CN105787921B - The method for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data - Google Patents
The method for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data Download PDFInfo
- Publication number
- CN105787921B CN105787921B CN201510511068.0A CN201510511068A CN105787921B CN 105787921 B CN105787921 B CN 105787921B CN 201510511068 A CN201510511068 A CN 201510511068A CN 105787921 B CN105787921 B CN 105787921B
- Authority
- CN
- China
- Prior art keywords
- structural unit
- viaduct
- volume elements
- bridge floor
- center line
- 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.)
- Expired - Fee Related
Links
Classifications
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10044—Radar image
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
Abstract
The present invention relates to a kind of methods for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data.This method step are as follows: extract viaduct point cloud from on-board LiDAR data using inverse iteration mathematical morphology filter and area information;It is split according to bridge floor connectivity pair viaduct point cloud;It is extracted using the segmentation of center line vertical line scan method without bifurcated or crosses structure and connection bridge floor that width is consistent, obtained " structural unit ";Make round buffer area at the endpoint of structural unit center line, passes through the barrier structure for judging to put cloud level journey infomation detection viaduct in buffer area;By structural unit matching, the fitting of two-dimensional center line and three-dimension curved surface fit procedure, complete bridge floor three-dimensional center line is obtained;Viaduct threedimensional model is rebuild in conjunction with bridge floor width information.It was verified that the present invention can effectively rebuild Large and Complicated Interchange bridge threedimensional model, solve the occlusion issue of viaduct point cloud, the threedimensional model of reconstruction accuracy with higher and percentage of head rice.
Description
Technical field
The present invention relates to a kind of reconstruction viaduct method for reconstructing three-dimensional model, utilize airborne LiDAR more particularly to a kind of
The method of data reconstruction Large and Complicated Interchange bridge threedimensional model.
Background technique
Viaduct makes urban transportation move towards three-dimensional from plane, avoids the current conflict of intersection, can both guarantee vehicle
Driving safety can improve the traffic capacity of road again, and be also equipped with stronger traffic management function.Viaduct is as new
The road circuit node of type three-dimensional, becoming can not in the building and city landscape design and visual features of modern city landmark-type
Or scarce a part.Complete spatial information can be presented in viaduct threedimensional model, declare in communication navigation, landscape design, city
The fields such as biography have great significance.
Cloverleaf Junction is more complicated multilayer atural object, and not with design requirement and locating terrain and its features environment
Together, morphosis differs greatly, and has complicated three-dimensional structure and spatial topotaxy.The three-dimensional reconstruction of viaduct is more at present
It is time-consuming and laborious using manual modeling pattern, inefficiency.Therefore cloverleaf Junction is automatically identified and is modeled as in order to
One new problem.
In the researchs such as the detection and extraction of bridge, remote sensing image is common data source, segmentation (Chaudhuri.D,
2012), mathematical morphology (Valero.S, 2010), feature extraction (Shao.Y, 2011), region growing (Amo.M, 2006) etc.
As common technological means, good effect is achieved.Wang and Zheng in 1998 is integrated with CFAR and detects, Image Segmentation Methods Based on Features,
The methods of morphologic filtering and Hough transform extract road and bridge area in High-resolution SAR Images;2008
Soergel et al. extracts water surface bridge using high-resolution InSAR data and orthography and carries out visual analyzing;2013
He et al. fusion optical image and Synthetic Aperture Radar images simultaneously propose a kind of method for extracting overpass network.It is simple to utilize
Image data, current research object are mostly water surface bridge or road, and the viaduct structure being related to is also relatively simple, it is overall and
Speech, solely using remote sensing image attempt realize Large and Complicated Interchange bridge automatically extract and three-dimensional modeling, still have very big
Technical difficulty.
Laser radar is the novel measuring technique of one kind fast-developing in recent ten years, is widely used in surface exploration
(Vosselman.G, 2005), feature detect (Tong.L, 2013), Objects extraction (Boyko.A, 2011), reconstructing three-dimensional model
(Cheng.L, 2012) etc., shows huge application prospect.2009, Oude Elberink et al. merged laser
Radar data and two dimensional terrain data realize the automatic three-dimensional rebuilding method of highway intersection;Chen Zhuo in 2012 et al. is utilized
LiDAR data rebuilds flyover model, is modeled using the staff cultivation triangulation network, and block knot using triangulation network infomation detection and recovery
Structure, but effect is undesirable in the case where serious shielding;2012, Li Hui et al. propose a kind of integrated LIDAR data and
The viaduct automatic testing method of remote sensing image.It in general, is mostly to build using the object that LiDAR data carries out three-dimensional reconstruction
Object or road (Denis.E, 2010;Cheng.L,2011;Sun.S, 2013), shorter mention large-sized multiple layer complexity viaduct,
And correlative study is insufficient to the barrier structure detection of viaduct and its repairing research.
