CN106023312A - Automatic 3D building model reconstruction method based on aviation LiDAR data - Google Patents
Automatic 3D building model reconstruction method based on aviation LiDAR data Download PDFInfo
- Publication number
- CN106023312A CN106023312A CN201610319784.3A CN201610319784A CN106023312A CN 106023312 A CN106023312 A CN 106023312A CN 201610319784 A CN201610319784 A CN 201610319784A CN 106023312 A CN106023312 A CN 106023312A
- Authority
- CN
- China
- Prior art keywords
- point
- garret
- lidar
- roof
- building
- 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 54
- 238000005457 optimization Methods 0.000 claims abstract description 9
- 238000012952 Resampling Methods 0.000 claims description 16
- 230000011218 segmentation Effects 0.000 claims description 16
- 239000000284 extract Substances 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 5
- 238000012706 support-vector machine Methods 0.000 claims description 4
- 230000015572 biosynthetic process Effects 0.000 claims description 3
- 238000001914 filtration Methods 0.000 abstract description 2
- 230000000877 morphologic effect Effects 0.000 abstract description 2
- 238000009499 grossing Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 6
- 239000010410 layer Substances 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 239000011229 interlayer Substances 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011017 operating method Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000009466 transformation Effects 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
- G06T17/05—Geographic models
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Remote Sensing (AREA)
- Computer Graphics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Processing Or Creating Images (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an automatic 3D building model reconstruction method based on aviation LiDAR data. The method comprises steps of: extracting a building roof point cloud from the aviation LiDAR data by using reverse iteration mathematic morphological filtering and a method based on point cloud density; extracting and optimizing roof surface patches according to a strategy of seed area selection-roof surface patch growth-surface path smoothing and optimization; constructing 2D regular grids to resample different roof layers to obtain the internal points and the edge points of the roof layers; optimizing the edge points of different roof layers; connecting the internal points and the edge points of the roof layers to construct a building roof surface and wall surface so as to finally achieve 3D model reconstruction of the building roof. It has been proved by practices that the method may reconstruct the 3D model of a building roof, provides a new idea for connection among different roof layers, and has high 3D model reconstruction precision.
Description
Technical field
The present invention relates to a kind of three-dimensional building object model automatic reconstruction method based on aviation LiDAR data, particularly relate to
And a kind of three-dimensional building object model automatic reconstruction method taking smooth strategy and interlayer to connect based on aviation LiDAR data.
Background technology
Three-dimensional building object model is the important expression means of building three dimensional structure information, urban planning, disaster monitoring,
The fields such as communications facility construction and digital city have application widely.Three-dimensional model building is rebuild as digital city
Most important and the most challenging task in construction, its research is paid close attention to the most greatly in the past few decades.Utilize aviation
Stereogram obtains three-dimensional model building, is a kind of traditional photogrammetric survey method, still uses a large amount of, but needs more
Manual intervention, automaticity is the highest.
In recent years, the development of aviation LiDAR technology is exceedingly fast, and aviation LiDAR data has been widely used in surface exploration
(Vosselman G, 2005), feature detection (Tong L H, 2013), Objects extraction (Boyko A, 2011), threedimensional model weight
Build aspects such as (Cheng L, 2012), show huge application prospect.Therefore, aviation LiDAR technology is the three of building
Dimension reconstruction provides a kind of alternative.Aviation LiDAR equipment can directly obtain the three-dimensional information of ground target, can improve automatically
Change level.But, a large amount of irregular point data bring new challenge also to the Model Reconstruction of building.To this end, it is the most effective
Utilize the advantage of aviation LiDAR data, it is achieved automatization's high-quality of three-dimensional model building is rebuild, be still that one is worth deep
Enter the proposition of research.
Although the research utilizing aviation LiDAR data to carry out three-dimensional model building reconstruction gets more and more, but most of
Method be by extract building roof contour line realize final Model Reconstruction.Zhou Q Y, Cheng L, Susaki J,
Cheng Liang et al. respectively " 2.5D building modeling by discovering global regularities ",
《Integration of LiDAR data and optical multi-view images for 3D
reconstruction of building roofs》、《Knowledge-based modeling of buildings in
Dense urban areas by combining airborne LiDAR data and aerial images ", " integrated many
Rebuild three-dimensional building object model depending on aviation image and LiDAR data " etc. in article, propose to use profile reconstruction building three-dimensional
The method of model.But realize reconstructing three-dimensional model by contour line and there are some inherent defects: 1) influence of noise.The existence of noise
Imperfect by making to extract the contour line obtained, even if forming integrity profile line by the method for matching, with actual profile line phase
Than then there is bigger error.2) dot density impact.The size of dot density directly affects the extraction of contour line, and dot density is less, carries
Obtain the contour line arrived by imperfect.3) interlayer annexation.Extract the different garret contour lines obtained and remain discrete shape
State, the annexation being difficult between determining.
Before extracting contour of building line, the segmentation of building roof dough sheet to be realized.Currently, to building
The dividing method of roof dough sheet can be largely classified into two big classes: 1) model driven method.2) data-driven method.Mass
(1999), Oude (2009), Hebel (2012), Susaki (2013), Huang (2013), Henn (2013) et al. use respectively
First method realizes building roof patch division.The thinking of this method is first to determine to treat experimental architecture thing roof class
Type, then determines the roof model needing to use in building roof model database.But the dough sheet that this method obtains
Segmentation result accuracy can not get ensureing, when when experimental architecture thing roof type is beyond roof type set in advance, splits
Relatively serious mistake will be there is in result.The thinking of second method is to extract roof dough sheet with data-driven, the most not to room completely
Top type does any hypothesis.2012, Chen D et al. utilized progressive morphologic filtering, algorithm of region growing and adaptive
Stochastical sampling consistency algorithm completes the segmentation of building roof dough sheet；2014, Fan H C et al. proposed a kind of based on ridge
The roof dough sheet dividing method of line hierachical decomposition, utilizes the feature of connectivity and coplanarity, realizes roof dough sheet along ridge line
Segmentation；2014, original aerial LiDAR point cloud was divided into ground point and non-ground points by Awrangjeb M et al., for non-ground
Point uses the coplanarity of point and the local feature of point to complete the segmentation of plane roof.But these methods are affected relatively by initial data
Greatly, there is not annexation between the roof dough sheet obtained, and the profile ratio of roof dough sheet is more tortuous.
