CN107610223A - Power tower three-dimensional rebuilding method based on LiDAR point cloud - Google Patents

Power tower three-dimensional rebuilding method based on LiDAR point cloud Download PDF

Info

Publication number
CN107610223A
CN107610223A CN201710853206.2A CN201710853206A CN107610223A CN 107610223 A CN107610223 A CN 107610223A CN 201710853206 A CN201710853206 A CN 201710853206A CN 107610223 A CN107610223 A CN 107610223A
Authority
CN
China
Prior art keywords
mrow
msub
tower
point
point cloud
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.)
Granted
Application number
CN201710853206.2A
Other languages
Chinese (zh)
Other versions
CN107610223B (en
Inventor
陈浩
翟瑞聪
张峰
许志海
彭炽刚
李雄刚
廖如超
杨成城
杜俊明
罗世奇
李威
李闪闪
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Wuhan Hui Zhuo Airlines Technology Co Ltd
Guangdong Power Grid Co Ltd Patrol Operation Center
Original Assignee
Wuhan Hui Zhuo Airlines Technology Co Ltd
Guangdong Power Grid Co Ltd Patrol Operation Center
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Wuhan Hui Zhuo Airlines Technology Co Ltd, Guangdong Power Grid Co Ltd Patrol Operation Center filed Critical Wuhan Hui Zhuo Airlines Technology Co Ltd
Priority to CN201710853206.2A priority Critical patent/CN107610223B/en
Publication of CN107610223A publication Critical patent/CN107610223A/en
Application granted granted Critical
Publication of CN107610223B publication Critical patent/CN107610223B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

The invention belongs to laser radar point cloud data information extraction technology field, is related to a kind of power tower three-dimensional rebuilding method based on LiDAR point cloud.Comprise the following steps:S1. the decomposition of power tower, by carrying out statistical analysis to density and width, power tower is decomposed into tower body and tower head two parts;S2. tower body is rebuild based on data-driven, angle steel joint is extracted and split, and the progress 3 d-line fitting of RANSAC algorithms is based on to four main wheel profiles of tower body;S3. tower head is rebuild based on model-driven, predefine a model library for including tower head fundamental type, then tower head fundamental type is identified using Shape context algorithm, then Metropolis Hastings algorithms is combined to the optimized parameter of estimation tower head model with simulated annealing;S4. the position and direction in S2, S3 step acquired results are combined, and obtain complete three-dimensional power tower model.The present invention can effectively, accurately rebuild power tower, meet power transmission line three-dimensional visual, digitized demand.

