CN109147038A - Pipeline three-dimensional modeling method based on three-dimensional point cloud processing - Google Patents
Pipeline three-dimensional modeling method based on three-dimensional point cloud processing Download PDFInfo
- Publication number
- CN109147038A CN109147038A CN201810957920.0A CN201810957920A CN109147038A CN 109147038 A CN109147038 A CN 109147038A CN 201810957920 A CN201810957920 A CN 201810957920A CN 109147038 A CN109147038 A CN 109147038A
- Authority
- CN
- China
- Prior art keywords
- point
- point cloud
- pipeline
- dimensional
- label
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000012545 processing Methods 0.000 title claims description 13
- 238000005070 sampling Methods 0.000 claims abstract description 10
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000010276 construction Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000005259 measurement Methods 0.000 claims description 7
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 claims description 6
- 238000004519 manufacturing process Methods 0.000 claims description 5
- 230000004927 fusion Effects 0.000 claims description 4
- 239000000126 substance Substances 0.000 claims description 4
- 238000013519 translation Methods 0.000 claims description 4
- 230000000007 visual effect Effects 0.000 claims description 4
- 239000003345 natural gas Substances 0.000 claims description 3
- 238000011017 operating method Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 230000007246 mechanism Effects 0.000 claims description 2
- 238000005457 optimization Methods 0.000 claims description 2
- 230000009466 transformation Effects 0.000 claims description 2
- 238000012217 deletion Methods 0.000 claims 1
- 230000037430 deletion Effects 0.000 claims 1
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 238000001514 detection method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012356 Product development Methods 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000000280 densification Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/162—Segmentation; Edge detection involving graph-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20072—Graph-based image processing
Abstract
The invention discloses the pipeline three-dimensional modeling methods handled based on three-dimensional point cloud comprising, data prediction is carried out to input point cloud: removing the spot noise that peels off, estimates the normal vector of each point；The cylinder parameter model collection of candidate pipeline is obtained by the method for fitting circle cylindrical equation using stochastical sampling consistency algorithm；It using a graph model description point cloud, and constructs and marks energy function corresponding to point cloud using candidate family collection, cut algorithmic minimizing energy function using improved figure, obtain all corresponding cylindrical surface parameters of pipeline and its respective interior point set in a cloud；Using the syntople of viewpoint, the pipeline detected in the point cloud under different points of view is registrated using ICP algorithm, finally rebuilds the threedimensional model of entire factory's pipeline.The method proposed can be applied to the three-dimensional point cloud of plurality of devices acquisition, have the characteristics that flexibly easy-to-use；Model parameter estimation is accurate, and the complete three-dimensional model of factory's pipeline may be implemented.
Description
Technical field
The present invention relates to three-dimensional point cloud modeling technique fields, more particularly to a kind of pipeline three based on three-dimensional point cloud processing
Tie up modeling method.
Background technique
In recent years, as country carries forward vigorously intelligent plant construction, intelligent preferable factory applies in progress Pipe installing
Technology realization product development all is defined using based on model before work, completely expresses pipeline with an integrated three-dimensional entity model
Design information and manufacture information (product structure, three-dimensional dimension etc.).But for existing factory, in intelligent construction, need
Upgrading is carried out to it, for outmoded chemical manufacturing plant pipes equipment, lack the collection and analysis of comprehensive status data,
Existing factory's pipe-line equipment objective details in time how are obtained, are the main problem faced.The three-dimensional modeling energy of pipeline
The geometric parameter of enough intuitive description pipelines is conducive to the distribution situation and damage feelings of true reflection factory's pipeline comprehensively
Condition, therefore three-dimensional modeling is studied, the management and application for factory's pipeline data have actual meaning.Factory has environment
The characteristics of complexity, conduit volume is huge, bad environments.Existing factory's pipe detection is mostly by being accomplished manually, by naked eyes
Heavy workload is detected, precision is low, is easy to fail to judge.This, which results in data record, error, the initial parameter inaccuracy of pipeline.It utilizes
Sensor carries out pipe detection and has then liberated human resources, will not make one to face potential danger.Currently, utilizing sensor tube
There are mainly two types of detection modes: one is two-dimensional figure, image, another kind is a cloud, traditional image recording mode due to
The steric informations such as phase are lost, so that multi-angle observation and analysis and research can not be carried out to object；And it puts cloud and is then widely present
The every field of real world can truly record the various 3 D stereo information of body surface.For adopting for three-dimensional point cloud
Collection, the 3 D laser scanning equipment that international advanced means are all made of profession are completed.But such equipment there are equipment prices extremely
Valuableness, operating method is cumbersome, is difficult to complete professional, is not suitable for promoting and applying.It is obtained using Kinect sensor
The three-dimensional information of environment or object operates more easy, equipment price compared with the large-scale three dimensional Laser Scanning Equipment of profession
Far below professional equipment.This is a kind of non-contact scanning technology, without touching testee, so that it may which quick obtaining is true
The three-dimensional coordinate of scene can be widely applied to dangerous fields of measurement.Since its volume is less than normal, it is quickly real that unmanned plane can be carried
When obtain pipeline multiple views information, avoid as blocking and the influence caused by three-dimensional modeling of the missing of point cloud data.
Summary of the invention
There are aiming at the problem that, the technical problem to be solved by the present invention is based on three-dimensional point cloud processing how to realize pipe
Road identification and reconstructing three-dimensional model.
In order to solve the above-mentioned technical problems, the present invention provides the pipeline three-dimensional modeling method handled based on three-dimensional point cloud,
Method includes the following steps:
Step 1, the three-dimensional point cloud that chemical manufacturing plant pipes or natural gas transportation pipeline are obtained using depth transducer, are counted
Data preprocess removes the spot noise that peels off using statistical zero-knowledge, obtains pretreated three dimensional point cloud.