Summary of the invention
The technical problem to be solved by the present invention is overcoming prior art disadvantage, a kind of utilization on-board LiDAR data weight is proposed
The method for building Large and Complicated Interchange bridge threedimensional model is able to solve viaduct occlusion issue, accurately and efficiently rebuilds large-scale multiple
Miscellaneous viaduct threedimensional model.
A kind of method for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data of the present invention 1., including with
Lower step:
Step 1, viaduct data reduction --- ground in initial data is realized using inverse iteration mathematical morphology filter
The separation of point and non-ground points removes the higher tree point cloud of roughness, followed by a cloud level journey roughness according to default
The biggish viaduct region of area threshold Retention area, complete viaduct data reduction;
Step 2, the segmentation of viaduct point cloud connectivity --- three-dimensional grid is constructed according to the viaduct of extraction point cloud, to obtain
The three-dimensional volume elements obtained accommodates viaduct point cloud, is same category by interconnected cell combination, completes to viaduct point cloud
Primary segmentation;
Step 3, the segmentation of viaduct structural unit --- viaduct point cloud is projected to X/Y plane, constructs two dimension in X/Y plane
Regular grid, and area pattern is converted by subpoint, vectorized process then is carried out to area pattern and obtains bridge floor center line,
Using the endpoint of center line as starting point, continuously make the vertical line of center line with constant spacing, and be scanned to bridge floor along vertical line direction,
Count effective two-dimentional grid quantity containing point cloud data that vertical line passes through, the vertical line to mutate with effective two-dimentional grid quantity
Connection bridge floor is split, continuous and equivalent width several structural units are obtained;
The detection and reparation of step 4, barrier structure --- buffer area is made as the center of circle using the centerline end point of structural unit, is delayed
The dispersed elevation difference for rushing structural unit point cloud in area is greater than height difference threshold value between the structural unit set, then the decision structure unit
For the structural unit blocked, the structural unit blocked includes fault barrier structural unit and hangs barrier structure unit, described disconnected
It splits barrier structure unit and refers to and be fractured into two structural units after continuous deck is blocked, the suspension barrier structure unit is then attached
Other fault barrier structural units are closely not present;When reparation, first fault barrier structural unit is matched, to fracture after matching
The center line of barrier structure unit is repaired to obtain complete three-dimensional center line, utilizes center line to suspension barrier structure unit
Extension is repaired, and complete three-dimensional center line is obtained;
Step 5, width and viaduct Complete three-dimensional center line using structural unit, reconstruction obtain viaduct three-dimensional mould
Type.
The present invention also further characterized in that
1, in step 2, connectivity judgment method is as follows between volume elements: the height value of all LiDAR points in record volume elements, and will
Height value of the average value of point cloud level journey as the volume elements, if center volume elements and the absolute value of the elevation difference of its neighborhood volume elements are small
The height difference threshold value Th between preset volume elements1, then determine that the neighborhood volume elements and center volume elements are interconnected, otherwise determine the neighborhood body
Member is not connected to mutually with center volume elements.
2, in step 2, the height difference threshold value Th1Pass through formula Th1=2 × Gridsize × i, which is calculated, to be obtained, wherein
Gridsize is voxel size, and i is the bridge floor maximum longitudinal grade gradient.
3, in step 2, preset noise spot amount threshold Th is utilized2, reject the affiliated facility of non-bridge floor in segmentation result
Point cloud, the noise spot amount threshold Th2Pass through formula Th2=(W × W)/S, which is calculated, to be obtained, and wherein W is bridge floor minimum widith, S
For the equalization point spacing of original LIDAR data.
4, the structural unit width multiplies the size in two-dimentional grid by effective two-dimentional grid quantity and is calculated, two dimension
LiDAR point data are then effective two-dimensional grid if it exists in grid, are otherwise abortive haul lattice.
5, in step 4, the value of the radius r of buffer area is viaduct bridge floor bicycle road width.
6, in step 4, fault barrier structural unit matching process is as follows:
A), using the centerline end point of fault barrier structural unit as the center of circle, the distance threshold of 2-4 times of bridge deck width of setting is searched
The centerline end point of other fault barrier structural units of rope, search centerline end point fault barrier structural unit be used as to
Distribution structure unit;
B), the above bridge floor is that structural unit to be matched is grouped by boundary;
C), every group of center line is chosen respectively forms line pair;
If d), line is less than preset height difference threshold value Th to the elevation difference of endpoint3, then the line is to gone out for preliminary screening
With scheme;
E), the line in the matching scheme gone out to preliminary screening is fitted the two-dimensional curve carried out in X/Y plane, is with fitting
Number R2Fitting result is evaluated, to determine Optimum Matching scheme.