Description

Power tower three-dimensional rebuilding method based on LiDAR point cloud
Technical field
The invention belongs to laser radar point cloud data information extraction technology field, more specifically, is related to one kind and is based on The power tower three-dimensional rebuilding method of LiDAR point cloud.
Background technology
The power tower infrastructure important as ultra-high-tension power transmission line, it inhales in terms of being reconstituted in power transmission line three-dimensional visual Increasing concern is drawn.In the past few years, power tower models the most frequently used method and mainly carried out with CAD or 3dmax Manually model, this method expends substantial amounts of human and material resources and reconstruction precision is relatively low.One kind is badly in need of certainly in administration of power networks department Dynamicization, high efficiency, high-precision power tower method for reconstructing meet the visualization of modern transmission line of electricity, digitized demand.With The rapid development of Airborne LiDAR Technology, accurately high density, 3D point cloud provide an effective solution to the demand.
At present, the scholar for having many has done substantial amounts of research in terms of the reconstruction based on LiDAR point cloud objective, and Natural feature on a map and man-made features are rebuild and achieve larger progress.However, complexity and species due to electric power tower structure Diversity, to power tower rebuild research it is also very rare.Li Qingquan etc. proposes a kind of power tower based on model-driven Power tower is decomposed into tower head, tower body and column foot three parts by method for reconstructing, this method first, then to tower body part with four masters Plane is rebuild, and tower head part is classified using SVM classifier.This method is carried out according only to density feature to power tower Decompose, lack applicability, other common type power towers (such as cat-head transmission tower, wineglass tower, hexagonal tower) often go out when being decomposed Existing error result;Further, since this method, which introduces, manually establishes tower head model so that automaticity reduces, and lacks Practicality.Guo Bo etc. proposes a kind of power tower method for reconstructing based on random geometry, and this method is by RJMCMC samplers and mould Intend annealing to be combined, automatically solve the problems, such as power tower type and model parameter, but this method is inefficient, it is necessary to expend a large amount of Iterations judge power tower type, and in power tower model parameter estimation process, the geometry not accounted between parameter closes System, causes substantial amounts of nuisance parameter needs to be estimated.
At present, the method that building is rebuild is broadly divided into following two strategies:Based on data-driven strategy and based on mould The strategy of type driving.Strategy based on data-driven is using method for reconstructing from top to bottom.Most of building is rebuild treated Journey generally comprises two committed steps:The extraction at building roof edge and the reconstruction of topological relation.Miscellaneous roof structure or height The cloud data of density, method extraction edge of the generally use based on patch division, is such as increased based on edge or region, is three-dimensional RANSAC algorithms, classification or feature clustering etc..Reconstructed results based on data-driven are not influenceed by model library integrality, theoretical On can rebuild any shape.When data than it is more complete when, the method for data-driven is provided simple target and accurately described; However, when shortage of data or during comprising more noise, this method can produce reconstruction error or even failure.
For strategy based on model-driven using method for reconstructing from bottom to up, this method is based on pre-defined model library 's.This method mainly includes two committed steps:Optimum Matching and corresponding model parameter are carried out with the model in model library most Optimization Solution.In many based in the method for reconstructing of model-driven, most basic hypothesis is the set that building is many metopes. Many methods such as RJMCMC is introduced for certainly model parameter problem, and shows larger development potentiality.Driven based on model Dynamic method has robustness to the quality of data and is applied to large scene, because topological relation defines in a model in advance, because This, this method can play its superiority when a cloud density is smaller and ensure correct topological relation.However, the reconstruction of this method As a result influenceed by model library, when when modulus shape parameter is more, this method is less efficient.
Due to the complexity of object construction and the diversity of shape, some way of only voucher one is difficult to meet Modern Reconstruction Requirement.Therefore, it is suggested with reference to the combination drive strategy of data-driven and model-driven in recent years.This method is in target Phase of regeneration will build regular (such as collimation, coplanarity, symmetry) introducing and carry out Optimized model.It is tactful different from single, The strategy of combination drive blends the advantage of two methods, has both remained the flexibility of data-driven, it may have model-driven Robustness.
The content of the invention
The defects of it is an object of the invention to overcome prior art, there is provided a kind of power tower based on LiDAR point cloud is three-dimensional Method for reconstructing, can effectively, accurately rebuild power tower, meet power transmission line three-dimensional visual, digitized demand.
To solve the above problems, technical scheme provided by the invention is:A kind of power tower Three-dimensional Gravity based on LiDAR point cloud Construction method, wherein, comprise the following steps:
S1. the decomposition of power tower, the LiDAR point cloud data of power tower are obtained, passes through the density and width to power tower point cloud Degree carries out statistical analysis, and power tower is decomposed into tower body and tower head two parts;
S2. tower body is rebuild based on data-driven, angle steel joint is extracted and split, four main wheel profiles of tower body are based on RANSAC algorithms carry out 3 d-line fitting;
S3. tower head is rebuild based on model-driven, a model library for including tower head fundamental type is predefined, then using shape Shape contextual algorithms identify tower head fundamental type, then Metropolis-Hastings algorithms are combined into estimation with simulated annealing The optimized parameter of tower head model;
S4. the position and direction in S2, S3 step acquired results are combined, and obtain complete three-dimensional power tower mould Type.
In the present invention, it is contemplated that the architectural feature of power tower, will first by carrying out statistical analysis to density and width Power tower is decomposed into tower body and tower head two parts;Then data-driven is based on to tower body and carries out three-dimensional reconstruction, to four of tower body Main wheel profile is based on RANSAC algorithms and carries out 3 d-line fitting;Three-dimensional reconstruction is carried out being based on model-driven to tower head, first The tower head model library of a predefined parametrization is needed, model library identification tower head is then based on using shape context algorithms The tower head type of point cloud, corresponding column is estimated with the method that simulated annealing is combined using Metropolis-Hastings samplers The parameter of head model, finally it is combined according to two-part position and direction, obtains complete three-dimensional power tower model.
Power tower is that its tower body structure is simple, type is single with certain artificial atural object for building rule, tower head species compared with It is more and complicated.In view of the architectural feature of power tower, power tower can be decomposed into two parts:Tower body and tower head;It is although electric The type of power tower has a variety of, but characteristic face all includes two obvious features:(1) local density is maximum, and (2) part width is most It is small.Wherein, density is defined as the point cloud number of tower cross section, and width is defined as each section to the ultimate range of kernel of section; Therefore, the present invention to power tower using the method for statistical analysis come location feature face.It is straight in order to obtain the statistics of density and width Fang Tu, if power tower is divided into dried layer, by counting each layer of density and width, choose simultaneously have density local maxima and Position of the minimum layer of width as characteristic face.