Step 2, with fitting process to each point estimations vector in pretreated three dimensional point cloud.
Step 3 describes the obtained sextuple array constituted with position and normal vector using stochastical sampling consistency algorithm
Three dimensional point cloud sampled, and be fitted to obtain the cylinder parameter mould of candidate pipeline by the method for fitting circle cylindrical equation
Type collection.
Step 4, construction energy function, using the cylinder parameter model collection of the candidate pipeline fitted in step 3 as energy
The initial labels of function cut algorithmic minimizing energy function using improved figure and determine cylindrical surface global best estimates in point cloud,
To detect the corresponding cylindrical surface of multiple pipelines in three-dimensional point cloud parameter and interior point set.
Step 5, by the collected three dimensional point cloud operating procedure 1 of different points of view to step 4, respectively obtain different perspectives
Then lower pipeline reconstructed results are registrated it using ICP algorithm using the syntople of viewpoint two-by-two, are spliced, final
To the threedimensional model of entire factory's pipeline.
Method is improved the present invention has the advantages that this method makes statistical zero-knowledge remove outlier sound for the prior art
The accuracy of vector estimation obtains the cylinder parameter model collection of multiple candidate pipelines using the position and normal vector of point, will be traditional
Figure cut algorithm and be transplanted to the segmentation of three-dimensional point cloud from two dimensional image segmentation, cut algorithmic minimizing energy letter using improved figure
Number, fast and accurately detects the corresponding cylindrical surface parameter of multiple pipelines and the Models Sets in three-dimensional point cloud, uses ICP algorithm
To multiple views point cloud registering, splicing, the threedimensional model of entire factory's pipeline is finally obtained.
Detailed description of the invention
Fig. 1 is the flow chart of the embodiment of the present invention.
Fig. 2 is pipeline point cloud registering, splices flow chart.
Specific embodiment
The specific technical solution of the present invention is specifically described in conjunction with the accompanying drawings and embodiments.
Refering to Figure 1, the present invention provides a kind of pipeline three-dimensional modeling methods based on three-dimensional point cloud processing, including
Following steps:
Step 1, the three-dimensional point cloud that chemical manufacturing plant pipes or natural gas transportation pipeline are obtained using depth transducer, are counted
Data preprocess removes the spot noise that peels off using statistical zero-knowledge.
For any point in the pipeline point cloud of scanning, take the point k neighborhood spatially, calculate each point to it
The average distance of all k point of proximity.Assuming that obtain the result is that a Gaussian Profile, shape are determined by mean value and standard deviation,
Point of the average distance except critical field (being defined by global distance average and variance), is defined as outlier, and from number
It is deleted according to concentrating.
Step 2, with fitting process to point estimations vector each in point cloud data.
1) for any point p in the pipeline point cloud of scanningi, take the point spatially radius be r neighborhood, then this
Point in neighborhood is k neighbour's point set of the point.The value of radius of neighbourhood r and the value of k and acquisition equipment are between pipeline
Distance and put cloud sparse degree it is related, being usually 30 or so to setting k value radius should 0.06m-0.20m it
Between.
It the use of the distance between acquisition equipment to pipeline is 1.2m-4m, vertical visual field is to be up to 60 degree, and horizontal field of view is most
Greatly 70 degree, according to acquisition distance and the resolution ratio of acquisition equipment, it is estimated that the big densification of collected pipeline point cloud
Degree, if acquisition equipment at a distance from pipeline farther out, point cloud it is sparse, then the biggish radius of neighbourhood is set, if being closer, point cloud it is close
Collection, then be arranged the lesser radius of neighbourhood.Radius should be between 0.06m-0.20m being 30 or so to setting k value.
2) its covariance matrix is calculated by obtained k neighbour's point set, it is as follows that this puts corresponding covariance matrix C:
Herein, k is point piThe number of Neighbor Points,Indicate the three-dimensional mass center of nearest neighbors, λjIt is covariance matrix
J-th of characteristic value,It is j-th of feature vector.Feature vector corresponding to the minimal eigenvalue of covariance matrix is taken, as
Point piNormal vector ni。
Step 3 describes the obtained sextuple array constituted with position and normal vector using stochastical sampling consistency algorithm
Three-dimensional point cloud sampled, and the cylinder parameter model collection of candidate pipeline is obtained by the method for fitting circle cylindrical equation.
1) two point p are randomly selected1And p2And corresponding normal direction n1And n2, it is assumed that the two are put on the surface of cylinder, that
Obtain the axial n=n of cylinder1×n2。
2) by p1And p2Straight line where two o'clock, p1+t·n1=0 and p2+t·n2=0 projects on plane nx=0, and two
The intersection point of the projection of straight line is exactly the point p on cylinder axis0。
3) the radius r that cylinder is arranged is p1To point p0Distance.
The parameter for obtaining cylinder is a point p on axis0=(x0, y0, zo), axial unit normal vector n=(nx, ny, nz), circle
Column radius is r.Cylinder parameter is verified using the mechanism of stochastical sampling consistency.The process is repeated, is obtained by multiple circles
The candidate cylinder parameter model collection that column is constituted.
Step 4, construction energy function, using the cylinder parameter model collection of the candidate pipeline fitted in step 3 as energy
The initial labels of function cut algorithmic minimizing energy function using improved figure and determine cylindrical surface global best estimates in point cloud,
To detect the corresponding cylindrical surface of multiple pipelines in three-dimensional point cloud parameter and interior point set.