7, fitting coefficient R is utilized2The method evaluated the fitting of structural unit matched center line is as follows: statistical fit
Total sum of squares SST, the residuals squares SSR and and fitting coefficient R of curve2, the actual value y of match point yiAnd average valueDifference
Square the sum of be total sum of squares, the estimated value y ' of the y value of every bitiWith actual value yiDifference the sum of square be residual sum of squares (RSS),
Fitting coefficient R2Between 0-1, fitting coefficient R2Closer to 1, fitting effect is better, total sum of squares SST, residuals squares SSR,
Fitting coefficient R2Calculation formula it is as follows:
R2=1-SSR/SST.
8, height difference threshold value Th3Pass through formula Th3=D × i, which is calculated, to be obtained, and wherein D is adjacent fault barrier structural unit center
Maximum distance between line endpoints, i are the bridge floor maximum longitudinal grade gradient.
9, as follows using the method that center line extends reparation suspension barrier structure: in suspension barrier structure in step 4
Heart line resampling is fitted the two-dimensional curve obtained in X/Y plane according to sampled point, and quasi- using the point cloud near the bridge floor that is blocked
The three-dimension curved surface for obtaining the region that is blocked is closed, fitting two-dimensional curve closes on bridge floor center line with occlusion area and projects on X/Y plane
Intersection point be (X1, Y1), then this is closed on bridge floor center line there are point (X1, Y1, Z1), on fitting surface there are point (X1, Y1,
Z2), if | Z1-Z2 | be less than height difference threshold value Th3, then the bridge floor that determines to be blocked closed on bridge floor with this and is connected to, intended with two-dimensional curve
It closes the projection line reparation on curved surface to be blocked the center line of part, obtains complete three-dimensional center line.
Beneficial effects of the present invention are as follows:
Compared with prior art, the present invention realizes a kind of three-dimensional using on-board LiDAR data reconstruction Large and Complicated Interchange bridge
Model method.This method is by viaduct data reduction, the point cloud segmentation based on connectivity, and the segmentation of viaduct structural unit hides
The detection and reparation of structure are kept off, the processes such as reconstructing three-dimensional model realize automatically extracting and threedimensional model weight for viaduct point cloud
It builds.
(1) present invention is by point cloud hierarchical segmentation technique, including the step such as cloud connectivity segmentation and structural unit segmentation
Suddenly, viaduct structural unit point cloud is effectively obtained.Based on " structural unit " concept, complicated viaduct is integrally decomposed into structure
Single module, greatly reduces the technical difficulty of three-dimensional reconstruction, effectively improves the reconstruction efficiency and precision of flyover model.
(2) viaduct barrier structure automatic detection and rehabilitation technology proposed by the present invention efficiently solves bridge floor and blocks and asks
Topic, improves the quality of reconstruction model.Mutilevel overpass will appear the phenomenon that lower layer's bridge floor is blocked by upper deck of bridge, existing research
Shorter mention problems, part research solves occlusion issue using model triangulation network information or surface fitting mode, but imitates
Fruit is undesirable.The present invention detects barrier structure by putting the elevation information of cloud in buffer area at centerline end point, and passes through structure
Units match, the reparation of two-dimensional center line and three-dimension curved surface elevation interpolation and etc. barrier structure is repaired, improve reconstruction
The quality of model.
To sum up, the present invention proposes a kind of using on-board LiDAR data reconstruction Large and Complicated Interchange bridge model method, should
Method by inverse iteration mathematical morphology filter, point cloud Hierarchical Segmentation and barrier structure be automatically repaired three processes realize it is vertical
Hand over accurate extraction and the three-dimensional reconstruction of bridge point cloud.Test proves that this method can adapt to Large and Complicated Interchange bridge, has preferable
Stability, the accuracy and percentage of head rice of reconstruction model are higher, ideal to the repairing effect blocked.
Detailed description of the invention
The present invention will be further described below with reference to the drawings.
Fig. 1 is viaduct of embodiment of the present invention reconstructing three-dimensional model flow chart.
Fig. 2-a is region of embodiment of the present invention aviation LiDAR data schematic diagram.
Fig. 2-b is region of embodiment of the present invention high score aviation image schematic diagram data.
Fig. 3 is the viaduct point cloud schematic diagram extracted from Fig. 2-a.
Fig. 4-a is viaduct point cloud connectivity segmentation result schematic diagram.
Fig. 4-b is the bridge floor center line schematic diagram obtained from Fig. 4-a.
Fig. 4-c is viaduct point cloud structure unit segmentation result schematic diagram.
Fig. 5 is three-dimensional regular grid system and single volume elements schematic diagram.
Fig. 6 is that structural unit divides schematic illustration.
Fig. 7 is barrier structure detection schematic diagram.
