CN108510516A - A kind of the three-dimensional line segment extracting method and system of dispersion point cloud - Google Patents
A kind of the three-dimensional line segment extracting method and system of dispersion point cloud Download PDFInfo
- Publication number
- CN108510516A CN108510516A CN201810295190.2A CN201810295190A CN108510516A CN 108510516 A CN108510516 A CN 108510516A CN 201810295190 A CN201810295190 A CN 201810295190A CN 108510516 A CN108510516 A CN 108510516A
- Authority
- CN
- China
- Prior art keywords
- dimensional
- line segment
- planar
- 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.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
- G06T2207/20068—Projection on vertical or horizontal image axis
Abstract
The present invention relates to the three-dimensional line segment extracting methods and system of a kind of dispersion point cloud, and extraction is detected to the three-dimensional line segment of extensive dispersion point cloud.This method includes mainly three steps:Point cloud segmentation is three-dimensional planar by region growing and region merging technique by the first step;Second step realizes two-dimensional projection to each three-dimensional planar, extracts profile on 2d and least square fitting obtains two-dimensional line segment.Again by two-dimensional line segment back projection to three-dimensional planar, corresponding three-dimensional line segment is obtained.Third walks, and eliminates outlier by post-processing and merges adjacent three-dimensional line segment.Three-dimensional edges point is first extracted from traditional algorithm and is fitted that the method for three-dimensional line segment is different again, and the present invention realizes the extraction of three-dimensional line segment on the basis of point cloud segmentation and two-dimensional projection, is simple and efficient, and suitable for the dispersion point cloud line segments extraction of different scenes.
Description
Technical field
The present invention relates to points cloud processing technical fields, a kind of three-dimensional line segment extracting method more particularly to dispersion point cloud and
System.
Background technology
Three-dimensional point cloud is the set of one group of discrete point with three dimensional space coordinate information.Compared with two-dimensional image data, point
There is cloud data the advantage in dimension, the coordinate information of three-dimensional point to provide spatial description more more intuitive than two-dimensional pixel information.
Therefore, point cloud data more preferable can must describe the geometrical property and topological structure of real world.Not with laser scanner technique
Disconnected upgrading, three-dimensional laser scanner can quickly and easily be obtained the high-precision point cloud data of object, be built using point cloud data
Mould can be obtained corresponding threedimensional model, and it is more which has penetrated into smart city construction, machine-building, reverse-engineering etc.
A industry.Due to laser point cloud no chapter at random, the curved surface quality similarities and differences, calculation can be increased to the point cloud data unified Modeling processing of magnanimity
The difficulty of method and the complexity of mathematical notation.In order to reduce the complexity of Point Cloud Processing, a kind of feasible way is only to examine
The structural information for considering point cloud has due to there is structural informations, the extractions of structural information such as a large amount of face, line in artificial scene
The complexity of cloud is put conducive to reduction and does not lose the main characterization information of a cloud.
Current more general point cloud line segments extraction method can be divided into three classes:Based on the extraction of point, extraction based on face and
Extraction based on image.
Method based on point usually first detects boundary point, is then fitted three-dimensional line segment with least-square fitting approach.It is examining
When surveying boundary point, usually there are two types of algorithms:First, first recycling algorithm of convex hull to find out segmentation at small segmentation block point cloud segmentation
The boundary point of block;Another method is to calculate Point cloud curvature, the features such as normal vector, by entire point cloud data carry out boundary point cloud and
Two classifications of non-boundary point cloud.Method based on point is the biggest problems are that extraction effect depends on characteristic point itself, to noise
It is insensitive.
Method based on plane considers the intersection point that three-dimensional line segment is regarded as to two three-dimensional planars, is usually used in airborne laser thunder
The modeling reached, but this kind of cloud usually will not be very intensive.Based on plane method the shortcomings that be that the terminal of cross spider is difficult to determine,
And this method is not suitable for facet.
Input point cloud is converted into bidimensional image by the method based on image first, then line segment detector is utilized to extract image
On two-dimensional line segment, finally these two-dimensional line segments are projected to again on cloud, obtain final three-dimensional line segment.It will point cloud conversion
It is a kind of common points cloud processing at two dimensional image, such as by point cloud segmentation at several cross sections, pair cross-section point cloud is projected
Bidimensional image is obtained, recycles two-dimensional line segment extraction algorithm to extract straightway, back projection is at three-dimensional line segment.Based on bidimensional image
Extracting method fully depends on the generation strategy of image, and number, resolution ratio and the viewpoint of projected image are difficult to determine.
Invention content
The technical problem to be solved by the present invention is to solve the above shortcomings of the prior art and to provide the three of a kind of dispersion point cloud
Tie up line segments extraction method and system.
The technical solution that the present invention solves above-mentioned technical problem is as follows:A kind of three-dimensional line segment extracting method of dispersion point cloud,
Including:
Step 1 after building k-d tree to the dispersion point cloud of input, utilizes it in the k-dtree each point
Around closest K point, fitting obtains the planar curvature of the point;
The planar curvature that step 2, basis are each put, by region growing and region merging technique, by the dispersion point cloud
It is divided into multiple three-dimensional planars;
Step 3, for each three-dimensional planar, all the points in it are projected in preset plane, formed a two dimension
Image extracts the two-dimensional line segment of the bidimensional image, and by the two-dimensional line segment back projection to three-dimensional planar, obtains three-dimensional line
Section;
Step 4 is removed and line by carrying out abnormal line segment to the corresponding three-dimensional line segment of three-dimensional planar described in each
Section merges, and obtains the new three-dimensional line segment of the dispersion point cloud.
Based on the above technical solution, the present invention can also be improved as follows.
Further, the step 2 includes:
Step 2.1, the planar curvature sequence from small to large according to each point, to the click-through in the dispersion point cloud
Row sequence, and since the corresponding point of minimum curvature, determine the coplanar point each put successively, obtain multiple cloud sectors
Domain, wherein each point cloud sector domain is made of coplanar point;
Step 2.2 is fitted the multiple cloud sector domain one normal vector and determines a fit Plane one by one, to each
Point in the fit Plane sets label, by the label, determines the adjacent flat of each fit Plane, and should
The corresponding adjacent flat of fit Plane merges, and obtains multiple three-dimensional planars.
Further, the step 1 includes:
Step 1.1, the corresponding k-dtree of structure dispersion point cloud;
Step 1.2, to the every bit p in the dispersion point cloudiIf its K in the k-d tree closest points
The collection of composition is combined intoBuild covariance matrix Σ:
In formula, Σ represents 3 × 3 covariance matrix, and K isThe number at midpoint,It isThe average value at midpoint;
Step 1.3, solution seek the characteristic value of the covariance matrix Σ, and using minimal eigenvalue as the point piPlane
Curvature
Further, the step 3 includes:
Step 3.1 calculates the average value that each described three-dimensional planar corresponds to point set, according to the average value, determines flat
Face central point obtains two by other spot projections to the default two dimensional surface Jing Guo the planar central point of the point concentration
Tie up image;
Step 3.2 converts the bidimensional image to binary picture, and extracts image wheel on the binary picture
Exterior feature recycles least square fitting to extract the two-dimensional line segment in the image contour;
Step 3.3, by the two-dimensional line segment back projection to its corresponding three-dimensional planar, obtain three-dimensional line segment.