1) describe three-dimensional point cloud using non-directed graph: proposed by the invention based on the algorithm that figure is cut is to regard three-dimensional point cloud as
One non-directed graph G=(V, Σ), Point Set V indicate the set that all points are constituted in point cloud, and Σ indicates consecutive points phase in point cloud
The set on the side for the composition that connects.Use the method construct neighborhood of a point system of construction kd-tree, the side (V of weighted graphi, Vj)∈
The weight of Σ represents the similarity between consecutive points, this similarity can be gray value, position, brightness a little etc..The present invention
The difference of three dimensional local information is used to realize measurement.
2) construct energy function: point cloud in cylindrical surface global best estimates problem be exactly be in three-dimensional point cloud it is each point mark
Note one unique label fi.The label value of point constitutes tally set F={ f0, f1, fM, wherein fi(i ∈ [1M]) is represented
The point being marked is the point on i-th of cylindrical surface, f0Point in non-cylindrical is represented, the number M of label is improved by proposing
Figure cuts algorithm and automatically determines.To estimate the cylinder in a cloud under global optimum's meaning, following energy function is constructed:
Wherein, first item is unitary item DpResidual error of the () metric point on the corresponding cylinder model of label marked,
Section 2 is binary item V () measurement flatness, i.e., two neighboring point is labeled as the cost of different cylinders, and N is consecutive points
Set, Section 3 L () are tag energy, measure the number for using label in annotation process, prevent over-fitting.
For unitary item Dp(f (p)), by calculating point ppIt is indicated to the distance that label is the candidate cylinder of f (p), f (p) table
Show to point p mark as a result, f (p) ∈ F.I.e.
Dp(f (p))=D-r
Wherein D indicates point ppTo the distance of cylinder axis, are as follows:
Binary item V (f (p), f (q)) in formula, also referred to as smooth item are described using Potts model, it may be assumed that
V (f (p), f (q))=λ ωpq·δ(f(p)≠f(q))
Wherein δ () is indicator function, i.e.,
ωpqFor penalty coefficient, the present invention uses gaussian kernel function:F (p) is indicated to point p
Mark as a result, f (p) ∈ F, f (q) indicate to mark point q as a result, f (q) ∈ F.As point ppPoint p in neighborhoodq, it is noted as
With ppLabel difference when, corresponding cost function is not zero.σ value range is 3~5, the weight coefficient λ value model of smooth item
Enclose is 0.5~2.5.
Tag energy item L (F) in formula, the number of label after measurement segmentation, avoids generating over-fitting, prevents to have
A few data point of noise is accidentally fitted to cylinder, i.e.,
L (F)=hM+ η
Wherein M is the number of label, and h is the weight coefficient of tag energy item, the value usually between 10~20.η is 10
Exterior point ratio again, exterior point ratio refer to that those are not marked the ratio that the point of label accounts for entire point cloud, can effectively guide
The determination of label number.
3) algorithmic minimizing energy function is cut by figure.It utilizes proposed extension α-expansion figure to cut algorithm to find
Optimal tally setFor each point in pipeline point cloud, using the label in tag set to it
Be labeled so that energy function constantly reduces, i.e., constantly in optimization point cloud each point label, until energy function no longer
Reduce, determines global best estimates in this way, obtainA label corresponds in the three-dimensional point cloud detectedA pipe
The cylindrical surface parameter in road and its respectively interior collection.
Wherein, optimal tally set is determined by following steps, specifically:
1) the cylinder parameter model collection for the candidate pipeline for fitting step 3 is as the initial labels collection F of energy function；
2) each point in three-dimensional point cloud is labeled using extension α-expansion algorithm, finds satisfactionOptimal tally setWith corresponding cylindrical surface parameter and interior point set.
3) if label isThe corresponding interior number of model be less than a certain threshold value, then by its model parameter
Compared with other model parameters, if being with labelThe rotation angle of model parameter, the small Mr. Yu of translation distance
One threshold value then merges the corresponding interior point of two models, labelWithFusion is denoted asLabel numberSubtract 1, it is on the contrary by this
The corresponding interior point of model is set as exterior pointSimilarly, the label in tally set is traversed, realizes the fusion of label and the essence of model
Refinement.
4) to label obtained in the previous stepInterior point set sampled using stochastical sampling consistency algorithm, and by quasi-
The method for closing cylindrical surface equation obtains the cylinder parameter model collection of candidate pipelineIt willMark after merging fining with previous step
Label collectionIt asks simultaneously, i.e.,As the initial labels collection of energy function, iteration above-mentioned steps 2) -4), until energy no longer
Reduce.Obtain the tally set of global optimumEach label(whereinThe cylindrical surface parameter of corresponding multiple pipelines
And interior point set.
The collected three-dimensional point cloud operation above-mentioned steps 1 of different points of view to step 4, are respectively obtained different points of view by step 5
Then lower pipeline reconstructed results are registrated it using ICP algorithm using the syntople of viewpoint two-by-two, are spliced, final
To the threedimensional model of entire factory's pipeline.Please refer to shown in Fig. 2 the specific steps are
1) wherein cloud is denoted as Q as source point cloud for order, and taking its abutment points cloud P using the syntople at visual angle is target point
Cloud,
2) n point is taken to be denoted as P at random in target point cloud P1(P1∈P)；
3) method for using kd-tree to search in source point cloud Q, according to | | Q1-P1| | the criterion of=min finds corresponding point
Collect Q1(Q1∈Q)；
4) match parameter spin matrix R and translation matrix t and P are calculated1To Q1Average distance
5) to P1Point set is coordinately transformed to obtain P1' and seek P1' arrive Q1Average distance As error function；
6) whether error in judgement restrains dk-dk+1< τ, τ are setting value and τ > 0, then restrain, otherwise jump to step 2).