Carrying out the fast automatic three-dimensional modeling of building roof by aviation LiDAR data is automatic modeling developing direction.So
Although and present stage is more to the research of this kind of method, but all there is certain problem, how to make full use of aviation LiDAR number
According to advantage, and realize accuracy, automatic three-dimensional reconstruction that degree of accuracy is higher still needs further to be studied.
Summary of the invention
The technical problem to be solved in the present invention is: more by extracting contour of building for existing structure method for reconstructing
Line realizes, and contour line extracts the more doubt problem of topological relation between defect and the contour line existed, and the present invention carries
For a kind of three-dimensional building object model automatic reconstruction method based on aviation LiDAR data, the method ensure that roof dough sheet divides
The accuracy cut, and use internal point that garret resampling obtains and marginal point to achieve different garret fast and efficiently
Between connection, rebuild the threedimensional model of building accurately, solved annexation between different garret more difficult really
Fixed problem.
A kind of based on aviation LiDAR data the three-dimensional building object model automatic reconstruction method that the present invention provides, step is such as
Under:
The first step, building roof data reduction extract building object point cloud, reject the some cloud on building wall,
To building roof point cloud；
Second step, segmentation roof dough sheet carry out roof patch division to building roof point cloud, then to the room obtained
Top sheet carries out smooth；
3rd step, merging roof dough sheet are when two roof dough sheets are neighbouring and place plane exists intersection, then by this
Two roof dough sheets merge formation garret；
4th step, garret resampling carry out resampling to all garrets, it is thus achieved that the internal point of garret and limit
Edge point；
Internal point and marginal point that 5th step, building model are rebuild belonging to same garret build the triangulation network,
Form deck；To belonging to adjacent floor and the marginal point structure triangulation network being positioned on same perpendicular, form deck and build
Build thing metope, be finally completed three-dimensional model building and rebuild.
The present invention also has a further feature:
1, in the described first step, the some cloud method on building wall is rejected as follows: with r as radius, to each LiDAR
Point searches neighborhood point, with the point searching the quantity of neighborhood point that the obtains volume divided by neighborhood region and obtaining each LiDAR point
Cloud density, when the some cloud density of LiDAR point is less than when specifying threshold value, and this LiDAR point is metope point, rejects.
2, in described second step, selected seed region, use algorithm of region growing to realize roof patch division, seed region
Choosing method as follows: first estimate the normal vector of each LiDAR point, as a example by calculating some p, find centered by a p, half
Point set N in the range of the r of footpathp, it is calculated three eigenvalue λ according to formula (1) and (2)1、λ2、λ3, minimum in three eigenvalues
Eigenvalue λminCorresponding characteristic vector is the normal vector of a p.
Q in formulai∈Np, n is point set NpThe quantity at midpoint, CpIt is the covariance matrix of a p, according to formula
Calculate the curvature of some p, when the curvature of a pLess than designated curvature threshold value λTTime, so that it may think the neighborhood point N of a ppPut down at one
On face, choose curvatureCorresponding neighborhood point NpSeed region as satisfied requirement.
3, in described second step, algorithm of region growing is it needs to be determined that two standards: one is the number of intra-office point, another
It it is the standard deviation of fit Plane；When the distance of LiDAR point to seed region place plane is less than distance to a declared goal threshold value, it is believed that be
Intra-office point；After calculating a little, then calculate the standard deviation of intra-office point, when standard deviation is less than when specifying threshold value, it is believed that meet
Roof patch division requirement.
4, in described second step, it is smooth smooth with local projecting point cloud that the smooth optimization of dough sheet includes splitting dough sheet point cloud:
Smooth for segmentation dough sheet point cloud, use least square fitting to go out to split dough sheet point cloud equation, then according to side
Point is smoothed on matching dough sheet by journey, and fit Plane equation calculates according to formula (3):
X in formulai, yi, ziRepresent the three-dimensional coordinate of point, a0, a1, a2Represent the coefficient of fit Plane equation；
Smooth for local projecting point cloud, utilizes distance threshold to scan for local projecting point cloud, calculate point
Point in face sheet, to the distance of segmentation dough sheet, when distance is less than threshold value, is designated as required point, is obtained according to above-mentioned by required point
Fit Plane equation be smoothed to split on dough sheet.