Further, the step 4 includes:
Step 4.1 is ranked up all three-dimensional line segments in three-dimensional planar described in each according to length;
Step 4.2, according to it is described sequence and preset classifying rules, corresponding to three-dimensional planar described in each described three
Dimension line segment is classified, and first category line segment and second category line segment are obtained;
Step 4.3 is unsatisfactory for when the corresponding first category line segment of the first three-dimensional planar and the second category line segment
Formula:l(c_1)+l(c_2)>When 0.5*l (all), then its corresponding first three-dimensional planar is deleted, wherein l (c_1) and l
(c_2) the length summation of all line segments in first category line segment is indicated respectively, and the length of all line segments is total in second category line segment
With l (all) indicates the length summation of all line segments in the three-dimensional planar;
Step 4.4, according to each corresponding two-dimensional silhouette of the second three-dimensional planar, one by one determine its corresponding three-dimensional line
Section determines and removes the abnormal three-dimensional line segment by abnormal three-dimensional line segment criterion, obtain each second three-dimensional planar
First three-dimensional line segment group, wherein second three-dimensional planar be the multiple three-dimensional planar in except first three-dimensional planar it
Outer three-dimensional planar;
Three-dimensional line segment in the first three-dimensional line segment group of each second three-dimensional planar is passed through z by step 4.5
Direction normalizes, and obtains histogram;
Step 4.6 determines that each three-dimensional line segment is adjacent thereto in the histogram in the first three-dimensional line segment group
Three-dimensional line segment;
Step 4.7, according to the adjacent three-dimensional line segment and it is default merge criterion, carry out three-dimensional line segment merging, obtain institute
State the new three-dimensional line segment of dispersion point cloud.
The present invention also provides a kind of storage mediums, the journey of the three-dimensional line segment extracting method for storing above-mentioned dispersion point cloud
Sequence, details are not described herein for specific method.
The present invention also provides a kind of three-dimensional line segment extraction systems of dispersion point cloud, including:
Three-dimensional planar generation module utilizes it after the dispersion point cloud structure k-d tree to input to each point
K closest point around in the k-d tree, fitting obtain the planar curvature of the point, and according to described in each point
The dispersion point cloud is divided into multiple three-dimensional planars by planar curvature by region growing and region merging technique;
Three-dimensional line segment generation module is flat for each three-dimensional for being generated for the three-dimensional planar generation unit
Face projects to all the points in it in preset plane, forms a bidimensional image, extracts the two-dimensional line of the bidimensional image
Section, and by the two-dimensional line segment back projection to three-dimensional planar, obtain three-dimensional line segment;
Data processing module, for each described three-dimensional planar pair by being generated to the three-dimensional line segment generation unit
The three-dimensional line segment answered carries out abnormal line segment removal and line segment merges, and obtains the new three-dimensional line segment of the dispersion point cloud.
Further, the three-dimensional planar generation module includes:
Fitting unit, after the dispersion point cloud structure k-d tree to input, to each point using it in the k-
Closest K point around in dtree, fitting obtain the planar curvature of the point;
Area generation unit, for according to the planar curvature put of each of fitting unit fitting from small to large
Sequentially, the point in the dispersion point cloud is ranked up, and since the corresponding point of minimum curvature, successively determine each put with
Its coplanar point obtains multiple cloud sectors domain, wherein each point cloud sector domain is made of coplanar point;
Region merging technique unit, the multiple cloud sector domain for being obtained to the Area generation unit are fitted one one by one
Normal vector simultaneously determines a fit Plane, and label is set to the point in each fit Plane, by the label, determines every
The adjacent flat of a fit Plane, and the corresponding adjacent flat of the fit Plane is merged, it obtains multiple three-dimensional flat
Face.
Further, the fitting unit is specifically used for:
Build the corresponding k-d tree of dispersion point cloud;To the every bit p in the dispersion point cloudiIf it is in the k-d
The collection that K closest points in tree are constituted is combined intoBuild covariance matrix Σ:
In formula, Σ represents 3 × 3 covariance matrix, and K isThe number at midpoint,It isThe average value at midpoint;Solution asks institute
The characteristic value of covariance matrix Σ is stated, and using minimal eigenvalue as the point piPlanar curvature
Further, the three-dimensional line segment generation module is specifically used for:
It calculates each described three-dimensional planar and corresponds to the average value of point set and planar central point is determined according to the average value,
On other spot projections to the default two dimensional surface Jing Guo the planar central point that the point is concentrated, bidimensional image is obtained;It will
The bidimensional image is converted into binary picture, and image contour is extracted on the binary picture, recycles least square
The two-dimensional line segment in the image contour is extracted in fitting;By in the two-dimensional line segment back projection to its corresponding three-dimensional planar, obtain
To three-dimensional line segment.
Further, the data processing module is specifically used for:
All three-dimensional line segments in three-dimensional planar described in each are ranked up according to length;According to the sequence and in advance
If classifying rules, the three-dimensional line segment corresponding to three-dimensional planar described in each classifies, and obtains first category line segment
With second category line segment;When the corresponding first category line segment of the first three-dimensional planar and the second category line segment are unsatisfactory for public affairs
Formula:l(c_1)+l(c_2)>When 0.5*l (all), then its corresponding first three-dimensional planar is deleted, wherein l (c_1) and l
(c_2) the length summation of all line segments in first category line segment is indicated respectively, and the length of all line segments is total in second category line segment
With l (all) indicates the length summation of all line segments in the three-dimensional planar;According to each corresponding two dimension of the second three-dimensional planar
Profile determines its corresponding three-dimensional line segment one by one, by abnormal three-dimensional line segment criterion, determines and remove the abnormal three-dimensional line segment,
Obtain the first three-dimensional line segment group of each second three-dimensional planar, wherein second three-dimensional planar is the multiple three
Three-dimensional planar in dimensional plane in addition to first three-dimensional planar;By the described 1st of each second three-dimensional planar the
Three-dimensional line segment in dimension line segment group is normalized by the directions z, obtains histogram;Determine each in the first three-dimensional line segment group
Three-dimensional line segment three-dimensional line segment adjacent thereto in the histogram;Sentenced according to the adjacent three-dimensional line segment and default merging
According to progress three-dimensional line segment merging obtains the new three-dimensional line segment of the dispersion point cloud.
Description of the drawings
Fig. 1 is a kind of flow signal of three-dimensional line segment extracting method of dispersion point cloud provided by one embodiment of the present invention
Figure;
Step 110 in a kind of three-dimensional line segment extracting method for dispersion point cloud that Fig. 2 provides for another embodiment of the present invention
Flow diagram;
Step 120 in a kind of three-dimensional line segment extracting method for dispersion point cloud that Fig. 3 provides for another embodiment of the present invention
Flow diagram;
Step 140 in a kind of three-dimensional line segment extracting method for dispersion point cloud that Fig. 4 provides for another embodiment of the present invention
Flow diagram;
Fig. 5 is a kind of schematic frame of the three-dimensional line segment extraction system of dispersion point cloud provided by one embodiment of the present invention
Figure;
Three-dimensional planar in a kind of three-dimensional line segment extraction system for dispersion point cloud that Fig. 6 provides for another embodiment of the present invention
The schematic block diagram of generation module.