Further, density and width the progress statistical analysis in described S1 steps to power tower point cloud include following step Suddenly:
If power tower point cloud is S101. divided into dried layer according to Δ h at equal intervals;
S102. calculate each layer of density and width, and established respectively as X-axis, using the number of plies as Y-axis using density and width close Spend histogram and width histogram;
S103. one is defined centered on current layer number, and size is the moving window of 2L*1 sizes respectively in density and width Spend and very dense value and width minimum are found in histogram, and layer where maximum and minimum is marked;
S104. power tower select first from bottom to up and meanwhile meet this layer of density local maxima and it is upper and lower two layers in the range of Position of the number of plies as characteristic face of width Local Minimum be present, decomposed by boundary line of characteristic face, portion more than characteristic face It is decomposed into tower head, the decomposed below characteristic face is tower body.
Distinguished point based extracts and the method for segmentation is a kind of method commonly used in data-driven strategy.In order to improve segmentation Accuracy, the present invention only requires that angle steel joint is extracted;In order to correctly extract the angle point of tower body, tower body is first according to elevation If it is divided into dried layer at equal intervals.Each layer can be regarded as with concentric rectangle.To each layer, extracted using algorithm of convex hull Its profile point, not all profile point is all located at four angles of rectangle, therefore, it is necessary to is based on pipe to the profile point after extraction Road algorithm is simplified.
Further, in described S2 steps angle steel joint carry out extraction comprise the following steps with segmentation:
S2011. current point and adjacent 2 points of angle are calculated, if angle is more than predetermined threshold df;Then retain current point, it is no Then, reject;
S2012. simplified by pipeline algorithm, delete most of non-angle point;
S2013. the angle point simplified to each layer, split according to it in the position at minimum enclosed rectangle center, will Angle point is divided into the subset of four direction.
After above-mentioned steps are handled, the subset of four tower body angle points is obtained.Due to not all letter in each subset The angle point of change is all at four angles, and therefore, the present invention is using the 3 d-line fitting algorithm based on RANSAC to each angle point subset It is fitted.
Further, the contour line fitting based on RANSAC comprises the following steps in described S2 steps:
S2021. two point P are randomly selected from the subset obtained by S1 steps1(x1,y1,z1) and P2(x2,y2,z2), and calculate 3 d-line equation L where it;
S2022. for the every bit in subset, calculate it and arrive straight line L distance, if the distance of current point to straight line is small In distance threshold Td, then it is available point, otherwise, is defined as Null Spot;And the ratio inner_ratio of available point is counted, Wherein, inner_ratio is that available point accounts for the ratio always counted;
S2023. stochastical sampling is repeated, until meeting that P is less than default fiducial probability TP or sampling number M and is more than default maximum Sampling number TN, choose parameter of the one group of minimum parameter of regression criterion as the subset 3 d-line equation;Wherein, P is to adopt The probability of appearance collection;
S2024. after four subsets are fitted, symmetry is added, coplanarity restrictive condition optimizes to model.
Preferably, described predetermined threshold df is 30 °.The angle point at four angles is only needed in the present invention, therefore df=is set 30°;When df set it is bigger when, profile point is fewer, but Tf should be less than 90 °.
Further, described 3 d-line equation L calculation is:
Further, the probability P of described sampling subset meets following relationship:
P=1- (1-inner_ratiom)M,
Wherein, m is to build the minimum parameters number required for a judgment criteria.
Further, identify that tower head fundamental type comprises the following steps in described S3 steps:
S3011. it is two-value image by all tower head model conversations in tower head point cloud and model library to be identified;
S3012. the outline of all images is extracted, contour edge is sampled to obtain n sampled point point set P={ p1, p2, p3... ..pn};
S3013. the Shape context h of each sampled point is calculated according to Shape context algorithmi(k), calculation formula such as formula (1), k={ 1,2 ..K }, K=M*N;
hi(k)={ q ≠ pi:(q-pi∈ bin (k) }, (1)
S3014. calculate between the shape histogram of image to be matched and the shape histogram each pair point of each phantom images Matching power flow, calculation formula such as formula (2), wherein, hi(k) the point p for being target PiShape histogram;hj(k) for target Q's Point qiShape histogram,
S3015. Point matching operation is carried out based on the cost matrix C being calculated, chooses the mould with minimum shape distance Type image is best matching result, and the model is that tower head point cloud corresponds to model.
Further, parameter optimization comprises the following steps in described S3 steps:
S3021.Gibbs energy definitions, it is (3) that Gibbs energy, which is expressed as formula,:
Using alpha shape algorithms come the key point of extraction model and point cloud, distance um(xi) it is each model key point To the distance of nearest tower head point cloud key point, distance up(xi) for each tower head point cloud key point to nearest model key point away from From, wherein,
Wherein,Average distance us of the degree of similarity u (x) by model to point cloudm(xi) and put cloud to model Average distance up(xi) composition, and u (x)=aum(x)+bup(x), a+b=1, a and b are respectively the weights of two distances;N is mould Type key point number, m are tower head point cloud key point number;
S3022.Metropolis-Hastings is sampled and simulated annealing, first, according to suggestion distribution q (x*| x) searching A current parameter value x candidate parameter value x is provided in rope space X*, then, candidate parameter value is calculated according to calculation formula (6) Receptance A (x, x*);If receptance A (x, x*) be more than default acceptance threshold Ta, then by candidate value x*It is no instead of currency x Then, currency x is retained;
Further, described acceptance threshold Ta=0.9+0.1*random (0,1), random (0,1) are 0~1 In the range of carry out random sampling result.
Metropolis-Hastings (MH) is sampled as the most popular MCMC method of samplings, and it is widely used In the approximation of appraising model parameter.MH sampling core concept be by statistical sampling, it is near with a simple question Like the combinatorial problem of complexity.It is related to two distributions in MH samplings:Target distribution p (x) and suggestion distribution q (x*|x).First, root It is suggested that distribution q (x*| a current parameter value x candidate parameter value x x) is provided in the X of search space*.Then, according to formula (10) receptance A (x, the x of candidate parameter value are calculated*);If receptance A (x, x*) be more than default acceptance threshold Ta, then by candidate Value x*Instead of currency x, otherwise, retain currency x.Pass through the repeated sampling of statistics, candidate value x*Finally tend towards stability.For Sampling is prevented to be absorbed in local optimum, acceptance threshold Ta=0.9+0.1*random (0,1), random (0,1) are the model 0~1 Enclose interior progress random sampling result.Although MH sampling algorithm principle very simples, realizing for it are relatively difficult.This be mainly because For the selection of object function and suggestion function.Different selection results can cause sampled result different.The present invention is selected Gauss Distribution is as distribution is suggested, using Gibbs energy as target distribution.Due to suggesting the symmetry of distribution, q (x*| x)=q (x | x*), therefore, receptance can be reduced to formula:Because stochastical sampling is seldom in the attached of actual value Closely, therefore, MH sampling efficiencies are than relatively low.It expends the more non-interest region of time search.In order that sampling is more efficiently, mould Intend annealing to be used in the present invention.Simulated annealing formulates an Annealing Strategy first since a higher initial temperature, With the decline of temperature, target distribution progressively tends towards stability, and finally reaches global optimum with higher receptance.In simulated annealing In, the selection of Annealing Strategy is an important influence factor, and it is by simulating a nonhomogeneous Markov chain come appraising model Optimized parameter.Unlike MH samplings, target distributionIt instead of p (x).Wherein, temperature when Ti is iterations i Degree, and meet limi→∞Ti=0.