5, in described 4th step, garret method for resampling is as follows: first project some cloud to X/Y plane, and to X/Y plane
Build two dimension grid, then make the following judgment:
A (), LiDAR point in the LiDAR point in four neighborhood grid units and central square net unit belong to same room
Top layer, now the center of central square net unit is as the internal point of garret, centered by the height of this internal point in grid unit
The average height of LiDAR point；
(b), when in some the neighborhood grid unit in four neighborhood grid units of center grid unit LiDAR point belong to
When two or more garrets, if most two garret of counting in this neighborhood grid unit is respectively garret j
With garret k, obtain the LiDAR point of garret j in this neighborhood grid unit and the cut-off rule of the LiDAR point of garret k, then
Calculate this cut-off rule and central square net unit center line intersecting point coordinate (x in this neighborhood grid unit0, y0), then (x0, y0, zj)
For the marginal point of garret j, zjFor the average height of the LiDAR point of garret j, then (x0, y0, zk) it is the edge of garret k
Point, zkAverage height for the LiDAR point of garret k；
(c), when in some the neighborhood grid unit in the LiDAR point in the grid unit of center and four neighborhood grid units
LiDAR point when being belonging respectively to two different garrets, if two different garrets are garret j and garret k, calculate center
Middle point coordinates (the x in the grid sideline between grid unit and described neighborhood grid unit1, y1), then (x1, y1, zj) it is garret j
Marginal point, zjFor the average height of the LiDAR point of garret j, then (x1, y1, zk) it is the marginal point of garret k, zkFor roof
The average height of the LiDAR point of layer k.
6, size is LiDAR point cloud average headway 2 times of described two dimension grid.
7, in described 4th step, the LiDAR point of garret j in neighborhood grid unit and the LiDAR point of garret k are obtained
The method of cut-off rule is: the LiDAR point of garret j in neighborhood grid unit and the LiDAR point of garret k are considered as two classes
Not, algorithm of support vector machine is used to obtain cut-off rule.
Beneficial effects of the present invention is as follows:
A (), the present invention propose the strategy of " seed region chooses the roof dough sheet smooth optimization of aufwuchsplate sheet ", this strategy
While extracting roof dough sheet, also roof dough sheet is carried out smooth optimization, follow-up reconstructing three-dimensional model.
B (), the present invention utilize garret resampling to be calculated internal point and marginal point, it is to avoid traditional method for extracting room
After the contour line of top, it is more difficult to the problem determining topological relation between differently contoured line.And the calculating of internal point and marginal point is efficient,
Convenient, it is to avoid loaded down with trivial details work, improve automatization level.
In sum, the present invention discloses a kind of based on aviation LiDAR data taking and smooths strategy and the three of interlayer connection
Dimension building model automatic reconstruction method, the method is closed by building roof data reduction, roof patch division, roof dough sheet
And, garret resampling, building model rebuild five steps and achieve the automatic three-dimensional reconstruction of building roof.Test card
Bright, the method has higher reconstruction precision, can reflect the geometry feature of top of building well.
Accompanying drawing explanation
Below in conjunction with the accompanying drawings the inventive method is further described.
Fig. 1 is the overview flow chart of the embodiment of the present invention.
Fig. 2 is embodiment of the present invention region aviation LiDAR data schematic diagram.
Fig. 3 is embodiment of the present invention dough sheet planarization procedure schematic diagram.
Fig. 4 is embodiment of the present invention internal point and marginal point schematic diagram.
Fig. 5 a～Fig. 5 e is embodiment of the present invention internal point and marginal point calculating schematic diagram.
Fig. 6 is embodiment of the present invention internal point and marginal point final result schematic diagram.
Fig. 7 is the embodiment of the present invention whole regional model reconstructed results figure.
Fig. 8 a～Fig. 8 f is 6 solitary building Model Reconstruction result figure in the embodiment of the present invention.
Detailed description of the invention
Elaborate the present invention below according to accompanying drawing, make the technology path of the present invention and operating procedure become apparent from.
The aviation LiDAR data using Optech ALTM Gemini laser scanner to acquire is experimental data, as
Shown in Fig. 2.Aviation LiDAR data equalization point spacing 0.4m, height accuracy 15cm, plane precision 30cm.
The present embodiment is that a kind of three-dimensional building object model automatic reconstruction method based on aviation LiDAR data (be shown in by flow chart
Fig. 1), comprise the following steps:
The type of ground objects that the first step, building roof data reduction original aerial LiDAR data are comprised is mainly
Building, vegetation and ground three class.The present invention filters based on the inverse iteration mathematical morphology in patent CN201110432421.8
Ripple algorithm extracts building LiDAR point cloud.
Method is described as follows: the original LiDAR point cloud of random distribution is carried out resampling, and (resampling spacing is set as
1m), the equidistantly point after resampling is carried out inverse iteration mathematical morphology filter, is gradually reduced spectral window with less step-length
The size of mouth, uses morphology "ON" to operate each window, the poorest to the filter result of adjacent two windows,
When difference is more than the height of minimum building, corresponding point is flagged as non-ground points.Extract the non-ground points bag obtained
Including construction zone and intensive trees region, followed by a roughness for cloud level journey (variance of elevation), setting threshold value will be thick
The tree point cloud that rugosity is higher is rejected.Finally, building roof LiDAR point cloud is extracted.In the present embodiment, the maximal window of selection
Mouth is 106m, and minimum window is 6m, and it is 10m that window reduces step-length, and the height of minimum building is 3m, and elevation variance threshold values is
0.4m。
In extracting the building LiDAR point cloud obtained, however it remains metope point cloud, the present invention uses side based on density
Method rejects metope point cloud.
Method is described as follows: for a some p, can obtain the neighbor point N of a p in r radiusp, close for a p
Degree can be defined as NpThe quantity at midpoint is divided by the volume of r radius.When the density of a p is less than density threshold, it is considered as
Point p is metope point.Otherwise, then it is building roof point.In the present embodiment, r radius size is 1m, density threshold be 2 points/
m3。
After second step, segmentation roof dough sheet extract building roof LiDAR point cloud, need to building not
Logical roof dough sheet extracts.Roof dough sheet extracts and uses " seed region chooses the roof dough sheet smooth optimization of aufwuchsplate sheet "
Strategy, first pass through the estimation selected seed region of a curvature, then use algorithm of region growing to realize dough sheet growth, finally
The roof dough sheet obtained is carried out smooth optimization, as shown in Figure 3.