Specific implementation mode
The principle and features of the present invention will be described below with reference to the accompanying drawings, and the given examples are served only to explain the present invention, and
It is non-to be used to limit the scope of the present invention.
Embodiment one
A kind of three-dimensional line segment extracting method 100 of dispersion point cloud, as shown in Figure 1, including:
Step 110 after building k-d tree to the dispersion point cloud of input, utilizes it in k-d tree each point
K closest point of surrounding, fitting obtain the planar curvature of the point.
Dispersion point cloud is divided into more according to the planar curvature each put by region growing and region merging technique by step 120
A three-dimensional planar.
Step 130, for each three-dimensional planar, all the points in it are projected in preset plane, formed one two
Image is tieed up, extracts the two-dimensional line segment of bidimensional image, and by two-dimensional line segment back projection to three-dimensional planar, obtain three-dimensional line segment.
Step 140, abnormal line segment removes and line segment merges by being carried out to the corresponding three-dimensional line segment of each three-dimensional planar,
Obtain the new three-dimensional line segment of dispersion point cloud.
Point cloud segmentation is three-dimensional planar by region growing and region merging technique by the present embodiment first step;Second step, it is right
Each three-dimensional planar realizes two-dimensional projection, extracts profile on 2d and least square fitting obtains two-dimensional line segment.Again
By in two-dimensional line segment back projection to three-dimensional planar, corresponding three-dimensional line segment is obtained.Third walk, by post-processing eliminate outlier and
Merge adjacent three-dimensional line segment.Three-dimensional edges point is first extracted from traditional algorithm is fitted that the method for three-dimensional line segment is different, and the present invention exists again
The extraction that three-dimensional line segment is realized on the basis of point cloud segmentation and two-dimensional projection, is simple and efficient, and suitable for the at random of different scenes
Point cloud line segments extraction.
Embodiment two
On the basis of embodiment one, as shown in Fig. 2, step 120 includes:
Step 121, the planar curvature sequence from small to large according to each point, are ranked up the point in dispersion point cloud,
And since the corresponding point of minimum curvature, the coplanar point each put is determined successively, obtains multiple cloud sectors domain, wherein
Each point cloud sector domain is made of coplanar point.
Step 122 is fitted multiple cloud sectors domain one normal vector and determines a fit Plane one by one, to each fitting
Point setting label in plane determines the adjacent flat of each fit Plane, and the fit Plane is corresponding by label
Adjacent flat merge, obtain multiple three-dimensional planars.
It should be noted that step 121 is region growing.Specifically, by each of point cloud data point according to each point
Amount of curvature be ranked up, first select curvature minimum point psAs seed point, chained list T is created to store and psCoplanar
Point.The mode that region increases is proceeded by from seed point to calculate segmentation plane.It is rightIn each candidate point pjIf met
(1)pjNormal vectorWith the normal vector of coplanar point in chained list TAngle be less than threshold θ (θ=15 °);(2)pjTo ps
Orthogonal distance be less than threshold value(3)pjTo psParallel distance be less than threshold value thp
Then by pjIt is added in chained list T, indicates pjWith psIt is coplanar.Formula indicates as follows:
WhenMiddle all the points are all traversed, then mark piFor processed point set, until point all in T is all handled
It crosses, then stops growing, all points form a plane domain in T.Iteration above-mentioned steps, until the point cloud data of all inputs
It is all processed, it obtains all plane domains and is referred to as R.
Step 122 merges multiple cloud sectors domain, obtains multiple three-dimensional planars.Specifically, for institute in step 121
Each region R obtainedi, it is fitted a plane using the method mentioned in above-mentioned calculating normal vector, is used in combinationWith
Indicate plane RiNormal vector, curvature and scale.For plane RiIn each point, distribute a label lable, indicate
It belongs to plane.Each plane is traversed, its adjacent flat is obtained.For plane RiIn each point pm, define and traverse it
Neighborhood point setIfThere are certain point pnLabel be different from pm, then consider pnFor plane RiBoundary point, and pnBelong to Ri
A certain adjacent flat, plane R so also can be obtainediAll of its neighbor plane.With reference to the algorithm of region growing in step 121,
All coplanar planes are clustered, i.e., the candidate point in step 121 is replaced with adjacent flat, and region growing condition is:(1)
The normal vector angle of two planes is less than threshold θ (θ=15 °);The orthogonal distance of (2) two interplanars is less than threshold value tho
It is after step 120 point cloud segmentation the result is that a series of three-dimensional facet, the normal vector of facet is represented by n∏,
T∏It is the correspondence point set of facet.
Embodiment three
On the basis of embodiment two, step 110 includes:
Step 111, the corresponding k-dtree of structure dispersion point cloud.
Step 112, to the every bit p in dispersion point cloudiIf the collection that its K in k-d tree closest points are constituted
It is combined intoBuild covariance matrix Σ:
In formula, Σ represents 3 × 3 covariance matrix, and K isThe number at midpoint,It isThe average value at midpoint.
Step 113, solution seek the characteristic value of covariance matrix Σ, and using minimal eigenvalue as the point piPlanar curvature
Specifically, constitutive characteristic value equation first:λ V=∑s V.Carrying out singular value decomposition (SVD decomposition) to the equation can be with
Obtain three characteristic values and corresponding three feature vectors, also referred to as principal component (PCs).By these three characteristic values from greatly to
Minispread:λ2>λ1>λ0, corresponding three feature vectors are denoted as v successively2、v1And v0.The first two principal component v2And v1Mutually just
It hands over, constitutesBest fit three-dimensional surface, third principal component v0It is orthogonal with the first two principal component, therefore can be used as
piNormal vectorWith v0Corresponding eigenvalue λ0Represent beIn point cloud in v0Variation journey on representative direction
Degree, therefore p can be used asiCurvatureDue to known point piThe collection that constitute of the closest points of K be combined intoIt can letter
It is single to define piScaleFor piThe distance between third consecutive points nearest with it.
Example IV
In embodiment one to embodiment three on the basis of any embodiment, as shown in figure 3, step 130 includes:
Step 131 calculates the average value that each three-dimensional planar corresponds to point set and determines planar central point according to average value,
By on other spot projections to the default two dimensional surface Jing Guo planar central point concentrated, bidimensional image is obtained.
Step 132 converts bidimensional image to binary picture, and extracts image contour on binary picture, then profit
The two-dimensional line segment in image contour is extracted with least square fitting.
Step 133, by two-dimensional line segment back projection to its corresponding three-dimensional planar, obtain three-dimensional line segment.