Fig. 8-a is the matched fault structure unit spot cloud schematic diagram of the present embodiment structural unit.
Fig. 8-b is the structural unit center line and endpoint schematic diagram obtained from Fig. 7.
Fig. 8-c is the structural unit centerline points schematic diagram obtained from Fig. 7.
Fig. 8-d is 1 schematic diagram of matching scheme that structural unit is broken in Fig. 7.
Fig. 8-e is 2 schematic diagram of matching scheme that structural unit is broken in Fig. 7.
Fig. 9-a is matched fault structure unit spot cloud schematic diagram.
Fig. 9-b is two-dimensional curve fitted area schematic diagram.
Fig. 9-c is that region of fracture two dimension repairs curve synoptic diagram.
Fig. 9-d is full two-dimensional center line schematic diagram.
Fig. 9-e is center line resampling point schematic diagram.
Fig. 9-f is three-dimension curved surface fitted area schematic diagram.
Fig. 9-g is surface fitting result schematic diagram.
Fig. 9-h is Complete three-dimensional center line schematic diagram.
Fig. 9-i is three-dimension curved surface fitted area schematic diagram.
Figure 10-a is the present embodiment overhang region image schematic diagram.
Figure 10-b is the present embodiment overhang region point cloud schematic diagram.
Figure 10-c is that the present embodiment overhang region extends node height difference schematic diagram.
Figure 11-a the present embodiment flyover model and point cloud TIN overlap schematic diagram.
Figure 11-b the present embodiment flyover model and original point cloud overlap schematic diagram.
Specific embodiment
Region of the embodiment of the present invention is Jiangsu Province's viaduct.Airborne LiDAR initial data (Fig. 2-a), it is big as rebuilding
The embodiment data of type complexity viaduct threedimensional model, equalization point spacing are 1m, height accuracy 0.20m, and plane precision is
0.50m.The size in embodiment region is about 520m × 450m, a total of 820,000 scanning elements of initial data.High resolution image
Data (Fig. 2-b), as the reference data of model evaluation, image resolution 0.10m.
The embodiment of the present invention rebuilds Large and Complicated Interchange bridge model method (overall flow using on-board LiDAR data
See Fig. 1), comprising the following steps:
Step 1 extracts viaduct point cloud.Firstly the need of the extraction viaduct point cloud from raw LiDAR data.Initial data
In mainly contain viaduct, four class point cloud of building, vegetation and ground.The present invention is based in patent CN201110432421.8
Reversed mathematical morphology iterative filtering method extracts viaduct point cloud, and extraction process is briefly described:
Resampling is carried out to original LiDAR point cloud, reversed mathematical morphology iteration is carried out to the equidistant point after resampling
Filtering, is gradually reduced window size with lesser step-length, uses morphology "ON" to operate each window, while constantly to phase
It is poor that two adjacent window filtering results are made, and the separation of ground point and non-ground points is realized to difference given threshold size.By anti-
It is filtered to iterated morphism, ground and independent tree point cloud can be effectively removed.Followed by the roughness (elevation of cloud level journey
Variance), given threshold removes the higher tree point cloud of roughness.The occupied area of viaduct is greater than single building surface
Product sets area threshold herein to remove building point cloud.Viaduct data reduction is completed, as a result sees Fig. 3.In addition to this, also
Extraction viaduct point cloud with other methods can be adopted, is easy to accomplish for a person skilled in the art.
Step 2, the viaduct point cloud segmentation based on connectivity.The purpose of this step is according to bridge floor connectivity by viaduct point
Cloud is divided into different classes of, including building three-dimensional regular grid system, connectivity judgement between volume elements, specific as follows:
It constructs three-dimensional regular grid first 2a) to accommodate the viaduct point cloud of discrete distribution, converts the segmentation of a cloud to
The classification of volume elements.Fig. 5 is the schematic diagram of three-dimensional grid system and volume elements, in three-dimensional grid coordinate system, X-direction record
It is the row number of volume elements, what Y direction recorded is the line number of volume elements, and what Z-direction recorded is the level number of volume elements.Volume elements is a mark
Quasi- cube, records the serial number of LiDAR point contained by its inside, and no point data is then hollow body member.To each LiDAR point, record
Row, column and level number of its affiliated volume elements in three-dimensional grid coordinate system, and two-way index is constructed with this.Wherein calculate point cloud
The formula of affiliated volume elements row, column and level number is as follows:
I=int ((y-ymin)/Gridsize)
J=int ((x-xmin)/Gridsize) (1)
K=int ((z-zmin)/Gridsize)
(i, j, k) respectively represents the row, column of volume elements, level number in formula (1), contained LiDAR point in (x, y, z) representative volume element
Space coordinate, (xmin,ymin,zmin) indicate viaduct point cloud min coordinates value, Gridsize indicate volume elements size.