7) target point cloud is translated and is rotated in τ by error convergence, new transformation point set P' and source point cloud Q after being registrated
It merges, new point cloud is denoted as Q, then the successively point cloud adjacent with original point cloud Q and the point cloud progress adjacent with cloud P again
Registration, fusion.
8) syntople for utilizing viewpoint, is successively registrated all of its neighbor point cloud, finally obtains the pipe of entire factory
The threedimensional model in road.
The present invention propose it is a kind of based on three-dimensional point cloud processing pipeline three-dimensional modeling method, using statistical zero-knowledge filter out from
Group's spot noise, the cylinder parameter of candidate pipeline is fitted according to the location information of point and normal vector, uses the improvement cut based on figure
Figure cuts algorithm, minimizes energy function, determines global best estimates, obtain the optimized parameter of pipeline cylinder, finally calculated using ICP
Method registration splices to viewpoint pipeline model, finally obtains the threedimensional model of entire factory's pipeline.With save the cost, convenient for answering
With model parameter estimation is accurate, the characteristics of can efficiently obtaining the threedimensional model of factory's pipeline comprehensively.
The above content is the further descriptions for combining optimum implementation to be the present invention, it cannot be assumed that of the invention
Specific implementation be only limited to these explanations.Those skilled in the art should understand that being limited not departing from by the appended claims
In the case where fixed, it can carry out various modifications in detail, all should be considered as protection scope of the present invention.
Claims (6)
1. the pipeline three-dimensional modeling method based on three-dimensional point cloud processing, it is characterised in that: method includes the following steps:
Step 1, the three-dimensional point cloud that chemical manufacturing plant pipes or natural gas transportation pipeline are obtained using depth transducer, it is pre- to carry out data
Processing removes the spot noise that peels off using statistical zero-knowledge, obtains pretreated three dimensional point cloud；
Step 2, with fitting process to each point estimations vector in pretreated three dimensional point cloud；
Step 3, the obtained sextuple array constituted with position and normal vector is described using stochastical sampling consistency algorithm three
Dimension point cloud data is sampled, and is fitted to obtain the cylinder parameter model of candidate pipeline by the method for fitting circle cylindrical equation
Collection；
Step 4, construction energy function, using the cylinder parameter model collection of the candidate pipeline fitted in step 3 as energy function
Initial labels, using improved figure cut algorithmic minimizing energy function determine point cloud in cylindrical surface global best estimates, thus
Detect the corresponding cylindrical surface of multiple pipelines in three-dimensional point cloud parameter and interior point set；
Step 5, by the collected three dimensional point cloud operating procedure 1 of different points of view to step 4, respectively obtain different perspectives down tube
Then road reconstructed results are registrated it using ICP algorithm using the syntople of viewpoint two-by-two, are spliced, finally obtained whole
The threedimensional model of a factory's pipeline.
2. the pipeline three-dimensional modeling method according to claim 1 based on three-dimensional point cloud processing, it is characterised in that: this method
Step 1 in
For any point in the pipeline point cloud of scanning, take the point k neighborhood spatially, calculate each point to all of its
The average distance of k point of proximity；Assuming that obtain the result is that a Gaussian Profile, shape are determined by mean value and standard deviation, it is average
Distance is defined as outlier in the point that critical field is except being defined by global distance average and variance, and from data set
Middle deletion.
3. the pipeline three-dimensional modeling method according to claim 1 based on three-dimensional point cloud processing, it is characterised in that: step 2,
With fitting process to point estimations vector each in point cloud data；
1) for any point p in the pipeline point cloud of scanningi, take the point the neighborhood that spatially radius is r, then this neighborhood
In point be the point k neighbour's point set；The value of radius of neighbourhood r and the value of k and acquisition equipment between pipeline away from
From and point cloud sparse degree it is related, radius should be between 0.06m-0.20m being usually 30 or so to setting k value；
It the use of the distance between acquisition equipment to pipeline is 1.2m-4m, vertical visual field is to be up to 60 degree, and horizontal field of view is up to
70 degree, according to acquisition distance and the resolution ratio of acquisition equipment, it is estimated that the big consistency of collected pipeline point cloud, if
At a distance from pipeline farther out, point cloud is sparse for acquisition equipment, then the biggish radius of neighbourhood is arranged, if being closer, point cloud is intensive, then
The lesser radius of neighbourhood is set；Radius should be between 0.06m-0.20m being 30 or so to setting k value；
2) its covariance matrix is calculated by obtained k neighbour's point set, it is as follows that this puts corresponding covariance matrix C:
Herein, k is point piThe number of Neighbor Points,Indicate the three-dimensional mass center of nearest neighbors, λjIt is the jth of covariance matrix
A characteristic value,It is j-th of feature vector；Feature vector corresponding to the minimal eigenvalue of covariance matrix is taken, as point pi
Normal vector ni。
4. the pipeline three-dimensional modeling method according to claim 1 based on three-dimensional point cloud processing, it is characterised in that: step 3,
It is carried out using three-dimensional point cloud of the stochastical sampling consistency algorithm to the obtained sextuple array description constituted with position and normal vector
Sampling, and the cylinder parameter model collection of candidate pipeline is obtained by the method for fitting circle cylindrical equation；
1) two point p are randomly selected1And p2And corresponding normal direction n1And n2, it is assumed that the two points are on the surface of cylinder, then obtaining
To the axial n=n of cylinder1×n2；
2) by p1And p2Straight line where two o'clock, p1+t·n1=0 and p2+t·n2=0 projects on plane nx=0, and two straight
The intersection point of the projection of line is exactly the point p on cylinder axis0；
3) the radius r that cylinder is arranged is p1To point p0Distance；
The parameter for obtaining cylinder is a point p on axis0=(x0, y0, zo), axial unit normal vector n=(nx, ny, nz), cylindrical radius
For r；Cylinder parameter is verified using the mechanism of stochastical sampling consistency；The process is repeated, obtains being made of multiple cylinders
Candidate cylinder parameter model collection.