It should be noted that the three-dimensional line segment detection based on plane.For each three-dimensional facet, by the institute belonging to it
Pointed set projects in plane, forms bidimensional image, then carries out two-dimensional silhouette extraction and least square fitting, obtains each flat
The two-dimensional line segment in face.Then these two-dimensional line segments are projected on three-dimensional planar again, obtains three-dimensional line segment.Specific steps 131
It is as follows:
Step 1311, tripleplane.For obtained three-dimensional facet ∏ in step 120 according to its affiliated point set T∏'s
Average value can obtain planar central point pc.Randomly choose a point p of plane ∏0It is projected, corresponds to subpoint and be defined as p0'.If
pcWith p0' line be x-axis,It is denoted as v for positive directionx, then y-axis vy=vx×n∏, thus any point p in spacei's
Two-dimensional coordinate (xi,yi) can be obtained by following formula:
Step 1312, binary picture converts.Rule of thumb define planar dimension s∏=0.75 { Sp}0.9, wherein { SpTable
Show the set pressed scale and sorted from small to large in plane ∏, { Sp}0.9Indicate { SpIn first 90%th.Known plane meter
Degree, so that it may these points are converted into binary picture by grid.It is sat from plane ∏ can be obtained in step 1311 in two dimension
Maxima and minima in mark system in x-axis and y-axis, is respectively defined as xmax, xminAnd ymax, ymin.Thus it can create one wide
For [(xmax-xmin)/s∏]+1, a height of [(ymax-ymim)/s∏]+1 binary picture, and its pixel coordinate (ui,vi) can pass through
Following formula calculates:
ui=(xi-xmin)/s∏
vi=(yi-ymin)/s∏
The gray value for the point converted in bidimensional image is set as 255, rest of pixels is set as 0, obtains two-dimentional shadow
The binary picture of picture.Then expansive working is carried out to original binary image, is used in combination 3 × 3 kernel to be corroded, this is not only
The resolution ratio of original image can be maintained, moreover it is possible to compensate for most of cavity on original binary image.
Step 132, bidimensional image line segments extraction.Image contour is first extracted on the binary picture obtained in step 131,
Least square fitting is recycled to extract two-dimensional line segment.It is real that contour detecting can directly invoke the findContours functions in OpenCV
It is existing, and the profile that size is less than 40 pixels is deleted, because they can not possibly include useful line segment.Finally use least square fitting
Line segment is used in combination different colors to be labeled different line segments.
Step 133, bidimensional image back projection.The two-dimensional line segment extracted from binary picture, by anti-tripleplane
It can obtain corresponding three-dimensional line segment.Anti- three-dimensional is broadly divided into two steps:
Step 1331, using the pixel coordinate calculation formula in step 1312, image binary picture is converted to two dimension
Image.
Step 1332, any point p in the space in step 1311 is utilizediTwo-dimensional coordinate calculation formula, can obtain
Corresponding three-dimensional coordinate, to obtain the extraction result of three-dimensional line segment
Embodiment five
In embodiment one to example IV on the basis of any embodiment, as shown in figure 4, step 140 includes:
Step 141 is ranked up all three-dimensional line segments in each three-dimensional planar according to length.
Step 142, according to sequence and preset classifying rules, the corresponding three-dimensional line segment of each three-dimensional planar is divided
Class obtains first category line segment and second category line segment.
Step 143 is unsatisfactory for formula when the corresponding first category line segment of the first three-dimensional planar and second category line segment:l(c_
1)+l(c_2)>When 0.5*l (all), then its corresponding first three-dimensional planar is deleted, wherein l (c_1) and l (c_2) are indicated respectively
The length summation of all line segments in first category line segment, the length summation of all line segments in second category line segment, l (all) are indicated
The length summation of all line segments in the three-dimensional planar.
Step 144, according to each corresponding two-dimensional silhouette of the second three-dimensional planar, one by one determine its corresponding three-dimensional line
Section determines and removes the abnormal three-dimensional line segment by abnormal three-dimensional line segment criterion, obtain the first of each the second three-dimensional planar
Three-dimensional line segment group, wherein the second three-dimensional planar is the three-dimensional planar in addition to the first three-dimensional planar in multiple three-dimensional planars.
Three-dimensional line segment in first three-dimensional line segment group of each the second three-dimensional planar is passed through the directions z normalizing by step 145
Change, obtains histogram.
Step 146 determines the three-dimensional line that each three-dimensional line segment is adjacent thereto in histogram in the first three-dimensional line segment group
Section.
Step 147 obtains scattered points according to adjacent three-dimensional line segment and default merging criterion, progress three-dimensional line segment merging
The new three-dimensional line segment of cloud.
It should be noted that specifically including:
Step 141~step 143, exceptional value remove, and include the removal of the exceptional value of three-dimensional planar and three-dimensional line segment.
Three-dimensional planar exceptional value removes.For each three-dimensional planar ∏, all three-dimensional line segments that will belong in the plane
It is ranked up from big to small by length, wherein reference line of the longest line segment as first classification.Consider reference line and three-dimensional
All non-classified line segment L in planeiBetween angular separation, if less than L if 10 °iWith reference line in same classification;If three
All non-classified line segment L in dimensional planeiAngle between reference line is more than 30 °, then with LiFor reference line, one is created newly
Classification.But in general, if three-dimensional planar is regular, such as the facade and window of building, have it is very strong parallel and
The structural informations such as orthogonal, and those exceptional values do not have this structural information usually such as trees and vegetables.Therefore, for rule
Three-dimensional planar, the present invention proposes the standard that a kind of exceptional value is distinguished:
l(c1)+l(c2)>0.5*l(all)
l(c1) and l (c2) indicate that the length summations of all line segments in the first and second classifications, l (all) indicate plane respectively
The summation of all line segments in ∏.
Three-dimensional line segment exceptional value is removed using the affiliated profile of line segment.The length threshold th of line segment removal is setlIf profile
Structure it is notable, the smaller length threshold of setting, conversely, using larger length threshold.It is s for a scale∏Plane
∏ traverses all three-dimensional line segment in plane first, therefrom find out contoured structure lines (definition of structure lines is:Its side
It is less than certain threshold value to the deviation between any one structure direction of plane ∏, is rule of thumb set as the threshold value
10°).The structure line slope t of profile is defined as the length of all structure lines divided by the length of all line segments on profile.It is long
Spend threshold value thlObtaining value method it is as follows:(1) if t>90%, then thl=20s∏;(2) if 90% >=t >=30%, thl=
60s∏;(3) if t<30%, then thl=100s∏.Meanwhile if the profile of line segment is less than 60s∏, remove entire profile.
Step 144~step 147, line segment merge.Its purpose is to merge excessively similar line segment so that line
The final result of section extraction is more clear.Key step has:
The latitude of point line segment (inlier) is arcsin (z) in defining, and wherein z is after line segment direction normalizes in the directions z
On component.The latitude that all inlier are divided between being with 6 ° of step-length establishes a histogram.By all inlier according to it
Length value be ranked up from big to small.
Select the longest line segment L in untreated line segmentiAs initial segment, L is traversediPlace bin and its left and right are adjacent
Each three-dimensional line segment L in two binj, find and meet the line segment below for merging and assuming:
Wherein 0.1 is empirical value,WithIndicate L in original point cloudiTo LjThe distance between, mag=| | P0| | it indicates
The size of input point cloud coordinate, i.e. the first of input point cloud point arrive the distance between original point.