Smooth the specifically comprising the following steps that of building roof dough sheet
A1), utilize Curvature Estimate selected seed region for the Curvature Estimate of point, it is necessary first to the normal direction of estimation point
Amount.The such as normal vector of estimation point p, it is thus necessary to determine that the neighborhood point N of some pp=q | (p, q) < r}, wherein r is search neighbour for q ∈ P, d
The search radius of territory point, which determines the size of neighborhood.Method a little can be obtained by calculating the covariance matrix of field point
Vector, covariance matrix is defined as follows:
Q in formulai∈Np, n is point set NpThe quantity at midpoint, CpCovariance matrix for a p.By covariance square above
Battle array can be calculated three eigenvalues, respectively λ1、λ2、λ3.Assume λ1<λ2<λ3, the minimum feature corresponding to eigenvalue to
Amount is just for the normal vector of some p.Curvature by calculating point comes selected seed region the most again, and curvature is defined as follows:
WhenLess than curvature threshold λTTime, it is possible to think a p neighborhood point in one plane, choose less curvatureInstitute
Corresponding neighborhood point is as the seed region of satisfied requirement.In the present embodiment, curvature threshold is 0.005.
A2), dough sheet growth seed region obtain after, i.e. obtain initial plane parameter.Here definition two meets
The standard of dough sheet growth: one is the number of intra-office point, and another is the standard deviation of segmentation plane.As a piArrive initial plane
When distance is less than distance threshold, it is believed that some piBelong to the plane at initial plane place, be intra-office point, when point does not meets again
When the distance of initial plane is less than distance threshold, dough sheet growth is complete, the number of point in statistics bureau.And calculate the face that growth obtains
The standard deviation of sheet, when standard deviation is less than when setting threshold value, it is believed that roof dough sheet has extracted, and meets extraction requirement.This enforcement
In example, distance threshold is 0.5m, and standard deviation is 0.95m.
A3), dough sheet is smooth optimizes mainly for 2 points: one is the smooth of the three-dimensional point set to segmentation dough sheet；
Two is that thin portion is disturbed the smooth of information.
The present embodiment is in above-mentioned steps a3) in the concrete grammar of the smooth optimization of dough sheet as follows:
B1), three-dimensional point set smooth in a practical situation, even plane roof, some cloud be also not entirely in
On one two dimensional surface, i.e. extracting on the roof dough sheet obtained, there is fluctuation in point.Plane equation z=a can be used0x+a1y+
a2It is fitted, for parameter a0, a1, a2Calculating following system of linear equations can be used to calculate:
After obtaining plane equation, the point on segmentation dough sheet can be smoothed to same flat according to the plane equation obtained
On face.
B2), thin portion interference information smooth for the roof dough sheet that obtains of extraction, there is a cloud disappearance, this portion
Branch cloud needs to be flattened on corresponding roof dough sheet, upwards finds for segmentation dough sheet known to each and carries at roof dough sheet
Taking non-existent LiDAR point in result, after finding, the plane equation according to corresponding roof dough sheet is smoothed on roof dough sheet.
Upwards finding distance a little in the present embodiment is 2m.
3rd step, merging roof dough sheet, before garret resampling, need all roofs dough sheet according to necessarily
Rule merges to form garret.When two roof dough sheets meet two conditions: dough sheet place, (1) two roof plane is deposited
At intersection；2), when two roof dough sheets are mutually adjacent, two roof dough sheets can be merged into a garret.
4th step, garret resampling calculate internal point and the marginal point of garret, and k garret is calculated k
Internal point or marginal point, this k point has identical x, y-coordinate, and different height values is illustrated in figure 4 internal point and limit
Edge point schematic diagram.Garret resampling is carried out based on two dimension regular grid.
Specifically comprising the following steps that of garret resampling
C1), two dimension regular grid build the size of two dimension regular grid by LiDAR point cloud in the x and y direction
Four to coordinate determines, in this example, grid unit size is 1m.Calculate each some grid unit belonging to correspondence, finally build
Vertical Grid Index.
C2) there is one and the internal point of above quantity during, internal point and marginal point calculate each grid unit
Or marginal point, during calculating, needs to consider four neighborhood grid units of this grid unit.Internal point and marginal point
Calculating have five kinds of situations, 1. the point in 5 grid units broadly falls into same garret, 2. there is the neighborhood of different garret
On the left side of central square net unit or the right, 3. there is the neighborhood grid unit of different garret at center grid list in grid unit
The top of unit or bottom, 4. the horizontal level of marginal point is on the border of central square net unit, 5. exists in neighborhood grid unit
Belong to plural different garret.
C3), different garret marginal points optimize the optimization employing principal component analytical method of marginal point.First, pass through
Original building roof point cloud boundary point is calculated the principal direction of building, then by the method for iteration by marginal point matching
On the straight line that principal direction is corresponding.The internal point of the building roof layer finally given and marginal point are as shown in Figure 6.
For c2) in the specifically comprising the following steps that of five kinds of different situations
D1), broadly fall into same garret for the point in situation 1. (Fig. 5 a) 5 grid units, illustrate not deposit
At metope.The center of central square net unit is defined as the x of central square net unit internal point, y-coordinate, and the height of this internal point is
The average height of LiDAR point in central square net unit.
D2), there is the LiDAR point in some neighborhood grid unit for situation 2. (Fig. 5 b) and belong to two differences
Garret, and neighborhood grid unit location is on the left side of central square net unit or the right.If two different garrets
For garret j and garret k.Firstly, it is necessary to use algorithm of support vector machine to be calculated two garrets on two dimensional surface
Cut-off rule.Then, using the intersection point of calculated cut-off rule and the horizontal central line of central square net unit as two garrets
The x of marginal point, y-coordinate (x0, y0), then (x0, y0, zj) it is the marginal point of garret j, zjLiDAR point average for garret j
Highly, then (x0, y0, zk) it is the marginal point of garret k, zkAverage height for the LiDAR point of garret k.