Compared with prior art, beneficial effect is:A kind of power tower Three-dimensional Gravity based on LiDAR point cloud provided by the invention Construction method, the construction feature of power tower is taken into full account, to strategy corresponding to different electric power tower section uses;Rebuild in tower head During, geometrical relationship is taken full advantage of between tower body reconstruction information, original point cloud information and parameter to reduce parameter as far as possible Number and parameter search space;The present invention can effectively, accurately rebuild power tower, meet power transmission line three-dimensional visual, number The demand of word.
Brief description of the drawings
Fig. 1 is power tower structural representation of the present invention.
Fig. 2 is angle point grid of the present invention with being layered schematic diagram in segmentation.
Fig. 3 is angle point grid of the present invention and angle point grid schematic diagram in segmentation.
Fig. 4 is angle point grid of the present invention and angle point rough schematic view in segmentation.
Fig. 5 is that angle point grid of the present invention splits schematic diagram with angle point in segmentation.
Fig. 6 is some subset schematic diagram of angle point during tower body of the present invention is rebuild.
Fig. 7 is angle point fitting result schematic diagram during tower body of the present invention is rebuild.
Fig. 8 is angle point tower body optimum results schematic diagram during tower body of the present invention is rebuild.
Fig. 9 is angle point optimum results top view illustration during tower body of the present invention is rebuild.
Figure 10 is the structural representations of tower head model library Model 1 that the present invention parameterizes.
Figure 11 is the structural representations of tower head model library Model 2 that the present invention parameterizes.
Figure 12 is the structural representations of tower head model library Model 3 that the present invention parameterizes.
Figure 13 is the structural representations of tower head model library Model 4 that the present invention parameterizes.
Figure 14 is Shape context algorithm point cloud configuration sampling point schematic diagram of the present invention.
Figure 15 is Shape context algorithm model configuration sampling point schematic diagram of the present invention.
Figure 16 is Shape context algorithm Shape context schematic diagram of the present invention.
Wherein, 1 is tower head, and 2 be tower body.
Embodiment
Accompanying drawing being given for example only property explanation, it is impossible to be interpreted as the limitation to this patent;It is attached in order to more preferably illustrate the present embodiment Scheme some parts to have omission, zoom in or out, do not represent the size of actual product;To those skilled in the art, Some known features and its explanation may be omitted and will be understood by accompanying drawing.Being given for example only property of position relationship described in accompanying drawing Explanation, it is impossible to be interpreted as the limitation to this patent.
A kind of power tower three-dimensional rebuilding method based on LiDAR point cloud, wherein, comprise the following steps:
Step 1:The decomposition of power tower, the LiDAR point cloud data of power tower are obtained, pass through the density to power tower point cloud Statistical analysis is carried out with width, power tower is decomposed into tower body and tower head two parts.
Power tower is that its tower body structure is simple, type is single with certain artificial atural object for building rule, tower head species compared with It is more and complicated, it is contemplated that the architectural feature of power tower, power tower can be decomposed into two parts:Tower body and tower head, such as Fig. 1 institutes Show.In order to clearly distinguish tower head and tower body, lower edge or the tower cross section sharp change in elevation that cross-arm is descended defined in the present invention are spy Sign face, as shown in Fig. 1 a, 1b, part more than characteristic face is defined as tower head, and the part below characteristic face is defined as tower body.Although The type of power tower has a variety of, but characteristic face all includes two obvious features:(1) maximum (2) part width of local density is most It is small.Wherein, density is defined as the point cloud number of tower cross section, and width is defined as each section to the ultimate range of kernel of section; Therefore, herein to power tower using the method for statistical analysis come location feature face.In order to obtain the statistics Nogata of density and width Figure, if power tower is divided into dried layer, by counting each layer of density and width, choose has density local maxima and width simultaneously Spend position of the minimum layer as characteristic face.
Comprise the following steps that:
1. by power tower point cloud, according to elevation, if Δ h is divided into dried layer at equal intervals.Δ h value sets smaller, characteristic face positioning Precision is higher, generally Δ h/2, but to ensure each layer all a little.In the present invention, Δ h=0.5 rice;
2. calculating each layer of density and width, and density is established as X-axis, by Y-axis of the number of plies using density and width respectively Histogram and width histogram;
3. defining one centered on current layer number, size is straight in density and width respectively for the moving window of 2L*1 sizes Very dense value and width minimum are found in square figure.If the density of current layer is maximum in the range of 2L*1, the layer is carried out Mark.If the width of current layer is minimum in the range of 2L*1, the layer is marked.The size 2*L of window and electric power tower height Spend relevant, L should be less than the minimum range of adjacent two cross section when setting.L=3 rice is set in the present invention;
4. from bottom to up select first simultaneously meet this layer of density local maxima and it is upper and lower two layers in the range of width be present Position of the number of plies of Local Minimum as characteristic face.Decomposed by boundary line of characteristic face, decomposed more than characteristic face is Tower head, the decomposed below characteristic face is tower body.
Step 2:Tower body is rebuild based on data-driven, angle steel joint is extracted and split, to four main wheel profiles of tower body 3 d-line fitting is carried out based on RANSAC algorithms.
Wherein, the extraction and segmentation of angle point are carried out first;As shown in Figure 2 and Figure 3, not all profile point is all located at Four angles of rectangle, therefore, it is necessary to the profile point after extraction is simplified based on pipeline algorithm.Calculate current point and adjacent two The angle of point, if angle is more than predetermined threshold df, retains current point, otherwise, rejects.Due to only needing four angles herein Angle point, therefore set df=30 °.When Tf set it is bigger when, profile point is fewer, but df should be less than 90 °.As shown in figure 4, pass through Pipeline algorithm is simplified, and can delete most of non-angle point.Finally, the angle point simplified to each layer, according to it outside minimum The position for connecing rectangular centre is split, and angle point is divided into the subset of four direction, and final result is as shown in Figure 5.Each subset It is approximately 3 d-line in shape.
Secondly, the contour line fitting based on RANSAC;After above-mentioned steps are handled, the son of four tower body angle points is obtained Collection.Because the angle point of not all simplification in each subset is at four angles, therefore, the three-dimensional based on RANSAC is used herein Algorithm of fitting a straight line is fitted to each angle point subset.
As shown in fig. 6, in order to obtain the 3 d-line equation of each subset, the point of selection two P random first1(x1,y1, z1) and P2(x2,y2,z2), and calculate the 3 d-line equation L where it.Wherein, 2 points of 3 d-line equation L is according to formula (1) Calculated;Calculation formula is as follows:
Then, for the every bit in subset, calculate it and arrive straight line L distance.If the distance of current point to straight line is small In distance threshold Td, then it is available point, otherwise, is defined as Null Spot.Distance threshold Td is noise spot degrees of tolerance, and Td is set Put it is smaller, participate in calculate noise spot it is fewer, the present invention in set Td=0.1 rice.Statistics available point accounts for total points of the subset Ratio inner_ratio.Stochastical sampling is repeated, until meeting that P is less than default fiducial probability TP or sampling number M more than default Maximum sampling number TN, parameter of the one group of minimum parameter of regression criterion as the subset 3 d-line equation is chosen, such as Fig. 7 institutes Show.
Wherein, under certain fiducial probability TP, bases subset minimum number of samples M in RANSAC algorithms and at least The probability P of a correct sampling subset meet following relationship:
P=1- (1-inner_ratiom)M,
In above formula, inner_ratio is that available point accounts for the ratio always counted, required for m is one judgment criteria of structure Minimum parameters number.It is 0.98 that the present invention, which sets fiducial probability, and maximum sampling number is 30 times.In the plan that four subsets are all independent After conjunction, tower body reconstructed results are optimized using specification (such as symmetry, coplanarity, collimation) is built so that rebuild knot Fruit meets actual architectural feature, as shown in Figure 8.In addition, tower body reconstructed results can be used in tower head weight as auxiliary information During building.As shown in figure 9, it is determined that behind the position of characteristic face, the long tdx and wide tdy of characteristic face are calculated, and as tower The nose heave known parameters built.
Step 3:Tower head is rebuild based on model-driven, a model library for including tower head fundamental type is predefined, then adopts Tower head fundamental type is identified with Shape context algorithm, then Metropolis-Hastings algorithms are combined with simulated annealing Estimate the optimized parameter of tower head model.
First, the tower head model library of parametrization is established.
Tower head is complicated, wide variety, and the reconstruction of tower head is difficult to realize using the strategy of data-driven.Because tower head has There is certain construction rule, and only main material part needs to rebuild, therefore, using carrying out weight to tower head based on the method for model-driven Build.The strategy of model-driven is to be based on predefined model library.With reference to the fundamental type of Chinese high-voltage power tower, a three-dimensional ginseng The model library of numberization is defined first.The integrality of model library directly affects the result of reconstruction.The model driven method of the present invention It is only capable of rebuilding the tower head type tower head point cloud defined in model library.If tower head types of models does not define, need first to exist The type is defined in model library, then rebuild again.
As shown in Figure 10 to Figure 13, in the present invention, tower head model library includes four kinds in Chinese 220-500kV high voltage power transmissions The widely used power tower type of circuit.To put it more simply, tower head model is represented with key point.In order to obtain the coordinate of key point, Tower head parameter is mainly formed by feature is high with feature length.Due to the complexity of tower head, tower head model includes many parameters, causes to hold high Expensive calculation cost carrys out the optimized parameter of appraising model.In order to reduce model parameter number and parameter search space, the present invention will Make full use of the geometrical relationship between tower body reconstructed results, original point cloud information and parameter.The tower head parameter of the four types such as institute of table 1 Show.Wherein, the tower head parameter that unknown parameter is estimated for needs, it is known that parameter is that can be entered according to information such as the geometrical relationships between parameter The parameter that row derives, is not involved in parameter estimation.
The tower head model parameter of table 1
From table 1 it follows that tower head model includes four basic parameters:tdx,tdy,Height,Length.tdx It is special positive length and width respectively with tdy, can be obtained from tower body reconstructed results.Height and Length is tower head respectively Size, it can be obtained from the length and width of original tower head point cloud.Other parameters can be entered according to the geometrical relationship with other specification Row is derived by.By taking Model1 as an example, H3、L3Difference can obtain from formula (2) and formula (3).
H3=Height-H1-H2; (2)
In addition, parameter search space can also be according to the reduction of geometrical relationship high degree.So, the efficiency that tower body is rebuild It can be greatly improved.
Secondly, tower head type is identified.
Difference is very big in shape for different types of tower head, therefore, tower head type can be entered according to the shape of tower head Row identification.In the present invention, tower head point cloud and tower head model are first converted into two-value image, are then calculated using Shape context Method realizes the matching of tower head point cloud image and corresponding phantom images.
Shape context is the Feature Descriptor measured shape similarity and find point correspondence.It knows in numeral Not, the retrieval based on similitude, trade mark retrieval and Three-dimensional target recognition etc. extensive use.As shown in Figure 14, Figure 15, shape The most basic thought of contextual algorithms is the n sampled point of selection on profile, then calculates the Shape context each put.Its In, the Shape context each put is defined as histogram of the remaining n-1 point to the relative coordinate of current point, with current point piFor In the center of circle, the local that R is radius N number of concentric circles is established by logarithm distance interval.By this region along the circumferential direction M deciles, formed Target shape template as shown in figure 16.Current point is reduced in template the point minute in each sector to the vectorial relative position of other each points Cloth number.The statistical distribution histogram h of these pointsi(k), referred to as point piShape context.By calculating image and model to be matched Shape distance between image each pair point, finally the type of phantom images of the selection with minimum cost is as the tower head point cloud Type.
Comprise the following steps that:
1. it is two-value image by all tower head model conversations in tower head point cloud and model library to be identified;
2. the outline of all images of extraction, samples to obtain n sampled point point set P={ p to contour edge1,p2,p3, ...pn};
3. the Shape context h of each sampled point is calculated according to Shape context algorithmi(k), formula such as formula (4), k= { 1,2 ..K }, K=M*N;
hi(k)={ q ≠ pi:(q-pi∈ bin (k) }, (4)
4. calculate between the shape histogram of image to be matched and the shape histogram each pair point of each phantom images With cost, formula such as formula (5), wherein, hi(k) the point p for being target PiShape histogram;hj(k) the point q for being target QiShape Shape histogram;
5. carrying out Point matching operation based on the cost matrix C being calculated, the model shadow with minimum shape distance is chosen As being best matching result, the model is that tower head point cloud corresponds to model.
Finally, tower head process of reconstruction is changed into determining for topic, Gibbs energy for Gibbs energy optimizations solution by the present invention Justice is the similitude of tower head point cloud and model, is used for estimating using Metropolis-Hastings samplers and simulated annealing Calculate the most optimized parameter of model.
Step 4:According to Step 1: Step 2: the position and direction in step 3 acquired results are combined, obtain Whole three-dimensional power tower model.
Obviously, the above embodiment of the present invention is only intended to clearly illustrate example of the present invention, and is not pair The restriction of embodiments of the present invention.For those of ordinary skill in the field, may be used also on the basis of the above description To make other changes in different forms.There is no necessity and possibility to exhaust all the enbodiments.It is all this All any modification, equivalent and improvement made within the spirit and principle of invention etc., should be included in the claims in the present invention Protection domain within.