For the combined line segment L of each hypothesisj, two endpoint is calculated to LiVertical range, if the two distances
Any one of be more than 4s∏(s∏For LiThe scale of place plane), then give up merging hypothesis.If line segment LjWith LiBetween foot
It is enough close, it is assumed that two lines section has coincidence or the intersegmental gap threshold of two lines, and sufficiently small (setting gap threshold value is less than 10s∏),
L will be usedjIt is projected in Li, choose four endpoint (LiTwo endpoints and LjIt is projected in LiOn two endpoints) in distance it is farthest
Two endpoints update LiTwo endpoints, and mark line segment LjIt is merged.
Export final result of all not labeled combined line segments as Line segment detection.
Embodiment six
A kind of storage medium, the program of the three-dimensional line segment extracting method for storing above-mentioned dispersion point cloud, specific method
Details are not described herein.
Embodiment seven
A kind of three-dimensional line segment extraction system 200 of dispersion point cloud, as shown in figure 5, including:
Three-dimensional planar generation module utilizes it after the dispersion point cloud structure k-d tree to input to each point
Closest K point around in k-d tree, fitting obtain the planar curvature of the point, and according to the Plane Curved of each point
Dispersion point cloud is divided into multiple three-dimensional planars by rate by region growing and region merging technique.
Three-dimensional line segment generation module, each three-dimensional planar for being generated for three-dimensional planar generation unit, will be in it
All the points project in preset plane, form a bidimensional image, extract the two-dimensional line segment of bidimensional image, and by two-dimensional line segment
In back projection to three-dimensional planar, three-dimensional line segment is obtained.
Data processing module, for the corresponding three-dimensional of each three-dimensional planar by being generated to three-dimensional line segment generation unit
Line segment carries out abnormal line segment removal and line segment merges, and obtains the new three-dimensional line segment of dispersion point cloud.
Embodiment eight
On the basis of embodiment six, as shown in fig. 6, three-dimensional planar generation module includes:
Fitting unit, after the dispersion point cloud structure k-d tree to input, to each point using it in k-dtree
In around closest K point, fitting obtains the planar curvature of the point.
Area generation unit, it is right for each of being fitted the sequence of the planar curvature put from small to large according to fitting unit
Point in dispersion point cloud is ranked up, and since the corresponding point of minimum curvature, determines the coplanar point each put successively,
Obtain multiple cloud sectors domain, wherein each point cloud sector domain is made of coplanar point.
Region merging technique unit, the multiple cloud sectors domain for being obtained to Area generation unit are fitted a normal vector simultaneously one by one
It determines a fit Plane, the adjoining of each fit Plane is determined by label to the point setting label in each fit Plane
Plane, and the corresponding adjacent flat of the fit Plane is merged, obtain multiple three-dimensional planars.
Embodiment nine
On the basis of embodiment seven, fitting unit is specifically used for:
Build the corresponding k-d tree of dispersion point cloud;To the every bit p in dispersion point cloudiIf its K in k-d tree
The collection that a closest point is constituted is combined intoBuild covariance matrix Σ:
In formula, Σ represents 3 × 3 covariance matrix, and K isThe number at midpoint,It isThe average value at midpoint;Solution asks association
The characteristic value of variance matrix Σ, and using minimal eigenvalue as point piPlanar curvature
Embodiment ten
In embodiment six to embodiment eight on the basis of any embodiment, three-dimensional line segment generation module is specifically used for:
It calculates the average value that each three-dimensional planar corresponds to point set and planar central point is determined according to average value, concentrated
Other spot projections to the default two dimensional surface Jing Guo planar central point on, obtain bidimensional image;Bidimensional image is converted into two
Into imaged, and image contour is extracted on binary picture, recycle the two dimension in least square fitting extraction image contour
Line segment;By in two-dimensional line segment back projection to its corresponding three-dimensional planar, three-dimensional line segment is obtained.
Embodiment 11
In embodiment six to embodiment nine on the basis of any embodiment, data processing module is specifically used for:To each
All three-dimensional line segments in a three-dimensional planar are ranked up according to length;According to sequence and preset classifying rules, to each
The corresponding three-dimensional line segment of three-dimensional planar is classified, and first category line segment and second category line segment are obtained;When the first three-dimensional planar
Corresponding first category line segment and second category line segment are unsatisfactory for formula:l(c_1)+l(c_2)>When 0.5*l (all), then delete
Its corresponding first three-dimensional planar, wherein l (c_1) and l (c_2) indicates the length of all line segments in first category line segment respectively
Summation, the length summation of all line segments, l (all) indicate that the length of all line segments in the three-dimensional planar is total in second category line segment
With;According to each corresponding two-dimensional silhouette of the second three-dimensional planar, its corresponding three-dimensional line segment is determined one by one, by abnormal three-dimensional
Line segment criterion determines and removes the abnormal three-dimensional line segment, obtains the first three-dimensional line segment group of each the second three-dimensional planar,
In, the second three-dimensional planar is the three-dimensional planar in addition to the first three-dimensional planar in multiple three-dimensional planars;By each the second three-dimensional
Three-dimensional line segment in first three-dimensional line segment group of plane is normalized by the directions z, obtains histogram;Determine the first three-dimensional line segment group
In each three-dimensional line segment three-dimensional line segment adjacent thereto in histogram;Sentenced according to adjacent three-dimensional line segment and default merging
According to progress three-dimensional line segment merging obtains the new three-dimensional line segment of dispersion point cloud.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent replacement, improvement and so on should all be included in the protection scope of the present invention.
Claims (10)
1. a kind of three-dimensional line segment extracting method of dispersion point cloud, which is characterized in that including:
After building k-d tree to the dispersion point cloud of input, its week in the k-d tree is utilized to each point for step 1
K closest point is enclosed, fitting obtains the planar curvature of the point;
The dispersion point cloud is divided into according to the planar curvature each put by region growing and region merging technique by step 2
Multiple three-dimensional planars;
Step 3, for each three-dimensional planar, all the points in it are projected in preset plane, formed a bidimensional image,
The two-dimensional line segment of the bidimensional image is extracted, and by the two-dimensional line segment back projection to three-dimensional planar, obtains three-dimensional line segment;
Step 4, by being carried out to the corresponding three-dimensional line segment of three-dimensional planar described in each, abnormal line segment removes and line segment closes
And obtain the new three-dimensional line segment of the dispersion point cloud.
2. a kind of three-dimensional line segment extracting method of dispersion point cloud according to claim 1, which is characterized in that the step 2
Including:
Step 2.1, the planar curvature sequence from small to large according to each point, arrange the point in the dispersion point cloud
Sequence, and since the corresponding point of minimum curvature, determine the coplanar point each put successively, obtain multiple cloud sectors domain,
In, each cloud sector domain of putting is made of coplanar point;
Step 2.2 is fitted the multiple cloud sector domain one normal vector and determines a fit Plane one by one, to each described
Point in fit Plane sets label, by the label, determines the adjacent flat of each fit Plane, and by the fitting
The corresponding adjacent flat of plane merges, and obtains multiple three-dimensional planars.