D3), there is the LiDAR point in some neighborhood grid unit for situation 3. (Fig. 5 c) and belong to two differences
Garret, and neighborhood grid unit location is in the top of central square net unit or bottom.If two different garrets
For garret j and garret k.In like manner, it is necessary first to use algorithm of support vector machine to be calculated two rooms on two dimensional surface
The cut-off rule of top layer.Then, using the intersection point of calculated cut-off rule and the vertical center line of central square net unit as two rooms
The x of top edge point, y-coordinate.The height of each garret marginal point is the average of corresponding garret LiDAR point in this grid unit
Highly.
D4), in the LiDAR point in situation 4. (Fig. 5 d) central square net unit and neighborhood grid unit
LiDAR point is just at two different garrets.If two different garrets are garret j and garret k, calculate central square
Middle point coordinates (the x in the grid sideline between net unit and described neighborhood grid unit1, y1), then (x1, y1, zj) it is garret j's
Marginal point, zjFor the average height of the LiDAR point of garret j, then (x1, y1, zk) it is the marginal point of garret k, zkFor garret k
The average height of LiDAR point.
D5), belong to plural for situation 5. (Fig. 5 e) when the LiDAR point in some neighborhood grid unit
Garret, determines the x of the marginal point of different garret with two garrets that LiDAR point number is most, and y-coordinate, concrete condition can
With reference to d2) or d3).The height of each garret marginal point is the average height of corresponding garret LiDAR point in this grid unit.
4th step, the reconstruction of building model reconstruction model include the reconstruction in building roof face and the weight of metope
Build.
Rebuild for building model, specifically comprise the following steps that
E1), deck is rebuild and the internal point and marginal point belonging to same garret is built the triangulation network, formation roof
Face.
E2), metope is rebuild belonging to adjacent floor layer and the marginal point structure triangle being positioned on same perpendicular
Net, forms deck building wall, is finally completed three-dimensional model building and rebuilds.
The present embodiment is with building roof aviation LiDAR data as actual value.Calculate with whole region and 6 solitary buildings
The precision of reconstruction model is estimated by the offset distance between model and true point.Whole regional model is as it is shown in fig. 7,6
Solitary building model is as shown in Figure 8.
Table 1 building model evaluation result
From the point of view of the statistical result of table 1, the building model from single 6 solitary building models and whole test block is with true
From the point of view of real-valued skew average distance, mostly it is in about 0.04m, and the percentage ratio less than 0.3m is more than 96%, furtherly
The bright present invention has degree of precision to the reconstructing three-dimensional model of building roof.
In addition to the implementation, the present invention can also have other embodiments.All employing equivalents or equivalent transformation shape
The technical scheme become, all falls within the protection domain of application claims.
Claims (8)
1. a three-dimensional building object model automatic reconstruction method based on aviation LiDAR data, step is as follows:
The first step, building roof data reduction extract building object point cloud, reject the some cloud on building wall, are built
Build thing roof point cloud；
Second step, segmentation roof dough sheet carry out roof patch division to building roof point cloud, then to the deck obtained
Sheet carries out smooth；
3rd step, merging roof dough sheet are when two roof dough sheets are neighbouring and place plane exists intersection, then by the two
Roof dough sheet merges formation garret；
4th step, garret resampling carry out resampling to all garrets, it is thus achieved that the internal point of garret and marginal point；
Internal point and marginal point that 5th step, building model are rebuild belonging to same garret build the triangulation network, are formed
Deck；To belonging to adjacent floor layer and the marginal point structure triangulation network being positioned on same perpendicular, form deck building
Thing metope, is finally completed three-dimensional model building and rebuilds.
A kind of three-dimensional building object model automatic reconstruction method based on aviation LiDAR data the most according to claim 1, its
It is characterised by: in the described first step, rejects the some cloud method on building wall as follows: with r as radius, to each LiDAR
Point searches neighborhood point, with the point searching the quantity of neighborhood point that the obtains volume divided by neighborhood region and obtaining each LiDAR point
Cloud density, when the some cloud density of LiDAR point is less than when specifying threshold value, and this LiDAR point is metope point, rejects.
A kind of three-dimensional building object model automatic reconstruction method based on aviation LiDAR data the most according to claim 1, its
It is characterised by: in described second step, selected seed region, uses algorithm of region growing to realize roof patch division, seed region
Choosing method as follows: first estimate the normal vector of each LiDAR point, as a example by calculating some p, find centered by a p, half
Point set N in the range of the r of footpathp, it is calculated three eigenvalue λ according to formula (1) and (2)1、λ2、λ3, minimum in three eigenvalues
Eigenvalue λminCorresponding characteristic vector is the normal vector of a p.
Q in formulai∈Np, n is point set NpThe quantity at midpoint, CpIt is the covariance matrix of a p, according to formula
Calculate the curvature of some p, when the curvature of a pLess than designated curvature threshold value λTTime, so that it may think the neighborhood point N of a ppPut down at one
On face, choose curvatureCorresponding neighborhood point NpSeed region as satisfied requirement.
A kind of three-dimensional building object model automatic reconstruction method based on aviation LiDAR data the most according to claim 3, its
Being characterised by: in described second step, algorithm of region growing is it needs to be determined that two standards: one is the number of intra-office point, another
It it is the standard deviation of fit Plane；When the distance of LiDAR point to seed region place plane is less than distance to a declared goal threshold value, it is believed that be
Intra-office point；After calculating a little, then calculate the standard deviation of intra-office point, when standard deviation is less than when specifying threshold value, it is believed that meet
Roof patch division requirement.