Claims (10)

1. a kind of power tower three-dimensional rebuilding method based on LiDAR point cloud, it is characterised in that comprise the following steps:
S1. the decomposition of power tower, the LiDAR point cloud data of power tower are obtained, is entered by the density to power tower point cloud and width Row statistical analysis, power tower is decomposed into tower body and tower head two parts;
S2. tower body is rebuild based on data-driven, angle steel joint is extracted and split, four main wheel profiles of tower body are based on RANSAC algorithms carry out 3 d-line fitting;
S3. tower head is rebuild based on model-driven, predefines a model library for including tower head fundamental type, then using in shape Hereafter algorithm identification tower head fundamental type, then Metropolis-Hastings algorithms are combined estimation tower head with simulated annealing The optimized parameter of model;
S4. the position and direction in S2, S3 step acquired results are combined, and obtain complete three-dimensional power tower model.
2. the power tower three-dimensional rebuilding method according to claim 1 based on LiDAR point cloud, it is characterised in that described Density and width progress statistical analysis in S1 steps to power tower point cloud comprises the following steps:
If power tower point cloud is S101. divided into dried layer according to Δ h at equal intervals;
S102. each layer of density and width is calculated, and it is straight using density and width as X-axis, using the number of plies as Y-axis to establish density respectively Side's figure and width histogram;
S103. one is defined centered on current layer number, and size is straight in density and width respectively for the moving window of 2L*1 sizes Very dense value and width minimum are found in square figure, and layer where maximum and minimum is marked;
S104. power tower select first from bottom to up and meanwhile meet this layer of density local maxima and it is upper and lower two layers in the range of exist Position of the number of plies of width Local Minimum as characteristic face, is decomposed by boundary line of characteristic face, part more than characteristic face point Solve as tower head, the decomposed below characteristic face is tower body.
3. the power tower three-dimensional rebuilding method according to claim 1 based on LiDAR point cloud, it is characterised in that described Angle steel joint extract and comprised the following steps with segmentation in S2 steps:
S2011. current point and adjacent 2 points of angle are calculated, if angle is more than predetermined threshold df;Then retain current point, otherwise, Reject;
S2012. simplified by pipeline algorithm, delete most of non-angle point;
S2013. the angle point simplified to each layer, split according to it in the position at minimum enclosed rectangle center, by angle point It is divided into the subset of four direction.
4. the power tower three-dimensional rebuilding method according to claim 3 based on LiDAR point cloud, it is characterised in that described The contour line fitting based on RANSAC comprises the following steps in S2 steps:
S2021. two point P1 (x1, y1, z1) and P2 (x2, y2, z2) are randomly selected from the subset obtained by S1 steps, and calculate it The 3 d-line equation L at place;
S2022. for the every bit in subset, calculate it and arrive straight line L distance, if current point to straight line distance less than away from From threshold value Td, then it is available point, otherwise, is defined as Null Spot;And the ratio inner_ratio of available point is counted, wherein, Inner_ratio is that available point accounts for the ratio always counted;
S2023. stochastical sampling is repeated, until meeting that P is more than default maximum sampling less than default fiducial probability TP or sampling number M Number TN, choose parameter of the one group of minimum parameter of regression criterion as the subset 3 d-line equation;Wherein, P is sampling The probability of collection;
S2024. after four subsets are fitted, symmetry is added, coplanarity restrictive condition optimizes to model.
5. the power tower three-dimensional rebuilding method according to claim 3 based on LiDAR point cloud, it is characterised in that described Predetermined threshold df is 30 °.
6. the power tower three-dimensional rebuilding method according to claim 4 based on LiDAR point cloud, it is characterised in that described 3 d-line equation L calculation is:
<mrow> <mfrac> <mrow> <mi>x</mi> <mo>-</mo> <msub> <mi>x</mi> <mn>1</mn> </msub> </mrow> <mrow> <msub> <mi>x</mi> <mn>2</mn> </msub> <mo>-</mo> <msub> <mi>x</mi> <mn>1</mn> </msub> </mrow> </mfrac> <mo>=</mo> <mfrac> <mrow> <mi>y</mi> <mo>-</mo> <msub> <mi>y</mi> <mn>1</mn> </msub> </mrow> <mrow> <msub> <mi>y</mi> <mn>2</mn> </msub> <mo>-</mo> <msub> <mi>y</mi> <mn>1</mn> </msub> </mrow> </mfrac> <mo>=</mo> <mfrac> <mrow> <mi>z</mi> <mo>-</mo> <msub> <mi>z</mi> <mn>1</mn> </msub> </mrow> <mrow> <msub> <mi>z</mi> <mn>2</mn> </msub> <mo>-</mo> <msub> <mi>z</mi> <mn>1</mn> </msub> </mrow> </mfrac> <mo>.</mo> </mrow>
7. the power tower three-dimensional rebuilding method according to claim 4 based on LiDAR point cloud, it is characterised in that described The probability P of sampling subset meets following relationship:
P=1- (1-inner_ratiom)M,
Wherein, m is to build the minimum parameters number required for a judgment criteria.
8. the power tower three-dimensional rebuilding method according to claim 1 based on LiDAR point cloud, it is characterised in that described Identify that tower head fundamental type comprises the following steps in S3 steps:
S3011. it is two-value image by all tower head model conversations in tower head point cloud and model library to be identified;
S3012. the outline of all images is extracted, contour edge is sampled to obtain n sampled point point set P={ p1, p2, p3... ..pn};
S3013. the Shape context h of each sampled point is calculated according to Shape context algorithmi(k), calculation formula such as formula (1), k ={ 1,2 ..K }, K=M*N;
hi(k)={ q ≠ pi:(q-pi∈ bin (k) }, (1)
S3014. between the shape histogram of image to be matched and the shape histogram each pair point of each phantom images is calculated With cost, calculation formula such as formula (2), wherein, hi (k) is target P point pi shape histogram;Hj (k) is target Q point qi Shape histogram,
<mrow> <msub> <mi>C</mi> <mrow> <mi>i</mi> <mo>,</mo> <mi>j</mi> </mrow> </msub> <mo>=</mo> <mi>C</mi> <mrow> <mo>(</mo> <msub> <mi>p</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>q</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> <mo>=</mo> <mn>0.5</mn> <mo>*</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>K</mi> </munderover> <mfrac> <msup> <mrow> <mo>&amp;lsqb;</mo> <msub> <mi>h</mi> <mi>i</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>-</mo> <msub> <mi>h</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>&amp;rsqb;</mo> </mrow> <mn>2</mn> </msup> <mrow> <msub> <mi>h</mi> <mi>i</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>+</mo> <msub> <mi>h</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>,</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>2</mn> <mo>)</mo> </mrow> </mrow>
S3015. Point matching operation is carried out based on the cost matrix C being calculated, chooses the model shadow with minimum shape distance As being best matching result, the model is that tower head point cloud corresponds to model.
9. the power tower three-dimensional rebuilding method according to claim 1 based on LiDAR point cloud, it is characterised in that described Parameter optimization comprises the following steps in S3 steps:
S3021.Gibbs energy definitions, Gibbs energy are expressed as formula and are:
<mrow> <mi>h</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <msup> <mi>e</mi> <mrow> <mo>-</mo> <mi>u</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> </mrow> </msup> <mi>Z</mi> </mfrac> <mo>,</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>3</mn> <mo>)</mo> </mrow> </mrow>
Using alpha shape algorithms come the key point of extraction model and point cloud, distance um(xi) for each model key point to most The distance of nearly tower head point cloud key point, distance up(xi) it is distance of each tower head point cloud key point to nearest model key point, its In:
<mrow> <msub> <mi>u</mi> <mi>m</mi> </msub> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mn>1</mn> <mi>n</mi> </mfrac> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>0</mn> </mrow> <mi>n</mi> </munderover> <msub> <mi>u</mi> <mi>m</mi> </msub> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> <mo>,</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>4</mn> <mo>)</mo> </mrow> </mrow>
<mrow> <msub> <mi>u</mi> <mi>p</mi> </msub> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mn>1</mn> <mi>m</mi> </mfrac> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>0</mn> </mrow> <mi>m</mi> </munderover> <msub> <mi>u</mi> <mi>p</mi> </msub> <mrow> <mo>(</mo> <msub> <msup> <mi>x</mi> <mo>&amp;prime;</mo> </msup> <mi>i</mi> </msub> <mo>)</mo> </mrow> <mo>,</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>5</mn> <mo>)</mo> </mrow> </mrow>
Wherein,Average distance us of the degree of similarity u (x) by model to point cloudm(xi) and point cloud putting down to model Distance up(xi) composition, and u (x)=aum(x)+bup(x), a+b=1, a and b are respectively the weights of two distances;N closes for model Key point number, m are tower head point cloud key point number;
S3022.Metropolis-Hastings is sampled and simulated annealing, first, according to suggestion distribution q (x*| x) in search space A current parameter value x candidate parameter value x is provided in X*, then, the receiving of candidate parameter value is calculated according to calculation formula (6) Rate A (x, x*);If receptance A (x, x*) be more than default acceptance threshold Ta, then by candidate value x*Instead of currency x, otherwise, protect Stay currency x;
<mrow> <mi>A</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <msup> <mi>x</mi> <mo>*</mo> </msup> <mo>)</mo> </mrow> <mo>=</mo> <mi>m</mi> <mi>i</mi> <mi>n</mi> <mo>{</mo> <mn>1</mn> <mo>,</mo> <mfrac> <mrow> <mi>p</mi> <mrow> <mo>(</mo> <msup> <mi>x</mi> <mo>*</mo> </msup> <mo>)</mo> </mrow> <mi>q</mi> <mrow> <mo>(</mo> <msup> <mi>x</mi> <mo>*</mo> </msup> <mo>|</mo> <mi>x</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mi>p</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mi>q</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>|</mo> <msup> <mi>x</mi> <mo>*</mo> </msup> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>}</mo> <mo>.</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>6</mn> <mo>)</mo> </mrow> </mrow>
10. the power tower three-dimensional rebuilding method according to claim 9 based on LiDAR point cloud, it is characterised in that described Acceptance threshold Ta=0.9+0.1*random (0,1), random (0,1) are that random sampling result is carried out in the range of 0~1.
CN201710853206.2A 2017-09-20 2017-09-20 Three-dimensional reconstruction method for power tower based on LiDAR point cloud Active CN107610223B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710853206.2A CN107610223B (en) 2017-09-20 2017-09-20 Three-dimensional reconstruction method for power tower based on LiDAR point cloud