2b) bridge floor connection and elevation change gentle, and the present invention is split using connectivity pair point cloud between volume elements.Connectivity
Judgment criteria is as follows: using the elevation mean value of all the points in volume elements as the height value of the volume elements.Calculating center volume elements and its neighborhood
The elevation difference of volume elements, if high absolute value of the difference is less than height difference threshold value Th between volume elements1, then it is assumed that neighborhood volume elements and center volume elements
It is interconnected, the point cloud genera in two volume elements is in same bridge floor.Otherwise, two volume elements are not connected to.
Height difference threshold value Th between volume elements1Pass through formula Th1=2 × Gridsize × i, which is calculated, to be obtained, and wherein Gridsize is body
First size, i are the bridge floor maximum longitudinal grade gradient, and the bridge top rake maximum value in Chinese city region is 9% (37-2012 city CJJ
City's road engineering design specification), the present embodiment is using 10% as longitudinal slope gradient maximum value.
Detailed process is as follows for the classification of volume elements connectivity:
1, the volume elements at origin is chosen as starting volume elements, and the initial attribute value of all volume elements is set as -1.
2, judge whether volume elements is unlabelled volume elements (whether volume elements attribute value is -1), (belong to if volume elements has been labeled
Value is non-1) then detects next volume elements for property.Next step operation is carried out if volume elements not labeled (body attribute value is -1).
3, judge that whether containing LiDAR point data in volume elements is set as -2 for the attribute value of the volume elements, together if no data
When detect next volume elements, and return step 2.If containing LiDAR point data in volume elements, which is assigned to N (N
Initial value be set as 0), while by the volume elements be added growth list in, carry out next step operation.
4, the first volume elements of list is grown as seed voxel.Seed voxel is located at 3 × 3 × 3 net centers of a lattice, by row,
Column, layer sequence successively traverse 26 neighborhood volume elements, and judge whether neighborhood volume elements meets connectivity, if meeting mark
Neighborhood volume elements is then added growth list and volume elements attribute value is labeled as N by standard.If not meeting connectivity, neighborhood body
The attribute value of member is constant and neighborhood volume elements is added without growth list.
5, the first volume elements in growth list is removed, volume elements in list is grown and moves forward one, original second volume elements conduct
The first volume elements.Judge to grow whether list is empty, if still making the first volume elements containing a little member, return step 4 in growth list
For new seed voxel continued growth.If list is empty for growth, N value adds 1 (N=N+1), indicates that a kind of point cloud classifications are completed,
Carry out the operation of next step.
6, judge that whether all volume elements are all labeled in three-dimensional grid system, if still having volume elements not to be labeled, choose not
Mark volume elements as seed voxel, return step 2 carries out the growth of a new round.If all labeled (attribute value of all volume elements
It is non-1), then entire growth course terminates.
By operation 1 to 6, all volume elements are labeled and classify, and are classified according to the attribute value of volume elements to a cloud,
The connectivity of point cloud, which is divided, to be completed, and connectivity segmentation result is shown in Fig. 4-a.
2c) after connectivity is divided, point cloud segmentation is multiple classifications.Bridge floor two sides there are affiliated facilities such as street lamps, because
This contains noise spot, needs to reject.Street lamp distribution is more discrete and is not connected to mutually with bridge floor, therefore can set amount threshold progress
Judgement.If the quantity of certain one kind point cloud is less than threshold value, then it is assumed that be noise spot.According to this to the point cloud cancelling noise of all categories
Point.During noise points deleting, noise spot quantity is much smaller than bridge floor point cloud.Noise spot amount threshold Th2Pass through formula Th2=
(W × W)/S, which is calculated, to be obtained, and wherein W is bridge floor minimum widith, and S is the equalization point spacing of raw LiDAR data.
Step 3, the segmentation of viaduct structural unit.Multiple structural units may be contained by being connected to bridge floor, subsequent in order to be conducive to
Modeling, needs further to divide.Center line preferably maintains the integrality and continuity of bridge deck structure, therefore set forth herein
The mode of central axis scanning determines structural unit.Bridge floor point cloud is converted to planar and wanted by the two-dimentional regular grid of building first
Element, then by binaryzation, refinement tracks isovectorization operation, obtains bridge floor center line.The endpoint of Selection Center line is used as
Initial point is scanned bridge floor point cloud iteration along vertical line direction, counts vertical line direction effective grid quantity.In the process of scanning
In, in the case that bridge deck width is constant, effective grid number is relatively stable.When there is bifurcated or cross structure when, grid number
Amount is mutated, in this, as detection bridge floor bifurcated or the foundation for the structure that crosses.According to scanning result, segmentation connection bridge floor is obtained
Structural unit, the present embodiment structural unit segmentation result are shown in Fig. 4-b.