A kind of three-dimensional building object model automatic reconstruction method based on aviation LiDAR data the most according to claim 4, its
Being characterised by: in described second step, it is smooth smooth with local projecting point cloud that the smooth optimization of dough sheet includes splitting dough sheet point cloud:
Smooth for segmentation dough sheet point cloud, use least square fitting to go out to split dough sheet point cloud equation, then will according to equation
Point is smoothed on matching dough sheet, and fit Plane equation calculates according to formula (3):
X in formulai, yi, ziRepresent the three-dimensional coordinate of point, a0, a1, a2Represent the coefficient of fit Plane equation；
Smooth for local projecting point cloud, utilizes distance threshold to scan for local projecting point cloud, calculates not at divisional plane
Point in sheet, to the distance of segmentation dough sheet, when distance is less than threshold value, is designated as required point, by required point according to plan obtained above
Close plane equation to be smoothed to split on dough sheet.
A kind of three-dimensional building object model automatic reconstruction method based on aviation LiDAR data the most according to claim 1, its
It is characterised by: in described 4th step, garret method for resampling is as follows: first some cloud is projected to X/Y plane, and to X/Y plane
Build two dimension grid, then make the following judgment:
A (), LiDAR point in the LiDAR point in four neighborhood grid units and central square net unit belong to same garret,
Now the center of central square net unit is as the internal point of garret, LiDAR in grid unit centered by the height of this internal point
The average height of point；
(b), belong to two when the LiDAR point in some the neighborhood grid unit in four neighborhood grid units of center grid unit
During individual or two or more garret, if most two garret of counting in this neighborhood grid unit is respectively garret j and room
Top layer k, obtains the LiDAR point of garret j in this neighborhood grid unit and the cut-off rule of the LiDAR point of garret k, then calculates
This cut-off rule and central square net unit center line intersecting point coordinate (x in this neighborhood grid unit0, y0), then (x0, y0, zj) it is room
The marginal point of top layer j, zjFor the average height of the LiDAR point of garret j, then (x0, y0, zk) it is the marginal point of garret k, zkFor
The average height of the LiDAR point of garret k；
(c), when in some the neighborhood grid unit in the LiDAR point in the grid unit of center and four neighborhood grid units
When LiDAR point is belonging respectively to two different garrets, if two different garrets are garret j and garret k, calculate central square
Middle point coordinates (the x in the grid sideline between net unit and described neighborhood grid unit1, y1), then (x1, y1, zj) it is garret j's
Marginal point, zjFor the average height of the LiDAR point of garret j, then (x1, y1, zk) it is the marginal point of garret k, zkFor garret k
The average height of LiDAR point.
A kind of three-dimensional building object model automatic reconstruction method based on aviation LiDAR data the most according to claim 6, its
It is characterised by: size is LiDAR point cloud average headway 2 times of described two dimension grid.
A kind of three-dimensional building object model automatic reconstruction method based on aviation LiDAR data the most according to claim 6, its
It is characterised by: in described 4th step, obtains the LiDAR point of garret j in neighborhood grid unit and the LiDAR point of garret k
The method of cut-off rule is: the LiDAR point of garret j in neighborhood grid unit and the LiDAR point of garret k are considered as two classes
Not, algorithm of support vector machine is used to obtain cut-off rule.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610319784.3A CN106023312B (en) | 2016-05-13 | 2016-05-13 | Three-dimensional building object model automatic reconstruction method based on aviation LiDAR data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610319784.3A CN106023312B (en) | 2016-05-13 | 2016-05-13 | Three-dimensional building object model automatic reconstruction method based on aviation LiDAR data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106023312A true CN106023312A (en) | 2016-10-12 |
| CN106023312B CN106023312B (en) | 2019-02-22 |
Family
ID=57100002
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610319784.3A Expired - Fee Related CN106023312B (en) | 2016-05-13 | 2016-05-13 | Three-dimensional building object model automatic reconstruction method based on aviation LiDAR data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106023312B (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107274422A (en) * | 2017-05-08 | 2017-10-20 | 燕山大学 | A kind of point cloud edge extracting method combined based on normal information and K neighborhood search |
| CN107944383A (en) * | 2017-11-21 | 2018-04-20 | 航天科工智慧产业发展有限公司 | Building roof patch division method based on three-dimensional Voronoi diagram |
| CN108875804A (en) * | 2018-05-31 | 2018-11-23 | 腾讯科技（深圳）有限公司 | A kind of data processing method and relevant apparatus based on laser point cloud data |
| CN108898553A (en) * | 2018-06-28 | 2018-11-27 | 中国地震局地震预测研究所 | Building destruction condition detection method |
| CN108955564A (en) * | 2018-06-20 | 2018-12-07 | 北京云迹科技有限公司 | Laser data method for resampling and system |