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710853206.2A CN107610223B (en) 2017-09-20 2017-09-20 Three-dimensional reconstruction method for power tower based on LiDAR point cloud

Publications (2)

Publication Number Publication Date
CN107610223A true CN107610223A (en) 2018-01-19
CN107610223B CN107610223B (en) 2020-09-25

Family

ID=61061715

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710853206.2A Active CN107610223B (en) 2017-09-20 2017-09-20 Three-dimensional reconstruction method for power tower based on LiDAR point cloud

Country Status (1)

Country Link
CN (1) CN107610223B (en)

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108830933A (en) * 2018-06-26 2018-11-16 广东电网有限责任公司 A kind of electric force pole tower tower body method for reconstructing, system, medium and equipment
CN109341667A (en) * 2018-11-29 2019-02-15 武汉汇卓航科技有限公司 A kind of three-dimensional power line measurement method based on region decision selection
CN109872384A (en) * 2018-12-29 2019-06-11 中国科学院遥感与数字地球研究所 A kind of shaft tower automation modeling method based on airborne LIDAR point cloud data
CN110136260A (en) * 2019-04-24 2019-08-16 广州供电局有限公司 Electric power line pole tower and its implementation based on laser point cloud full feature model library
CN110766750A (en) * 2019-09-09 2020-02-07 浙江大学 Mechanical arm visual servo target identification method based on local contour
CN111126211A (en) * 2019-12-13 2020-05-08 北京四维图新科技股份有限公司 Label identification method and device and electronic equipment
CN111739164A (en) * 2019-03-19 2020-10-02 北京京东尚科信息技术有限公司 Box body modeling method and device, robot picking system, electronic device and medium
CN112215867A (en) * 2020-10-23 2021-01-12 劢微机器人科技(深圳)有限公司 Point cloud data segmentation method, point cloud data segmentation equipment, storage medium and device
CN112698303A (en) * 2020-12-23 2021-04-23 国网电力科学研究院武汉南瑞有限责任公司 Method and system for measuring point cloud tower inclination parameters based on unmanned aerial vehicle laser radar
CN112767352A (en) * 2021-01-19 2021-05-07 国网通用航空有限公司 High-voltage line tower automatic fine extraction method for airborne LiDAR power inspection
CN112884823A (en) * 2021-01-13 2021-06-01 上海建工四建集团有限公司 Running rule simulation point cloud selection and calculation method for actual measurement application
CN113076870A (en) * 2021-03-31 2021-07-06 广东电网有限责任公司 Tower identification method and system based on laser point cloud
CN113345089A (en) * 2021-05-31 2021-09-03 西北农林科技大学 Regularization modeling method based on power tower point cloud
CN113554595A (en) * 2021-06-21 2021-10-26 国网山西省电力公司电力科学研究院 Unmanned aerial vehicle laser radar point cloud tower head deformation detection device and method
CN114418194A (en) * 2021-12-29 2022-04-29 广东电网有限责任公司 Tower damage prediction method and device based on data driving and model driving

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104866840A (en) * 2015-06-04 2015-08-26 广东中城规划设计有限公司 Method for recognizing overhead power transmission line from airborne laser point cloud data
CN106157361A (en) * 2016-05-31 2016-11-23 中国科学院遥感与数字地球研究所 A kind of multiple fission conductor full-automatic three-dimensional method for reconstructing based on LiDAR point cloud
US20170053438A1 (en) * 2014-06-13 2017-02-23 Shenzhen Institutes Of Advanced Technology Chinese Academy Of Sciences Method and system for reconstructing a three-dimensional model of point clouds
CN106683132A (en) * 2015-11-10 2017-05-17 星际空间(天津)科技发展有限公司 High-precision three-dimensional city modeling method
CN107102339A (en) * 2017-04-28 2017-08-29 上海华测导航技术股份有限公司 A kind of automatic generation method of above-ground route data