Specific cutting procedure exists simultaneously two kinds of segmentation feelings as shown in fig. 6, bridge floor center line contains 3 endpoints (A, B, C)
Condition: being a plurality of ring road one is a bifurcation structure i.e. bridge floor bifurcated;Another kind is that cross that the i.e. a plurality of ring road of structure crosses be one
Bridge floor.
Selected element A makees center line vertical line A as starting point1.Along A1It is scanned to two side point cloud of center line, when detecting sky
Grid is the internal grid without point cloud data, then A1Direction reaches bridge floor edge and stops scanning.Record A1Effective net on direction
Lattice quantity NA1.Selection next node spaced apart, and make vertical line A2.Node is successively chosen, and records every vertical line direction
On effective grid quantity.AkIt indicates kth time scan line, is crossed to form two nodes with middle line.At this point, there is bifurcated knot in bridge floor
Structure stops scanning.
Selected element B is as starting point.BkIndicate kth time scan line, NBkIndicate vertical line BkEffective grid quantity on direction.
During Scanning Detction, if NBk+1Greater than 1.5*NBkAnd NBk+2Greater than 1.5*NBk, then it is assumed that there is the structure that crosses in bridge floor, and stops
Only scan.If it is similar to point B to scan cutting procedure using endpoint C as starting point, it is not repeated to illustrate.So far, center line
Three endpoints scan completion, obtain 3 structural units after connection bridge floor segmentation.
Step 4, viaduct occlusion detection and recovery.Since airborne LiDAR overlooks the characteristic of acquisition data, lower layer will cause
The phenomenon that bridge floor is blocked by upper deck of bridge needs to detect barrier structure and repaired, the specific steps are as follows:
4a) the detection of barrier structure.As shown in fig. 7, make the round buffer area of radius r using centerline end point as the center of circle, half
Diameter r size is the width (r takes 3.5m in this example) in single lane.Point cloud in buffer area containing structural unit B, statistical framework
The point cloud level journey mean value h of unit A and BA、hB, and calculating difference Δ h, formula are as follows:
Δ h=hB-hA (2)
It is greater than between structural unit if height difference threshold value (this example takes 3m) if Δ h and thinks that structural unit A is hidden by structural unit B
Gear.Block to form two kinds of structures: one is fault structure, bridge floor forms structural unit C and D after being blocked by structural unit A, this
Structure can be repaired by the matched mode of structural unit;Another kind is suspended structure, and bridge floor blocks shape by structural unit B
At structural unit A, nearby there is no other structural units, this structure is known as suspended structure.
4b) the matching of fault structure unit.Structural unit A1,A2,B1,B2(Fig. 8-a), center line LA1,LA2,LB1,LB2
(Fig. 8-b), endpoint EA1,EA2,EB1,EB2(Fig. 8-b), centerline points PA1,PA2,PB1,PB2(Fig. 8-c).Fault structure unit
It is as follows with detailed step:
1, structural unit to be matched is searched for.Using the endpoint of center line as the center of circle, setting reasonable distance threshold is about 2-4 times
Bridge deck width, other endpoints in search range.
2, structural unit grouping to be matched.Structural unit to be matched is divided into two groups, LA1And LA2One group;LB1And LB2One
Group is located at upper deck of bridge two sides.Assuming that being respectively LA={ LA after center line groupingi, i=1,2 ..., n } and LB={ LBj,
J=1,2 ..., m } (m >=n).
3, matching scheme is formed.It chooses a center line respectively from LA and LB, forms a line pair.When LA center line is complete
Portion's selection finishes, and forms Amn kind matching scheme.The present embodiment forms two kinds of matching schemes: scheme 1:A1-B1,A2-B2;Scheme 2:
A1-B2,A2-B1。
4, matching scheme screens.Elevation difference of each line to endpoint in calculating matching scheme.Set height difference threshold value Th3,
Th3=D × i, which is calculated, to be obtained, wherein maximum distance of the D between structural unit endpoint, and i is the bridge floor maximum longitudinal grade gradient.This implementation
In example, D=20m is maximum search distance at endpoint, and i=10% is the maximum longitudinal grade gradient, Th3=2m, if difference is greater than height difference
Threshold value Th3Then think that the program is unreasonable.Proper Match scheme after screening is denoted as set C={ Ci, i=1,2 ..., k }
5, Optimum Matching scheme is determined.Two-dimensional curve is fitted to using equation 3 to line matched in scheme.LA and LB is corresponding
Central point be respectively PA={ PAi, i=1,2 ..., n } and PB={ PBi, i=1,2 ..., m }.Using least square method pair
Centerline points carry out curve fitting.Fitting coefficient R is used herein2Fitting result is evaluated.It is wired to calculate each scheme institute
Pair fitting coefficient, the sum of coefficient is recorded as Rsum.Wherein RsumBeing worth maximum matching scheme is Optimum Matching scheme.