| CN108986218A (en) * | 2018-06-06 | 2018-12-11 | 东南大学 | A kind of building point off density cloud fast reconstructing method based on PMVS |
| CN110009745A (en) * | 2019-03-08 | 2019-07-12 | 浙江中海达空间信息技术有限公司 | According to plane primitive and model-driven to the method for data reduction plane |
| CN110249741A (en) * | 2019-06-05 | 2019-09-20 | 中国农业大学 | A kind of seed potato block cutting method based on point cloud model |
| CN110414379A (en) * | 2019-07-10 | 2019-11-05 | 武汉大学 | In conjunction with the building extraction algorithm of elevation map Gabor textural characteristics and LiDAR point cloud feature |
| CN111369680A (en) * | 2020-02-28 | 2020-07-03 | 当家移动绿色互联网技术集团有限公司 | Method and device for generating three-dimensional image of building |
| CN112184900A (en) * | 2019-07-04 | 2021-01-05 | 北京四维图新科技股份有限公司 | Method and device for determining elevation data and storage medium |
| CN112683221A (en) * | 2020-12-21 | 2021-04-20 | 深圳集智数字科技有限公司 | Building detection method and related device |
| CN112700465A (en) * | 2021-01-08 | 2021-04-23 | 上海建工四建集团有限公司 | Actual measurement actual quantity oriented room body point cloud extraction and part segmentation method and device |
| CN113470095A (en) * | 2021-09-03 | 2021-10-01 | 贝壳技术有限公司 | Processing method and device for indoor scene reconstruction model |
| CN114445550A (en) * | 2020-11-06 | 2022-05-06 | 中煤航测遥感集团有限公司 | Three-dimensional reconstruction method and device, electronic equipment and storage medium |
| CN115457205A (en) * | 2022-09-14 | 2022-12-09 | 青岛海信医疗设备股份有限公司 | Method, device, equipment and medium for constructing three-dimensional visceral organ model |
| CN116578876A (en) * | 2023-07-12 | 2023-08-11 | 国网江苏省电力有限公司苏州供电分公司 | Safety improvement method based on resistive attack deep neural network |
| CN117218143A (en) * | 2023-11-07 | 2023-12-12 | 法奥意威(苏州)机器人系统有限公司 | Point cloud segmentation method and device |
| CN117934732A (en) * | 2024-03-22 | 2024-04-26 | 北京华航唯实机器人科技股份有限公司 | Reconstruction method and device of point cloud model and electronic equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101877128A (en) * | 2009-12-23 | 2010-11-03 | 中国科学院自动化研究所 | Method for segmenting different objects in three-dimensional scene |
| CN104484668A (en) * | 2015-01-19 | 2015-04-01 | 武汉大学 | Unmanned aerial vehicle multi-overlapped-remote-sensing-image method for extracting building contour line |
| CN104809759A (en) * | 2015-04-03 | 2015-07-29 | 哈尔滨工业大学深圳研究生院 | Large-area unstructured three-dimensional scene modeling method based on small unmanned helicopter |
-
2016
- 2016-05-13 CN CN201610319784.3A patent/CN106023312B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101877128A (en) * | 2009-12-23 | 2010-11-03 | 中国科学院自动化研究所 | Method for segmenting different objects in three-dimensional scene |
| CN104484668A (en) * | 2015-01-19 | 2015-04-01 | 武汉大学 | Unmanned aerial vehicle multi-overlapped-remote-sensing-image method for extracting building contour line |
| CN104809759A (en) * | 2015-04-03 | 2015-07-29 | 哈尔滨工业大学深圳研究生院 | Large-area unstructured three-dimensional scene modeling method based on small unmanned helicopter |
Non-Patent Citations (4)
| Title |
|---|
| LIANG CHENG ET AL: ""3D Building Model Reconstruction from Multi-view Aerial Imagery and Lidar Data"", 《PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING》 * |
| LIANG CHENG ET AL: ""Dynamic Triangle - based Method for 3D Building Rooftop Reconstruction from LiDAR Data"", 《IEEE》 * |
| 王方建 等: ""地面激光扫描数据在建筑物重建中的研究进展"", 《遥感信息》 * |
| 程亮 等: ""LiDAR数据与正射影像结合的三维屋顶模型重建方法"", 《武汉大学学报·信息科技版》 * |
Cited By (31)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107274422A (en) * | 2017-05-08 | 2017-10-20 | 燕山大学 | A kind of point cloud edge extracting method combined based on normal information and K neighborhood search |
| CN107944383A (en) * | 2017-11-21 | 2018-04-20 | 航天科工智慧产业发展有限公司 | Building roof patch division method based on three-dimensional Voronoi diagram |
| CN108875804A (en) * | 2018-05-31 | 2018-11-23 | 腾讯科技（深圳）有限公司 | A kind of data processing method and relevant apparatus based on laser point cloud data |
| CN108875804B (en) * | 2018-05-31 | 2019-12-20 | 腾讯科技（深圳）有限公司 | Data processing method based on laser point cloud data and related device |
| CN108986218A (en) * | 2018-06-06 | 2018-12-11 | 东南大学 | A kind of building point off density cloud fast reconstructing method based on PMVS |
| CN108955564A (en) * | 2018-06-20 | 2018-12-07 | 北京云迹科技有限公司 | Laser data method for resampling and system |
| CN108955564B (en) * | 2018-06-20 | 2021-05-07 | 北京云迹科技有限公司 | Laser data resampling method and system |
| CN108898553B (en) * | 2018-06-28 | 2021-01-22 | 中国地震局地震预测研究所 | Method for detecting destruction state of building |
| CN108898553A (en) * | 2018-06-28 | 2018-11-27 | 中国地震局地震预测研究所 | Building destruction condition detection method |
| CN110009745A (en) * | 2019-03-08 | 2019-07-12 | 浙江中海达空间信息技术有限公司 | According to plane primitive and model-driven to the method for data reduction plane |
| CN110009745B (en) * | 2019-03-08 | 2023-01-06 | 浙江中海达空间信息技术有限公司 | Method for extracting plane from point cloud according to plane element and model drive |
| CN110249741A (en) * | 2019-06-05 | 2019-09-20 | 中国农业大学 | A kind of seed potato block cutting method based on point cloud model |
| CN112184900A (en) * | 2019-07-04 | 2021-01-05 | 北京四维图新科技股份有限公司 | Method and device for determining elevation data and storage medium |