5. the pipeline three-dimensional modeling method according to claim 1 based on three-dimensional point cloud processing, it is characterised in that: step 4,
Energy function is constructed, using the cylinder parameter model collection of the candidate pipeline fitted in step 3 as the initial labels of energy function,
Algorithmic minimizing energy function is cut using improved figure and determines cylindrical surface global best estimates in point cloud, to detect three-dimensional point
The parameter on the corresponding cylindrical surface of multiple pipelines in cloud and interior point set；
1) three-dimensional point cloud is described using non-directed graph: based on the algorithm that figure is cut be three-dimensional point cloud is regarded as a non-directed graph G=(V,
Σ), Point Set V indicates the set that all points are constituted in point cloud, and Σ indicates the side that consecutive points are interconnected to constitute in point cloud
Set；Use the method construct neighborhood of a point system of construction kd-tree, the side (V of weighted graphi, Vj) ∈ Σ weight represent it is adjacent
Similarity between point, this similarity can be gray value a little, position, brightness；Come using the difference of three dimensional local information
Realize measurement；
2) construct energy function: point cloud in cylindrical surface global best estimates problem be exactly be in three-dimensional point cloud it is each point mark one
A unique label fi；The label value of point constitutes tally set F={ f0, f1... fM, wherein fiWhat (i ∈ [1M]) representative was marked
Point is the point on i-th of cylindrical surface, f0Point in non-cylindrical is represented, the number M of label cuts algorithm by proposed improved figure
It automatically determines；To estimate the cylinder in a cloud under global optimum's meaning, following energy function is constructed:
Wherein, first item is unitary item DpResidual error of the () metric point on the corresponding cylinder model of label marked, Section 2
For binary item V () measurement flatness, i.e., two neighboring point is labeled as the cost of different cylinders, and N is the set of consecutive points,
Section 3 L () is tag energy, measures the number for using label in annotation process, prevents over-fitting；
For unitary item Dp(f (p)), by calculating point ppIt is indicated to the distance that label is the candidate cylinder of f (p), f (p) expression pair
Point p mark as a result, f (p) ∈ F；I.e.
Dp(f (p))=D-r
Wherein D indicates point ppTo the distance of cylinder axis, are as follows:
Binary item V (f (p), f (q)) in formula, also referred to as smooth item are described using Potts model, it may be assumed that
V (f (p), f (q))=λ ωpq·δ(f(p)≠f(q))
Wherein δ () is indicator function, i.e.,
ωpqFor penalty coefficient, gaussian kernel function is used:F (p) indicate to point p mark as a result,
F (p) ∈ F, f (q) indicate to mark point q as a result, f (q) ∈ F；As point ppPoint p in neighborhoodq, it is noted as and ppLabel
When different, corresponding cost function is not zero；σ value range is 3~5, the weight coefficient λ value range of smooth item is 0.5~
2.5；
Tag energy item L (F) in formula, the number of label after measurement segmentation, avoids generating over-fitting, prevents having noise
A few data point be accidentally fitted to cylinder, i.e.,
L (F)=hM+ η
Wherein M is the number of label, and h is the weight coefficient of tag energy item, the value usually between 10~20；η is 10 times
Exterior point ratio, exterior point ratio refer to that those are not marked the ratio that the point of label accounts for entire point cloud, can effectively guide tag
The determination of number；
3) algorithmic minimizing energy function is cut by figure；Utilize proposed extension α-expansion figure cut algorithm find it is optimal
Tally setFor each point in pipeline point cloud, it is carried out using the label in tag set
Mark, so that energy function constantly reduces, i.e., the label of each point in constantly optimization point cloud, until energy function no longer subtracts
It is small, global best estimates are determined in this way, are obtainedA label corresponds in the three-dimensional point cloud detectedA pipeline
Cylindrical surface parameter and its respectively in collection；
Wherein, optimal tally set is determined by following steps, specifically:
1) the cylinder parameter model collection for the candidate pipeline for fitting step 3 is as the initial labels collection F of energy function；
2) each point in three-dimensional point cloud is labeled using extension α-expansion algorithm, finds satisfactionOptimal tally setWith corresponding cylindrical surface parameter and interior point set；
3) if label isThe corresponding interior number of model be less than a certain threshold value, then by its model parameter and its
Its model parameter compares, if being with labelThe rotation angle of model parameter, translation distance be less than a certain threshold
Value then merges the corresponding interior point of two models, labelWithFusion is denoted asLabel numberSubtract 1, it is on the contrary by this model
Corresponding interior point is set as exterior pointSimilarly, the label in tally set is traversed, realizes the fusion of label and the fining of model；
4) to label obtained in the previous stepInterior point set sampled using stochastical sampling consistency algorithm, and pass through fitting circle