3. a kind of three-dimensional line segment extracting method of dispersion point cloud according to claim 2, which is characterized in that the step 1
Including:
Step 1.1, the corresponding k-d tree of structure dispersion point cloud;
Step 1.2, to the every bit p in the dispersion point cloudiWhat if its K in the k-d tree closest points were constituted
Collection is combined intoBuild covariance matrix Σ:
In formula, Σ represents 3 × 3 covariance matrix, and K isThe number at midpoint,It isThe average value at midpoint;
Step 1.3, solution seek the characteristic value of the covariance matrix Σ, and using minimal eigenvalue as the point piPlanar curvature
4. a kind of three-dimensional line segment extracting method of dispersion point cloud according to claim 1, which is characterized in that the step 3
Including:
Step 3.1 calculates each described three-dimensional planar and corresponds to the average value of point set and determined in plane according to the average value
Heart point obtains bidimensional image by the spot projection to the default two dimensional surface Jing Guo the planar central point of the point concentration;
Step 3.2 converts the bidimensional image to binary picture, and extracts image contour on the binary picture,
Least square fitting is recycled to extract the two-dimensional line segment in the image contour;
Step 2.3, by the two-dimensional line segment back projection to its corresponding three-dimensional planar, obtain three-dimensional line segment.
5. a kind of three-dimensional line segment extracting method of dispersion point cloud according to any one of claims 1 to 4, which is characterized in that
The step 4 includes:
Step 4.1 is ranked up all three-dimensional line segments in three-dimensional planar described in each according to length;
Step 4.2, according to it is described sequence and preset classifying rules, the three-dimensional line corresponding to three-dimensional planar described in each
Duan Jinhang classifies, and obtains first category line segment and second category line segment;
Step 4.3 is unsatisfactory for formula when the corresponding first category line segment of the first three-dimensional planar and the second category line segment:
l(c_1)+l(c_2)>When 0.5*l (all), then its corresponding first three-dimensional planar is deleted, wherein l (c_1) and l (c_2)
The length summation for indicating all line segments in first category line segment respectively, the length summation of all line segments, l in second category line segment
(all) the length summation of all line segments in the three-dimensional planar is indicated;
Step 4.4, according to each corresponding two-dimensional silhouette of the second three-dimensional planar, one by one determine its corresponding three-dimensional line segment, lead to
Abnormal three-dimensional line segment criterion is crossed, determines and removes the abnormal three-dimensional line segment, obtain the first of each second three-dimensional planar
Three-dimensional line segment group, wherein second three-dimensional planar be the multiple three-dimensional planar in addition to first three-dimensional planar
Three-dimensional planar;
Three-dimensional line segment in the first three-dimensional line segment group of each second three-dimensional planar is passed through the directions z by step 4.5
Normalization, obtains histogram;
Step 4.6 determines each three-dimensional line segment in the first three-dimensional line segment group adjacent thereto three in the histogram
Tie up line segment;
Step 4.7, according to the adjacent three-dimensional line segment and it is default merge criterion, carry out three-dimensional line segment merging, obtain described dissipate
The disorderly new three-dimensional line segment of point cloud.
6. a kind of three-dimensional line segment extraction system of dispersion point cloud, which is characterized in that including:
Three-dimensional planar generation module, after the dispersion point cloud structure k-d tree to input, to each point using it in institute
K point closest around in k-d tree is stated, fitting obtains the planar curvature of the point, and according to the plane of each point
The dispersion point cloud is divided into multiple three-dimensional planars by curvature by region growing and region merging technique;
Three-dimensional line segment generation module, for for the three-dimensional planar generation unit generates for each three-dimensional planar, general
All the points in it project in preset plane, form a bidimensional image, extract the two-dimensional line segment of the bidimensional image, and will
In the two-dimensional line segment back projection to three-dimensional planar, three-dimensional line segment is obtained;
Data processing module, for corresponding by each the described three-dimensional planar generated to the three-dimensional line segment generation unit
The three-dimensional line segment carries out abnormal line segment removal and line segment merges, and obtains the new three-dimensional line segment of the dispersion point cloud.
7. a kind of three-dimensional line segment extraction system of dispersion point cloud according to claim 6, which is characterized in that described three-dimensional flat
Face generation module includes:
Fitting unit, after the dispersion point cloud structure k-d tree to input, to each point using it in the k-dtree
In around closest K point, fitting obtains the planar curvature of the point;
Area generation unit, for according to the planar curvature put of each of fitting unit fitting from small to large suitable
Sequence is ranked up the point in the dispersion point cloud, and since the corresponding point of minimum curvature, successively determine each put and its
Coplanar point obtains multiple cloud sectors domain, wherein each point cloud sector domain is made of coplanar point;
Region merging technique unit, the multiple cloud sector domain for being obtained to the Area generation unit are fitted a normal direction one by one
It measures and determines a fit Plane, each institute is determined by the label to the point setting label in each fit Plane
The adjacent flat of fit Plane is stated, and the corresponding adjacent flat of the fit Plane is merged, obtains multiple three-dimensional planars.
8. a kind of three-dimensional line segment extraction system of dispersion point cloud according to claim 7, which is characterized in that the fitting is single
Member is specifically used for:
Build the corresponding k-d tree of dispersion point cloud;To the every bit p in the dispersion point cloudiIf it is in the k-d tree
The collection that constitute of the closest points of K be combined intoBuild covariance matrix Σ:
In formula, Σ represents 3 × 3 covariance matrix, and K isThe number at midpoint,It isThe average value at midpoint;Solution asks the association
The characteristic value of variance matrix Σ, and using minimal eigenvalue as the point piPlanar curvature
9. a kind of three-dimensional line segment extraction system of dispersion point cloud according to claim 6, which is characterized in that the three-dimensional line
Section generation module is specifically used for:
It calculates each described three-dimensional planar and corresponds to the average value of point set and planar central point is determined according to the average value, by institute
It states on other spot projections to the default two dimensional surface Jing Guo the planar central point concentrated, obtains bidimensional image;It will be described
Bidimensional image is converted into binary picture, and image contour is extracted on the binary picture, recycles least square fitting
Extract the two-dimensional line segment in the image contour;By in the two-dimensional line segment back projection to its corresponding three-dimensional planar, three are obtained
Tie up line segment.