| CN112184900B (en) * | 2019-07-04 | 2024-03-19 | 北京四维图新科技股份有限公司 | Method, device and storage medium for determining elevation data |
| CN110414379A (en) * | 2019-07-10 | 2019-11-05 | 武汉大学 | In conjunction with the building extraction algorithm of elevation map Gabor textural characteristics and LiDAR point cloud feature |
| CN111369680B (en) * | 2020-02-28 | 2023-12-15 | 北京五一视界数字孪生科技股份有限公司 | Method and device for generating three-dimensional image of building |
| CN111369680A (en) * | 2020-02-28 | 2020-07-03 | 当家移动绿色互联网技术集团有限公司 | Method and device for generating three-dimensional image of building |
| CN114445550A (en) * | 2020-11-06 | 2022-05-06 | 中煤航测遥感集团有限公司 | Three-dimensional reconstruction method and device, electronic equipment and storage medium |
| CN112683221A (en) * | 2020-12-21 | 2021-04-20 | 深圳集智数字科技有限公司 | Building detection method and related device |
| CN112683221B (en) * | 2020-12-21 | 2022-05-17 | 深圳集智数字科技有限公司 | Building detection method and related device |
| CN112700465A (en) * | 2021-01-08 | 2021-04-23 | 上海建工四建集团有限公司 | Actual measurement actual quantity oriented room body point cloud extraction and part segmentation method and device |
| CN112700465B (en) * | 2021-01-08 | 2024-02-09 | 上海建工四建集团有限公司 | Actually measured real quantity-oriented room main body point cloud extraction and part segmentation method and equipment |
| CN113470095A (en) * | 2021-09-03 | 2021-10-01 | 贝壳技术有限公司 | Processing method and device for indoor scene reconstruction model |
| CN113470095B (en) * | 2021-09-03 | 2021-11-16 | 贝壳技术有限公司 | Processing method and device for indoor scene reconstruction model |
| CN115457205A (en) * | 2022-09-14 | 2022-12-09 | 青岛海信医疗设备股份有限公司 | Method, device, equipment and medium for constructing three-dimensional visceral organ model |
| CN115457205B (en) * | 2022-09-14 | 2024-08-20 | 青岛海信医疗设备股份有限公司 | Method, device, equipment and medium for constructing three-dimensional viscera model |
| CN116578876A (en) * | 2023-07-12 | 2023-08-11 | 国网江苏省电力有限公司苏州供电分公司 | Safety improvement method based on resistive attack deep neural network |
| CN116578876B (en) * | 2023-07-12 | 2024-02-13 | 国网江苏省电力有限公司苏州供电分公司 | Safety improvement method based on resistive attack deep neural network |
| CN117218143A (en) * | 2023-11-07 | 2023-12-12 | 法奥意威(苏州)机器人系统有限公司 | Point cloud segmentation method and device |
| CN117218143B (en) * | 2023-11-07 | 2024-01-23 | 法奥意威(苏州)机器人系统有限公司 | Point cloud segmentation method and device |
| CN117934732A (en) * | 2024-03-22 | 2024-04-26 | 北京华航唯实机器人科技股份有限公司 | Reconstruction method and device of point cloud model and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106023312B (en) | 2019-02-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106023312B (en) | Three-dimensional building object model automatic reconstruction method based on aviation LiDAR data | |
| CN110570428B (en) | Method and system for dividing building roof sheet from large-scale image dense matching point cloud | |
| CN107146280B (en) | Point cloud building reconstruction method based on segmentation | |
| CN104036544B (en) | A kind of building roof method for reconstructing based on on-board LiDAR data | |
| CN104392486B (en) | One kind point cloud scene reconstruction method | |
| Zolanvari et al. | Three-dimensional building façade segmentation and opening area detection from point clouds | |
| CN106127857B (en) | The on-board LiDAR data modeling method of integrated data driving and model-driven | |
| CN106780751A (en) | Three-dimensional point cloud method for reconstructing based on improved shielding Poisson algorithm | |
| CN115564926B (en) | Three-dimensional patch model construction method based on image building structure learning | |
| CN106570468A (en) | Method for reconstructing LiDAR original point cloud building contour | |
| CN116310192A (en) | Urban building three-dimensional model monomer reconstruction method based on point cloud | |
| Truong-Hong et al. | Octree-based, automatic building facade generation from LiDAR data | |
| CN106126816B (en) | Repeat the extensive ALS building point cloud modeling method under building automatic sensing | |
| CN103413297A (en) | Cutting method based on integrated three-dimensional GIS model | |
| CN107657659A (en) | The Manhattan construction method for automatic modeling of scanning three-dimensional point cloud is fitted based on cuboid | |
| CN103927788A (en) | Building ground feature DEM manufacturing method based on city vertical planning | |
| CN107644452A (en) | Airborne LiDAR point cloud roof dough sheet dividing method and system | |
| CN107944383A (en) | Building roof patch division method based on three-dimensional Voronoi diagram | |
| CN108919295A (en) | Airborne LiDAR point cloud road information extracting method and device | |
| CN104751479A (en) | Building extraction method and device based on TIN data | |
| CN115222883A (en) | Electric power tower reconstruction method based on foundation LiDAR point cloud | |
| CN107818338A (en) | A kind of method and system of building group pattern-recognition towards Map Generalization | |
| CN114898118A (en) | Automatic statistical method and system for power transmission line house removal amount based on multi-source point cloud | |
| CN116012613B (en) | Method and system for measuring and calculating earthwork variation of strip mine based on laser point cloud | |
| Vanderhaegen et al. | Developing urban metrics to describe the morphology of urban areas at block level |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190222 Termination date: 20200513 |