The method of cylindrical equation obtains the cylinder parameter model collection of candidate pipelineIt willTally set after merging fining with previous stepIt asks simultaneously, i.e.,As the initial labels collection of energy function, iteration above-mentioned steps 2) -4), until energy no longer subtracts
It is small；Obtain the tally set of global optimumEach labelThe cylindrical surface parameter of corresponding multiple pipelines and interior point set, wherein
6. the pipeline three-dimensional modeling method according to claim 1 based on three-dimensional point cloud processing, it is characterised in that: step 5
By the collected three-dimensional point cloud operation above-mentioned steps 1 of different points of view to step 4, pipeline construction knot under different points of view is respectively obtained
Then fruit is registrated it using ICP algorithm using the syntople of viewpoint two-by-two, is spliced, entire factory's pipe is finally obtained
The threedimensional model in road；
1) wherein cloud is denoted as Q as source point cloud for order, and taking its abutment points cloud P using the syntople at visual angle is target point cloud,
2) n point is taken to be denoted as P at random in target point cloud P1(P1∈P)；
3) method for using kd-tree to search in source point cloud Q, according to | | Q1-P1| | the criterion of=min finds corresponding point set Q1
(Q1∈Q)；
4) match parameter spin matrix R and translation matrix t and P are calculated1To Q1Average distance
5) to P1Point set is coordinately transformed to obtain P1' and seek P1' arrive Q1Average distance
As error function；
6) whether error in judgement restrains dk-dk+1< τ, τ are setting value and τ > 0, then restrain, otherwise jump to step 2)；
7) target point cloud is translated and is rotated in τ by error convergence, and the new transformation point set P' after being registrated is merged with source point cloud Q,
New point cloud is denoted as Q, then again successively the point cloud adjacent with original point cloud Q and a point cloud adjacent with cloud P be registrated,
Fusion；
8) syntople for utilizing viewpoint, is successively registrated all of its neighbor point cloud, finally obtains the pipeline of entire factory
Threedimensional model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810957920.0A CN109147038B (en) | 2018-08-21 | 2018-08-21 | Pipeline three-dimensional modeling method based on three-dimensional point cloud processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810957920.0A CN109147038B (en) | 2018-08-21 | 2018-08-21 | Pipeline three-dimensional modeling method based on three-dimensional point cloud processing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109147038A true CN109147038A (en) | 2019-01-04 |
| CN109147038B CN109147038B (en) | 2023-02-07 |
Family
ID=64790967
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810957920.0A Active CN109147038B (en) | 2018-08-21 | 2018-08-21 | Pipeline three-dimensional modeling method based on three-dimensional point cloud processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109147038B (en) |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109669425A (en) * | 2019-01-12 | 2019-04-23 | 大连理工大学 | A kind of method of urban duct construction site group to control |
| CN109903379A (en) * | 2019-03-05 | 2019-06-18 | 电子科技大学 | A kind of three-dimensional rebuilding method based on spots cloud optimization sampling |
| CN110334818A (en) * | 2019-05-22 | 2019-10-15 | 广州中船文冲船坞有限公司 | A kind of method and system of pipeline automatic identification |
| CN110889898A (en) * | 2019-12-11 | 2020-03-17 | 南京航空航天大学 | Modeling method suitable for appearance of single aviation conduit |
| CN111192358A (en) * | 2019-12-25 | 2020-05-22 | 五邑大学 | Pipeline crack detection method, device, equipment and storage medium based on three dimensions |
| CN111275815A (en) * | 2020-02-18 | 2020-06-12 | 中国建筑第八工程局有限公司 | Three-dimensional modeling method for existing complex pipeline |
| CN111429563A (en) * | 2020-03-10 | 2020-07-17 | 山东大学 | Pipeline three-dimensional reconstruction method, system, medium and equipment based on deep learning |
| CN111460974A (en) * | 2020-03-30 | 2020-07-28 | 华南理工大学 | Optimization-based global feature extraction method for scattered point cloud data |
| CN111489432A (en) * | 2019-12-16 | 2020-08-04 | 西安航天发动机有限公司 | Bent pipe reconstruction and allowance calculation method based on point cloud data |
| CN111583383A (en) * | 2020-04-07 | 2020-08-25 | 浙江网标物联有限公司 | Three-dimensional visual auxiliary method for high-pressure container inspection |
| CN111951401A (en) * | 2020-08-07 | 2020-11-17 | 中山大学 | Method for constructing precise three-dimensional geometric model of pipeline elbow capable of being used for laser scanning |
| CN112102474A (en) * | 2020-09-01 | 2020-12-18 | 长春工程学院 | Novel cylindrical three-dimensional reconstruction method and system |
| CN112417618A (en) * | 2020-11-20 | 2021-02-26 | 北京工商大学 | Four-dimensional space-time perception large-scene free bending pipeline detection and point cloud completion method |
| CN112509141A (en) * | 2020-12-22 | 2021-03-16 | 华南理工大学 | Method for extracting orthogonal plane pair in building indoor three-dimensional point cloud data |
| WO2021081783A1 (en) * | 2019-10-30 | 2021-05-06 | 深圳市大疆创新科技有限公司 | Point cloud fusion method, apparatus and detection system |
| CN112884886A (en) * | 2021-03-17 | 2021-06-01 | 南通大学 | Three-dimensional point cloud pipeline extraction and modeling method capable of self-adapting to search radius |
| CN113554614A (en) * | 2021-07-21 | 2021-10-26 | 中国人民解放军陆军工程大学 | Pipeline measurement system pose calibration method for point cloud splicing |
| CN113706505A (en) * | 2021-08-24 | 2021-11-26 | 凌云光技术股份有限公司 | Cylinder fitting method and device for removing local outliers in depth image |