10. according to a kind of three-dimensional line segment extraction system of dispersion point cloud of claim 6 to 8 any one of them, which is characterized in that
The data processing module is specifically used for:
All three-dimensional line segments in three-dimensional planar described in each are ranked up according to length;According to it is described sequence and it is preset
Classifying rules, the three-dimensional line segment corresponding to three-dimensional planar described in each are classified, and first category line segment and are obtained
Two classification line segments;When the corresponding first category line segment of the first three-dimensional planar and the second category line segment are unsatisfactory for formula:l
(c_1)+l(c_2)>When 0.5*l (all), then its corresponding first three-dimensional planar is deleted, wherein l (c_1) and l (c_2)
The length summation for indicating all line segments in first category line segment respectively, the length summation of all line segments, l in second category line segment
(all) the length summation of all line segments in the three-dimensional planar is indicated;According to each corresponding two-dimensional silhouette of the second three-dimensional planar,
Its corresponding three-dimensional line segment is determined one by one, by abnormal three-dimensional line segment criterion, is determined and is removed the abnormal three-dimensional line segment, obtain every
First three-dimensional line segment group of one second three-dimensional planar, wherein second three-dimensional planar is the multiple three-dimensional planar
In three-dimensional planar in addition to first three-dimensional planar;By first three-dimensional line segment of each second three-dimensional planar
Three-dimensional line segment in group is normalized by the directions z, obtains histogram;Determine each three-dimensional line in the first three-dimensional line segment group
Section three-dimensional line segment adjacent thereto in the histogram;According to the adjacent three-dimensional line segment and default merging criterion, carry out
Three-dimensional line segment merges, and obtains the new three-dimensional line segment of the dispersion point cloud.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810295190.2A CN108510516A (en) | 2018-03-30 | 2018-03-30 | A kind of the three-dimensional line segment extracting method and system of dispersion point cloud |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810295190.2A CN108510516A (en) | 2018-03-30 | 2018-03-30 | A kind of the three-dimensional line segment extracting method and system of dispersion point cloud |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108510516A true CN108510516A (en) | 2018-09-07 |
Family
ID=63380190
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810295190.2A Withdrawn CN108510516A (en) | 2018-03-30 | 2018-03-30 | A kind of the three-dimensional line segment extracting method and system of dispersion point cloud |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108510516A (en) |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109512547A (en) * | 2018-12-03 | 2019-03-26 | 刘彩刚 | A kind of breast prosthesis construction method and device based on k-means |
CN110285754A (en) * | 2019-07-02 | 2019-09-27 | 深圳市镭神智能系统有限公司 | Workpiece localization method, device, system and storage medium based on laser scanning |
CN110335282A (en) * | 2018-12-25 | 2019-10-15 | 广州启明星机器人有限公司 | A kind of profile line segment feature extraction algorithm based on grid |
CN110909749A (en) * | 2019-10-24 | 2020-03-24 | 同济大学 | Method for optimizing and extracting geometric linear contour from two-dimensional disordered point set |
CN111210427A (en) * | 2020-01-17 | 2020-05-29 | 云南大学 | Time-change-based method for analyzing post-operation shrinkage of in-vivo light-weight patch |
CN111583098A (en) * | 2020-02-21 | 2020-08-25 | 东南大学 | Line segment clustering and fitting method and system based on sequence image |
CN111815637A (en) * | 2020-06-19 | 2020-10-23 | 浙江大华技术股份有限公司 | Method and device for extracting 3D image edge point cloud and computer storage medium |
CN112053361A (en) * | 2020-10-15 | 2020-12-08 | 南京航空航天大学 | Aircraft skin butt joint detection method based on large-scale point cloud |
CN112164145A (en) * | 2020-10-30 | 2021-01-01 | 武汉大学 | Method for rapidly extracting indoor three-dimensional line segment structure based on point cloud data |
CN112861228A (en) * | 2021-02-18 | 2021-05-28 | 深圳辰视智能科技有限公司 | Large-scale steel structure positioning method and system |
CN112967311A (en) * | 2019-12-12 | 2021-06-15 | 浙江商汤科技开发有限公司 | Three-dimensional line graph construction method and device, electronic equipment and storage medium |
CN113012161A (en) * | 2021-03-05 | 2021-06-22 | 华南理工大学 | Stacking scattered target point cloud segmentation method based on convex region growth |
CN113283267A (en) * | 2020-02-19 | 2021-08-20 | 广东博智林机器人有限公司 | Minimum closed space extraction method and device based on two-dimensional space |
US11200351B2 (en) * | 2019-06-30 | 2021-12-14 | Huazhong University Of Science And Technology | Method for constructing curve of robot processing path of part with small curvature based on point cloud boundary |
CN116246038A (en) * | 2023-05-11 | 2023-06-09 | 西南交通大学 | Multi-view three-dimensional line segment reconstruction method, system, electronic equipment and medium |
Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100315412A1 (en) * | 2009-06-15 | 2010-12-16 | Microsoft Corporation | Piecewise planar reconstruction of three-dimensional scenes |
CN101958006A (en) * | 2010-09-03 | 2011-01-26 | 南京大学 | X-ray image-based three-dimensional object imaging method |
CN103413133A (en) * | 2013-06-28 | 2013-11-27 | 广东电网公司电力科学研究院 | Automatically-extracting power line method in random laser point cloud data |
CN104408689A (en) * | 2014-12-10 | 2015-03-11 | 武汉大学 | Holographic-image-based streetscape image fragment optimization method |
CN104463851A (en) * | 2014-11-19 | 2015-03-25 | 哈尔滨工业大学深圳研究生院 | Automatic shoe sole edge line tracking method based on robot |
CN104463871A (en) * | 2014-12-10 | 2015-03-25 | 武汉大学 | Streetscape facet extraction and optimization method based on vehicle-mounted LiDAR point cloud data |
CN104952107A (en) * | 2015-05-18 | 2015-09-30 | 湖南桥康智能科技有限公司 | Three-dimensional bridge reconstruction method based on vehicle-mounted LiDAR point cloud data |
CN105488847A (en) * | 2015-12-11 | 2016-04-13 | 中国测绘科学研究院 | Airborne laser radar point cloud based three-dimensional reconstruction method for single-stage single power lead wire |
CN105957076A (en) * | 2016-04-27 | 2016-09-21 | 武汉大学 | Clustering based point cloud segmentation method and system |
CN106485680A (en) * | 2016-10-13 | 2017-03-08 | 上海联影医疗科技有限公司 | Method for correcting image and device |
CN106570507A (en) * | 2016-10-26 | 2017-04-19 | 北京航空航天大学 | Multi-angle consistent plane detection and analysis method for monocular video scene three dimensional structure |
CN107392964A (en) * | 2017-07-07 | 2017-11-24 | 武汉大学 | The indoor SLAM methods combined based on indoor characteristic point and structure lines |
CN107463918A (en) * | 2017-08-17 | 2017-12-12 | 武汉大学 | Lane line extracting method based on laser point cloud and image data fusion |
-
2018
- 2018-03-30 CN CN201810295190.2A patent/CN108510516A/en not_active Withdrawn
Patent Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100315412A1 (en) * | 2009-06-15 | 2010-12-16 | Microsoft Corporation | Piecewise planar reconstruction of three-dimensional scenes |