Y=a*x2+b*x+c (3)
Total sum of squares SST, the residuals squares SSR and and fitting coefficient R of statistical fit curve2.The actual value of match point y
yiAnd average valueDifference the sum of square be known as total sum of squares, the estimated value y ' of the y value of every bitiWith actual value yiDifference it is flat
The sum of side is known as residual sum of squares (RSS).Fitting coefficient R2Between 0~1, closer to 1, fitting effect is better.SST,SSR,R2Three
Person's formula is as follows:
R2=1-SSR/SST (6)
Endpoint EA in Fig. 8-b1,EA2,EB1And EB2Height value be respectively 15.11,16.25,15.21 and 15.88m.Appoint
The threshold value that the elevation difference of two-end-point of anticipating respectively less than is set, therefore form two kinds of matching schemes.Fig. 8-d and 8-e is respectively scheme 1
With 2 fitting result, RsumValue is respectively 1.999 and 1.925.Obvious scheme 1 is optimal matching scheme.
4c) the reparation of fault structure.Including two-dimensional curve fitting and three-dimension curved surface elevation interpolation.Fig. 9-a is
The structural unit point cloud of fracture obtains two-dimensional center line (Fig. 9-b) by vector quantization step, then chooses the neighbour of the region of fracture
Near point cloud simultaneously obtains repairing curve (Fig. 9-c) using equation (3) fitting, thus obtains complete two-dimensional center line (Fig. 9-d), so
Resampling (Fig. 9-e) equidistantly is carried out to center line afterwards.Formula (7) are used herein and suitable region (Fig. 9-f) is selected to be fitted
Obtain curved surface (Fig. 9-g).The two-dimentional sampled point of the region of fracture carries out elevation interpolation (Fig. 9-h) by the curved surface that fitting obtains, most
Complete three-dimensional center line (Fig. 9-i) is obtained eventually.
Z=a*x2+b*x*y+c*y2+d*x+e*y+f (7)
4d) the reparation of suspended structure.It is likely to occur suspended structure under special circumstances, i.e., there is no matchings near blocking
Structural unit, the present invention extends reasonability judgement using center line and repairs suspended structure.As shown in Figure 10-a and 10-b, quilt
The lower layer's bridge floor 1 and 3 blocked forms bifurcated, destroys the integrality of bifurcation structure, makes bridge floor 1 that can not normally be divided into two sections of knots
Structure unit, therefore form suspended structure.Using equation (3) to 1 matched curve of bridge floor, and extend downward and center line 2 and 3
B, C two o'clock are intersected at respectively.Figure 10-c midpoint A is hanging node, and B, C two o'clock are located at the prolongation of center line 1, B1、C1Point minute
It Wei Yu not center line 2 and 3, B1、C1Subpoint be B point and C point.Surface fitting is carried out to overhang region according to equation (7), and
Point B and B are calculated separately according to fitting result1Elevation difference dB, point C and C1Elevation difference dCIf dB>dCAnd dCLess than height difference
Threshold value then thinks that bridge floor 1 is connect with bridge floor 3, if dC>dBAnd dBThen think that bridge floor 1 is connected to bridge floor 2 less than height difference threshold value.Connection
As a result elevation assignment is carried out to the sampled point of center line 1, obtains complete three-dimensional center line.
Step 5, viaduct reconstructing three-dimensional model.Complete bridge floor center line is obtained using the three-dimensional center line of structural unit.
The width of every bridge deck structure unit is obtained by the center line scan operation of step 3.Utilize bridge floor three-dimensional center line and width
Information, reconstruction obtain viaduct threedimensional model.Figure 11-a is model and the overlay effect diagram for putting cloud TIN base map, the entirety of model
Clear-cut, bridge floor is well arranged.Figure 11-b is the overlay effect diagram of model and original point cloud, and model and LiDAR point cloud are more
It coincide, barrier structure, that is, black surround region of viaduct achieves preferable repairing effect.
The present embodiment is using aviation image viaduct region as true value.The counterweight in terms of the quantity of model and area two
The accuracy and percentage of head rice of established model are calculated:
In formula (8) and (9), TP (True Positives) represents the correct number or correct area of flyover model;
FN (False Negatives) represents the omission number of viaduct reconstruction model or omits area;FP(False Positives)
Represent the error number or mistake area of viaduct reconstruction model.Viaduct based on number rebuilds correct, mistake, omission is sentenced
It is fixed as follows: using the structural unit of flyover model as a judging unit, if the area ratio in correct region and structural unit
Value is greater than 80%, then it is assumed that structural unit is rebuild correct;If missed areas and the area ratio of manual extraction structural unit are big
In 20%, then it is assumed that structural unit is omitted;If zone errors and the area ratio of structural unit are greater than 20%, then it is assumed that structure
Unit reconstruction errors.