| CN113781649A (en) * | 2021-09-07 | 2021-12-10 | 岱悟智能科技(上海)有限公司 | Building plane map generation method based on three-dimensional scanning point cloud |
| WO2022165672A1 (en) * | 2021-02-03 | 2022-08-11 | 深圳市大疆创新科技有限公司 | Point cloud processing method and apparatus, and computer readable storage medium |
| WO2023093085A1 (en) * | 2021-11-29 | 2023-06-01 | 上海商汤智能科技有限公司 | Method and apparatus for reconstructing surface of object, and computer storage medium and computer program product |
| CN116720233A (en) * | 2023-08-09 | 2023-09-08 | 琥崧智能装备(太仓)有限公司 | Three-dimensional space modeling method, device, computer equipment and readable storage medium |
| WO2023240907A1 (en) * | 2022-06-14 | 2023-12-21 | 广东粤海水务投资有限公司 | Pipeline modeling method based on fractional brownian motion |
| CN117576144A (en) * | 2024-01-15 | 2024-02-20 | 湖北工业大学 | Laser point cloud power line extraction method and device and electronic equipment |
| CN111951401B (en) * | 2020-08-07 | 2024-05-17 | 中山大学 | Precise three-dimensional geometric model construction method of pipeline elbow capable of being used for laser scanning |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102722907A (en) * | 2012-05-22 | 2012-10-10 | 清华大学 | Geometric modeling method based on pipe factory point cloud |
| CN102915561A (en) * | 2012-09-27 | 2013-02-06 | 清华大学 | Method of three-dimensional reconstruction for pipeline structures |
| CN105046688A (en) * | 2015-06-23 | 2015-11-11 | 北京工业大学 | Method for automatically identifying multiple planes in three-dimensional point cloud |
| CN106228539A (en) * | 2016-07-12 | 2016-12-14 | 北京工业大学 | Multiple geometric primitive automatic identifying method in a kind of three-dimensional point cloud |
| US20170193692A1 (en) * | 2015-12-30 | 2017-07-06 | Shenzhen Institutes Of Advanced Technology Chinese Academy Of Sciences | Three-dimensional point cloud model reconstruction method, computer readable storage medium and device |
-
2018
- 2018-08-21 CN CN201810957920.0A patent/CN109147038B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102722907A (en) * | 2012-05-22 | 2012-10-10 | 清华大学 | Geometric modeling method based on pipe factory point cloud |
| CN102915561A (en) * | 2012-09-27 | 2013-02-06 | 清华大学 | Method of three-dimensional reconstruction for pipeline structures |
| CN105046688A (en) * | 2015-06-23 | 2015-11-11 | 北京工业大学 | Method for automatically identifying multiple planes in three-dimensional point cloud |
| US20170193692A1 (en) * | 2015-12-30 | 2017-07-06 | Shenzhen Institutes Of Advanced Technology Chinese Academy Of Sciences | Three-dimensional point cloud model reconstruction method, computer readable storage medium and device |
| CN106228539A (en) * | 2016-07-12 | 2016-12-14 | 北京工业大学 | Multiple geometric primitive automatic identifying method in a kind of three-dimensional point cloud |
Non-Patent Citations (1)
| Title |
|---|
| LIANG WANG等: "Energy-based automatic recognition of multiple spheres in three-dimensional point cloud", 《PATTERN RECOGNITION LETTERS》 * |
Cited By (36)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109669425B (en) * | 2019-01-12 | 2020-04-28 | 大连理工大学 | Group-to-group control method for urban pipeline construction site |
| CN109669425A (en) * | 2019-01-12 | 2019-04-23 | 大连理工大学 | A kind of method of urban duct construction site group to control |
| CN109903379A (en) * | 2019-03-05 | 2019-06-18 | 电子科技大学 | A kind of three-dimensional rebuilding method based on spots cloud optimization sampling |
| CN110334818A (en) * | 2019-05-22 | 2019-10-15 | 广州中船文冲船坞有限公司 | A kind of method and system of pipeline automatic identification |
| CN110334818B (en) * | 2019-05-22 | 2022-04-26 | 广州文冲船舶修造有限公司 | Method and system for automatically identifying pipeline |
| WO2021081783A1 (en) * | 2019-10-30 | 2021-05-06 | 深圳市大疆创新科技有限公司 | Point cloud fusion method, apparatus and detection system |
| CN110889898A (en) * | 2019-12-11 | 2020-03-17 | 南京航空航天大学 | Modeling method suitable for appearance of single aviation conduit |
| CN111489432A (en) * | 2019-12-16 | 2020-08-04 | 西安航天发动机有限公司 | Bent pipe reconstruction and allowance calculation method based on point cloud data |
| CN111192358A (en) * | 2019-12-25 | 2020-05-22 | 五邑大学 | Pipeline crack detection method, device, equipment and storage medium based on three dimensions |
| CN111275815A (en) * | 2020-02-18 | 2020-06-12 | 中国建筑第八工程局有限公司 | Three-dimensional modeling method for existing complex pipeline |
| CN111275815B (en) * | 2020-02-18 | 2023-02-28 | 中国建筑第八工程局有限公司 | Three-dimensional modeling method for existing complex pipeline |
| WO2021179593A1 (en) * | 2020-03-10 | 2021-09-16 | 山东大学 | Deep learning-based three-dimensional pipeline reconstruction method, system, medium, and apparatus |
| CN111429563B (en) * | 2020-03-10 | 2021-08-13 | 山东大学 | Pipeline three-dimensional reconstruction method, system, medium and equipment based on deep learning |
| CN111429563A (en) * | 2020-03-10 | 2020-07-17 | 山东大学 | Pipeline three-dimensional reconstruction method, system, medium and equipment based on deep learning |
| CN111460974B (en) * | 2020-03-30 | 2023-04-28 | 华南理工大学 | Scattered point cloud data global feature extraction method based on optimization |
| CN111460974A (en) * | 2020-03-30 | 2020-07-28 | 华南理工大学 | Optimization-based global feature extraction method for scattered point cloud data |
| CN111583383A (en) * | 2020-04-07 | 2020-08-25 | 浙江网标物联有限公司 | Three-dimensional visual auxiliary method for high-pressure container inspection |
| CN111951401A (en) * | 2020-08-07 | 2020-11-17 | 中山大学 | Method for constructing precise three-dimensional geometric model of pipeline elbow capable of being used for laser scanning |