CN101958006A (en) * | 2010-09-03 | 2011-01-26 | 南京大学 | X-ray image-based three-dimensional object imaging method |
CN103413133A (en) * | 2013-06-28 | 2013-11-27 | 广东电网公司电力科学研究院 | Automatically-extracting power line method in random laser point cloud data |
CN104463851A (en) * | 2014-11-19 | 2015-03-25 | 哈尔滨工业大学深圳研究生院 | Automatic shoe sole edge line tracking method based on robot |
CN104408689A (en) * | 2014-12-10 | 2015-03-11 | 武汉大学 | Holographic-image-based streetscape image fragment optimization method |
CN104463871A (en) * | 2014-12-10 | 2015-03-25 | 武汉大学 | Streetscape facet extraction and optimization method based on vehicle-mounted LiDAR point cloud data |
CN104952107A (en) * | 2015-05-18 | 2015-09-30 | 湖南桥康智能科技有限公司 | Three-dimensional bridge reconstruction method based on vehicle-mounted LiDAR point cloud data |
CN105488847A (en) * | 2015-12-11 | 2016-04-13 | 中国测绘科学研究院 | Airborne laser radar point cloud based three-dimensional reconstruction method for single-stage single power lead wire |
CN105957076A (en) * | 2016-04-27 | 2016-09-21 | 武汉大学 | Clustering based point cloud segmentation method and system |
CN106485680A (en) * | 2016-10-13 | 2017-03-08 | 上海联影医疗科技有限公司 | Method for correcting image and device |
CN106570507A (en) * | 2016-10-26 | 2017-04-19 | 北京航空航天大学 | Multi-angle consistent plane detection and analysis method for monocular video scene three dimensional structure |
CN107392964A (en) * | 2017-07-07 | 2017-11-24 | 武汉大学 | The indoor SLAM methods combined based on indoor characteristic point and structure lines |
CN107463918A (en) * | 2017-08-17 | 2017-12-12 | 武汉大学 | Lane line extracting method based on laser point cloud and image data fusion |
Non-Patent Citations (2)
Title |
---|
欧元汉等: "基于密集点云重投影的三维线段模型重建算法", 《计算机工程与设计》 * |
牛长锋等: "基于SIFT特征和粒子滤波的目标跟踪方法", 《机器人》 * |
Cited By (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109512547A (en) * | 2018-12-03 | 2019-03-26 | 刘彩刚 | A kind of breast prosthesis construction method and device based on k-means |
CN110335282A (en) * | 2018-12-25 | 2019-10-15 | 广州启明星机器人有限公司 | A kind of profile line segment feature extraction algorithm based on grid |
CN110335282B (en) * | 2018-12-25 | 2023-04-18 | 广州启明星机器人有限公司 | Contour line segment feature extraction method based on grids |
US11200351B2 (en) * | 2019-06-30 | 2021-12-14 | Huazhong University Of Science And Technology | Method for constructing curve of robot processing path of part with small curvature based on point cloud boundary |
CN110285754A (en) * | 2019-07-02 | 2019-09-27 | 深圳市镭神智能系统有限公司 | Workpiece localization method, device, system and storage medium based on laser scanning |
CN110909749A (en) * | 2019-10-24 | 2020-03-24 | 同济大学 | Method for optimizing and extracting geometric linear contour from two-dimensional disordered point set |
CN110909749B (en) * | 2019-10-24 | 2022-06-14 | 同济大学 | Method for optimizing and extracting geometric linear contour from two-dimensional disordered point set |
CN112967311A (en) * | 2019-12-12 | 2021-06-15 | 浙江商汤科技开发有限公司 | Three-dimensional line graph construction method and device, electronic equipment and storage medium |
CN111210427A (en) * | 2020-01-17 | 2020-05-29 | 云南大学 | Time-change-based method for analyzing post-operation shrinkage of in-vivo light-weight patch |
CN111210427B (en) * | 2020-01-17 | 2023-03-10 | 云南大学 | Time-change-based method for analyzing post-operation shrinkage of in-vivo light-weight patch |
CN113283267A (en) * | 2020-02-19 | 2021-08-20 | 广东博智林机器人有限公司 | Minimum closed space extraction method and device based on two-dimensional space |
CN111583098A (en) * | 2020-02-21 | 2020-08-25 | 东南大学 | Line segment clustering and fitting method and system based on sequence image |
CN111815637A (en) * | 2020-06-19 | 2020-10-23 | 浙江大华技术股份有限公司 | Method and device for extracting 3D image edge point cloud and computer storage medium |
US11532121B2 (en) | 2020-10-15 | 2022-12-20 | Nanjing University Of Aeronautics And Astronautics | Method for measuring a seam on aircraft skin based on large-scale point cloud |
CN112053361A (en) * | 2020-10-15 | 2020-12-08 | 南京航空航天大学 | Aircraft skin butt joint detection method based on large-scale point cloud |
CN112164145A (en) * | 2020-10-30 | 2021-01-01 | 武汉大学 | Method for rapidly extracting indoor three-dimensional line segment structure based on point cloud data |
CN112861228A (en) * | 2021-02-18 | 2021-05-28 | 深圳辰视智能科技有限公司 | Large-scale steel structure positioning method and system |
CN112861228B (en) * | 2021-02-18 | 2024-02-13 | 深圳辰视智能科技有限公司 | Positioning method and system for large-scale steel structure |
CN113012161A (en) * | 2021-03-05 | 2021-06-22 | 华南理工大学 | Stacking scattered target point cloud segmentation method based on convex region growth |
CN116246038A (en) * | 2023-05-11 | 2023-06-09 | 西南交通大学 | Multi-view three-dimensional line segment reconstruction method, system, electronic equipment and medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108510516A (en) | A kind of the three-dimensional line segment extracting method and system of dispersion point cloud | |
Dong et al. | An efficient global energy optimization approach for robust 3D plane segmentation of point clouds | |
Bazazian et al. | Fast and robust edge extraction in unorganized point clouds | |
CN107292276B (en) | Vehicle-mounted point cloud clustering method and system | |
CN107742102B (en) | Gesture recognition method based on depth sensor | |
CN104268583B (en) | Pedestrian re-recognition method and system based on color area features | |
CN110222642B (en) | Plane building component point cloud contour extraction method based on global graph clustering | |
WO2012111236A1 (en) | Image identification device and program | |
CN109034065B (en) | Indoor scene object extraction method based on point cloud | |
CN103473551A (en) | Station logo recognition method and system based on SIFT operators | |
Yang et al. | Efficient plane extraction using normal estimation and RANSAC from 3D point cloud | |
CN112767485A (en) | Point cloud map creating and scene identifying method based on static semantic information | |
CN111145129A (en) | Point cloud denoising method based on hyper-voxels | |
CN108182705A (en) | A kind of three-dimensional coordinate localization method based on machine vision | |
Li et al. | A branch-trunk-constrained hierarchical clustering method for street trees individual extraction from mobile laser scanning point clouds | |
CN113012161A (en) | Stacking scattered target point cloud segmentation method based on convex region growth | |
CN110348478B (en) | Method for extracting trees in outdoor point cloud scene based on shape classification and combination | |
Chen et al. | A local tangent plane distance-based approach to 3D point cloud segmentation via clustering | |
CN115018999A (en) | Multi-robot-cooperation dense point cloud map construction method and device | |
Hao et al. | Structure-based object detection from scene point clouds | |
Sun et al. | Automated segmentation of LiDAR point clouds for building rooftop extraction | |
Omidalizarandi et al. | Segmentation and classification of point clouds from dense aerial image matching | |
CN111311643B (en) | Video target tracking method using dynamic search | |
CN111462123B (en) | Point cloud data segmentation method based on spectral clustering | |
Cupec et al. | Segmentation of depth images into objects based on local and global convexity |
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 | ||
WW01 | Invention patent application withdrawn after publication | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20180907 |