power line point cloud is (x)_P,y_P) Constructing a vector

Wherein the content of the first and second substances,

X_AB＝x_B-x_A，Y_AB＝y_B-y_A

X_AP＝x_P-x_A，Y_AP＝y_P-y_A

calculate the area S of Δ ABP_ABP＝(X_AB×Y_AP-X_AP×Y_AB) /2, the offset distance can be obtained

Accumulated distanceWherein the offset distance d_PThe right side of the tower connecting line is negative, and the left side is positive. If the condition is satisfied

If not, the next gear judgment is carried out.

Step 2, single-gear power line point cloud plane segmentation and gridding

The single-gear power line point cloud plane segmentation method in the step 2 comprises the following steps: when the airborne laser LiDAR technology is used for field collection of point cloud data, the requirements are generally that a single power line is approximately parallel to a connecting line of center points of towers and the point offset distances d on the single power line are approximately equal under the conditions of no wind and breeze. Therefore, the method realizes the segmentation calculation of the power line with non-coincident plane positions by detecting the offset jump, simultaneously designs the one-dimensional grid parallel to the connecting line of the central points of the tower, and gridds the segmented point set.

(1) Sorting any power Line point cloud set converted in the step one according to ascending order of offset distance to obtain ascending order set Line { P₁,P₂,…,P_n}. According to d_i-d_i-1And if the value is more than 2, n is more than or equal to i and is used as a constraint condition to search offset catastrophe points, wherein n is the number of ascending set points. Suppose t mutation points, P respectively, are retrieved_m1,P_m2,…,P_mtThen each partition subset is [ line ]₁{P₁,P₂,…P_m1-1},line₂{P_m1,P_m1+1,…P_m2-1},…,line_t+1{P_mt,P_mt+1,…P_n}]。

PS, in the dividing result, a plurality of electric lines of force are overlapped in the vertical direction.

(2) Setting the step length of accumulation distance as 1m and the number of grids asThe one-dimensional grid of (2) is used for dividing any point P in the point cloud subset obtained by the step (2.1) according to [ l ]_P]And putting the grid indexes into corresponding grids, and arranging the points in each sub-grid according to an ascending order of elevation.

Step 3, compensation point calculation

The method for calculating the supplementary points in the step 3 comprises the following steps:

(1) setting the ith Grid in the one-dimensional Grid as Grid in step 2.2_i{P₁,P₂,…,P_n}, according to l_t-l_t-1N (the value is set to be 4m and can be changed along with the change of the voltage level according to the safety requirement of the inter-phase distance) is used as a constraint condition to retrieve the accumulated distance catastrophe points, and if N accumulated distance catastrophe points are detected, the number L of the grid power lines is_iN + 1. Traversing the entire grid, the number of power lines currently partitioning the subset is Max ({ L)₁,L₂,…,L_n})。

(2) In the initial point collecting stage, the grid is traversed backwards from the small grid with the accumulation distance of 0, and when the number L of power lines of the grid is_i＝Max({L₁,L₂,…,L_n}) the lowest point (first point) of the grid is added to the initial sampling point array. And when the number of the initial sampling point groups is equal to 5, stopping traversing, and performing polynomial fitting to obtain an initial local power line equation.

The polynomial fitting method of the method adopts a least square theory, takes the accumulated distance l as an independent variable, performs linear fitting with the offset distance d, and performs 3 times of term fitting with the elevation h, and a specific equation is as follows:

in the formula, a₁,a₂And b₁,b₂,b₃,b₄Respectively, linear equation and polynomial equation parameters.

(3) After the initial local power line equation is obtained, extrapolation calculation in the front and rear directions of the accumulation distance needs to be performed. Taking the rear as an example, continuously traversing the grid backward from the grid where the last point of the initial sampling point group is located, and setting the lowest point of the grid as P_low(d_low,l_low,h_low) To do so byJudging whether the lowest point of the grid is a new sampling point or not for the constraint condition, wherein

b₁,b₂,b₃,b₄And (4) performing cubic equation parameters for the local power line equation. When searching for the number of new sampling points, etc. 5, P is utilized_endAnd carrying out polynomial fitting with the new sampling point again to obtain a new local power line equation. And then, taking the new local power line equation as a reference, continuing extrapolation calculation, and continuously repeating the process until the grid is traversed.

According to the method, whether the grid of the point to be acquired should lack the power line point of the point at this time is detected through the form of extrapolation of the local power line fitting equation, so that the influence of the upper-layer overlapped power line point on equation fitting is avoided, and the local power line fitting effect is most fit with the actual point cloud.

(4) Recording each section of local power line equation and the corresponding sampling point array, and setting the first point of the sampling point array as P_starThe last point is P_endUsing local power line equations, the accumulation distance is given by l_starStarting with_endFor termination, interpolation calculation is performed at intervals of 0.1m to obtain the local power line compensation point. And (4) putting all the interpolation results of the local power line equations into the same point set, and rotating to an original point cloud coordinate system through coordinates to obtain a point supplementing result of the lowest power line under the plane segmentation subset.

(5) And establishing a kd-tree by using a local power line point complement set, and removing the laser points on the power lines extracted this time by the grid through proximity search. Repeating the steps 3.1-3.4, and sequentially supplementing points to the upper power lines in the plane segmentation subsets.