| CN111951401B (en) * | 2020-08-07 | 2024-05-17 | 中山大学 | Precise three-dimensional geometric model construction method of pipeline elbow capable of being used for laser scanning |
| CN112102474A (en) * | 2020-09-01 | 2020-12-18 | 长春工程学院 | Novel cylindrical three-dimensional reconstruction method and system |
| CN112102474B (en) * | 2020-09-01 | 2022-12-06 | 长春工程学院 | Novel axle journal three-dimensional reconstruction method and system |
| CN112417618A (en) * | 2020-11-20 | 2021-02-26 | 北京工商大学 | Four-dimensional space-time perception large-scene free bending pipeline detection and point cloud completion method |
| CN112417618B (en) * | 2020-11-20 | 2022-03-11 | 北京工商大学 | Four-dimensional space-time perception large-scene free bending pipeline detection and point cloud completion method |
| CN112509141A (en) * | 2020-12-22 | 2021-03-16 | 华南理工大学 | Method for extracting orthogonal plane pair in building indoor three-dimensional point cloud data |
| WO2022165672A1 (en) * | 2021-02-03 | 2022-08-11 | 深圳市大疆创新科技有限公司 | Point cloud processing method and apparatus, and computer readable storage medium |
| CN112884886B (en) * | 2021-03-17 | 2023-08-25 | 南通大学 | Three-dimensional point cloud pipeline extraction and modeling method capable of adaptively searching radius |
| CN112884886A (en) * | 2021-03-17 | 2021-06-01 | 南通大学 | Three-dimensional point cloud pipeline extraction and modeling method capable of self-adapting to search radius |
| CN113554614A (en) * | 2021-07-21 | 2021-10-26 | 中国人民解放军陆军工程大学 | Pipeline measurement system pose calibration method for point cloud splicing |
| CN113706505A (en) * | 2021-08-24 | 2021-11-26 | 凌云光技术股份有限公司 | Cylinder fitting method and device for removing local outliers in depth image |
| CN113781649A (en) * | 2021-09-07 | 2021-12-10 | 岱悟智能科技(上海)有限公司 | Building plane map generation method based on three-dimensional scanning point cloud |
| WO2023093085A1 (en) * | 2021-11-29 | 2023-06-01 | 上海商汤智能科技有限公司 | Method and apparatus for reconstructing surface of object, and computer storage medium and computer program product |
| WO2023240907A1 (en) * | 2022-06-14 | 2023-12-21 | 广东粤海水务投资有限公司 | Pipeline modeling method based on fractional brownian motion |
| CN116720233A (en) * | 2023-08-09 | 2023-09-08 | 琥崧智能装备(太仓)有限公司 | Three-dimensional space modeling method, device, computer equipment and readable storage medium |
| CN116720233B (en) * | 2023-08-09 | 2023-10-27 | 琥崧智能装备(太仓)有限公司 | Three-dimensional space modeling method, device, computer equipment and readable storage medium |
| CN117576144A (en) * | 2024-01-15 | 2024-02-20 | 湖北工业大学 | Laser point cloud power line extraction method and device and electronic equipment |
| CN117576144B (en) * | 2024-01-15 | 2024-03-29 | 湖北工业大学 | Laser point cloud power line extraction method and device and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109147038B (en) | 2023-02-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109147038A (en) | Pipeline three-dimensional modeling method based on three-dimensional point cloud processing | |
| Lee et al. | Skeleton-based 3D reconstruction of as-built pipelines from laser-scan data | |
| Wang et al. | Fully automated generation of parametric BIM for MEP scenes based on terrestrial laser scanning data | |
| CN108764048B (en) | Face key point detection method and device | |
| CN113168717B (en) | Point cloud matching method and device, navigation method and equipment, positioning method and laser radar | |
| CN102804231B (en) | Piecewise planar reconstruction of three-dimensional scenes | |
| Zhang et al. | A 3D reconstruction method for pipeline inspection based on multi-vision | |
| CN105654483B (en) | The full-automatic method for registering of three-dimensional point cloud | |
| CN105469388A (en) | Building point cloud registration algorithm based on dimension reduction | |
| CN114332510B (en) | Hierarchical image matching method | |
| CN112444246B (en) | Laser fusion positioning method in high-precision digital twin scene | |
| Ceriani et al. | Pose interpolation slam for large maps using moving 3d sensors | |
| Liu et al. | High-precision detection method for structure parameters of catenary cantilever devices using 3-D point cloud data | |
| Yang et al. | Road constrained monocular visual localization using Gaussian-Gaussian cloud model | |
| CN116518864A (en) | Engineering structure full-field deformation detection method based on three-dimensional point cloud comparison analysis | |
| Zhang et al. | A posture detection method for augmented reality–aided assembly based on YOLO-6D | |
| Palma et al. | Detection of geometric temporal changes in point clouds | |
| Ye et al. | Ec-sfm: Efficient covisibility-based structure-from-motion for both sequential and unordered images | |
| Maiza et al. | Automatic classification of archaeological potsherds | |
| Liu et al. | An improved registration strategy for aligning incomplete blade measurement data to its model | |
| Zhao et al. | Accurate extraction of building roofs from airborne light detection and ranging point clouds using a coarse-to-fine approach | |
| Xin et al. | Accurate and complete line segment extraction for large-scale point clouds | |
| Chen et al. | Plane segmentation for a building roof combining deep learning and the RANSAC method from a 3D point cloud | |
| Cao et al. | Aircraft pipe gap inspection on raw point cloud from a single view | |
| Zhang et al. | Pose estimation of space objects based on hybrid feature matching of contour points |
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 |