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170053438A1 (en) * 2014-06-13 2017-02-23 Shenzhen Institutes Of Advanced Technology Chinese Academy Of Sciences Method and system for reconstructing a three-dimensional model of point clouds
CN104866840A (en) * 2015-06-04 2015-08-26 广东中城规划设计有限公司 Method for recognizing overhead power transmission line from airborne laser point cloud data
CN106683132A (en) * 2015-11-10 2017-05-17 星际空间(天津)科技发展有限公司 High-precision three-dimensional city modeling method
CN106157361A (en) * 2016-05-31 2016-11-23 中国科学院遥感与数字地球研究所 A kind of multiple fission conductor full-automatic three-dimensional method for reconstructing based on LiDAR point cloud
CN107102339A (en) * 2017-04-28 2017-08-29 上海华测导航技术股份有限公司 A kind of automatic generation method of above-ground route data

Cited By (22)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108830933A (en) * 2018-06-26 2018-11-16 广东电网有限责任公司 A kind of electric force pole tower tower body method for reconstructing, system, medium and equipment
CN108830933B (en) * 2018-06-26 2022-04-19 广东电网有限责任公司 Method, system, medium and equipment for rebuilding tower body of electric tower
CN109341667A (en) * 2018-11-29 2019-02-15 武汉汇卓航科技有限公司 A kind of three-dimensional power line measurement method based on region decision selection
CN109341667B (en) * 2018-11-29 2020-12-18 武汉汇卓航科技有限公司 Three-dimensional power line measuring method based on region judgment selection
CN109872384A (en) * 2018-12-29 2019-06-11 中国科学院遥感与数字地球研究所 A kind of shaft tower automation modeling method based on airborne LIDAR point cloud data
CN111739164A (en) * 2019-03-19 2020-10-02 北京京东尚科信息技术有限公司 Box body modeling method and device, robot picking system, electronic device and medium
CN110136260A (en) * 2019-04-24 2019-08-16 广州供电局有限公司 Electric power line pole tower and its implementation based on laser point cloud full feature model library
CN110766750B (en) * 2019-09-09 2022-04-05 浙江大学 Mechanical arm visual servo target identification method based on local contour
CN110766750A (en) * 2019-09-09 2020-02-07 浙江大学 Mechanical arm visual servo target identification method based on local contour
CN111126211A (en) * 2019-12-13 2020-05-08 北京四维图新科技股份有限公司 Label identification method and device and electronic equipment
CN111126211B (en) * 2019-12-13 2023-08-29 北京四维图新科技股份有限公司 Label identification method and device and electronic equipment
CN112215867A (en) * 2020-10-23 2021-01-12 劢微机器人科技(深圳)有限公司 Point cloud data segmentation method, point cloud data segmentation equipment, storage medium and device
CN112698303A (en) * 2020-12-23 2021-04-23 国网电力科学研究院武汉南瑞有限责任公司 Method and system for measuring point cloud tower inclination parameters based on unmanned aerial vehicle laser radar
CN112884823A (en) * 2021-01-13 2021-06-01 上海建工四建集团有限公司 Running rule simulation point cloud selection and calculation method for actual measurement application
CN112884823B (en) * 2021-01-13 2023-09-22 上海建工四建集团有限公司 Guiding rule simulation point cloud selection and calculation method for actual measurement actual quantity application
CN112767352A (en) * 2021-01-19 2021-05-07 国网通用航空有限公司 High-voltage line tower automatic fine extraction method for airborne LiDAR power inspection
CN112767352B (en) * 2021-01-19 2023-07-07 国网电力空间技术有限公司 Automatic refined extraction method for high-voltage line tower for airborne LiDAR power inspection
CN113076870A (en) * 2021-03-31 2021-07-06 广东电网有限责任公司 Tower identification method and system based on laser point cloud
CN113345089A (en) * 2021-05-31 2021-09-03 西北农林科技大学 Regularization modeling method based on power tower point cloud
CN113345089B (en) * 2021-05-31 2023-06-23 西北农林科技大学 Regularized modeling method based on power tower point cloud
CN113554595A (en) * 2021-06-21 2021-10-26 国网山西省电力公司电力科学研究院 Unmanned aerial vehicle laser radar point cloud tower head deformation detection device and method
CN114418194A (en) * 2021-12-29 2022-04-29 广东电网有限责任公司 Tower damage prediction method and device based on data driving and model driving

Also Published As

Publication number Publication date
CN107610223B (en) 2020-09-25

Similar Documents

Publication Publication Date Title
CN107610223A (en) Power tower three-dimensional rebuilding method based on LiDAR point cloud
Liu et al. Adaptive douglas-peucker algorithm with automatic thresholding for AIS-based vessel trajectory compression
CN110163439A (en) A kind of city size taxi trajectory predictions method based on attention mechanism
CN102279929B (en) Remote-sensing artificial ground object identifying method based on semantic tree model of object
CN102496034B (en) High-spatial resolution remote-sensing image bag-of-word classification method based on linear words
CN107392875A (en) A kind of cloud data denoising method based on the division of k neighbours domain
CN105654483B (en) The full-automatic method for registering of three-dimensional point cloud
CN104318622B (en) Triangular mesh modeling method of indoor scene inhomogeneous three dimension point cloud data
CN110322453A (en) 3D point cloud semantic segmentation method based on position attention and auxiliary network
CN104091321A (en) Multi-level-point-set characteristic extraction method applicable to ground laser radar point cloud classification
CN103533501A (en) Geofence generating method
CN110533048A (en) The realization method and system of combination semantic hierarchies link model based on panoramic field scene perception
CN106570468A (en) Method for reconstructing LiDAR original point cloud building contour
CN107330901B (en) Skeleton-based object component decomposition method
CN103077555B (en) The automatic marking method that a kind of three-dimensional model is formed
CN102622753B (en) Semi-supervised spectral clustering synthetic aperture radar (SAR) image segmentation method based on density reachable measure
CN105046694A (en) Quick point cloud registration method based on curved surface fitting coefficient features
CN103559705B (en) Computer method for comparing similarity of different plant forms
CN106548484A (en) Product model dispersion point cloud Boundary characteristic extraction method based on two-dimentional convex closure
CN101937453A (en) 3-D model (three-dimensional model) search method based on weighted topology constraint comparison
CN106886980A (en) A kind of enhanced method of point cloud density based on three-dimensional laser radar target identification
CN102314610B (en) Object-oriented image clustering method based on probabilistic latent semantic analysis (PLSA) model
CN104537353A (en) Three-dimensional face age classifying device and method based on three-dimensional point cloud
CN104036229A (en) Regression-based active appearance model initialization method
CN102945553A (en) Remote sensing image partition method based on automatic difference clustering algorithm

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