1 flyover model accuracy of table and percentage of head rice (quantity and area)
Tab.1Correctness and Completeness of the Reconstructed Model(number
and area)
According to the statistical result of table 1, evaluated in terms of quantity and area two, the viaduct threedimensional model of reconstruction all have compared with
High percentage of head rice and accuracy.
In addition to the implementation, the present invention can also have other embodiments.It is all to use equivalent substitution or equivalent transformation shape
At technical solution, fall within the scope of protection required by the present invention.
Claims (8)
1. a kind of method for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data, comprising the following steps:
Step 1, viaduct data reduction --- using inverse iteration mathematical morphology filter realize initial data in ground point with
The separation of non-ground points removes the higher tree point cloud of roughness, followed by a cloud level journey roughness according to preset face
The product biggish viaduct region of threshold value Retention area, completes viaduct data reduction;
Step 2, the segmentation of viaduct point cloud connectivity --- three-dimensional grid is constructed according to the viaduct of extraction point cloud, with the three of acquisition
It ties up volume elements and accommodates viaduct point cloud, be same category by interconnected cell combination, complete preliminary minute to viaduct point cloud
It cuts;In step 2, connectivity judgment method is as follows between volume elements: the height value of all LiDAR points in record volume elements, and will put the cloud level
Height value of the average value of journey as the volume elements is preset if the absolute value of the elevation difference of center volume elements and its neighborhood volume elements is less than
Volume elements between elevation difference threshold value Th1, then determine that the neighborhood volume elements and center volume elements are interconnected, otherwise determine the neighborhood body
Member is not connected to mutually with center volume elements;
Step 3, the segmentation of viaduct structural unit --- viaduct point cloud is projected to X/Y plane, constructs two dimension rule in X/Y plane
Grid, and area pattern is converted by subpoint, vectorized process then is carried out to area pattern and obtains bridge floor center line, in
The endpoint of heart line is starting point, continuously makees the vertical line of center line with constant spacing, and be scanned to bridge floor along vertical line direction, is counted
Effective two-dimentional grid quantity containing point cloud data that vertical line passes through, the vertical line to be mutated with effective two-dimentional grid quantity is to even
Logical bridge floor is split, and obtains continuous and equivalent width several structural units;
The detection and reparation of step 4, barrier structure --- make buffer area, buffer area by the center of circle of the centerline end point of structural unit
The dispersed elevation difference of interior structural unit point cloud is greater than the threshold value of elevation difference between the structural unit of setting, then determines the structure list
Member is the structural unit blocked, and the structural unit blocked includes fault barrier structural unit and hangs barrier structure unit, described
Fault barrier structural unit, which refers to, is fractured into two structural units after continuous deck is blocked, and the suspension barrier structure unit is then
For other fault barrier structural units are nearby not present;When reparation, first fault barrier structural unit is matched, it is right after matching
The center line of fault barrier structural unit is repaired to obtain complete three-dimensional center line, in suspension barrier structure unit utilization
The extension of heart line is repaired, and complete three-dimensional center line is obtained;
Step 5, width and viaduct Complete three-dimensional center line using structural unit, reconstruction obtain viaduct threedimensional model.
2. the method according to claim 1 for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data,
It is characterized in that: in step 2, the threshold value Th of the elevation difference1Pass through formula Th1=2 × Gridsize × i, which is calculated, to be obtained, wherein
Gridsize is voxel size, and i is the bridge floor maximum longitudinal grade gradient.
3. the method according to claim 1 for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data,
It is characterized in that: in step 2, utilizing preset noise spot amount threshold Th2, reject the affiliated facility point of non-bridge floor in segmentation result
Cloud, the noise spot amount threshold Th2Pass through formula Th2=(W × W)/S, which is calculated, to be obtained, and wherein W is bridge floor minimum widith, and S is
The equalization point spacing of raw LiDAR data.
4. the method according to claim 1 for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data,
Be characterized in that: the width of the structural unit is multiplied by effective two-dimentional grid quantity to be calculated in the size of two-dimentional grid, and two
LiDAR data is then effective two-dimensional grid if it exists in dimension grid, is otherwise abortive haul lattice.
5. the method according to claim 1 for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data,
Be characterized in that: in step 4, the value of the radius r of buffer area is viaduct bridge floor bicycle road width.
6. the method according to claim 1 for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data,
Be characterized in that: in step 4, fault barrier structural unit matching process is as follows:
A), using the centerline end point of fault barrier structural unit as the center of circle, the distance threshold of 2-4 times of bridge deck width of setting searches for it
The centerline end point of his fault barrier structural unit searches the fault barrier structural unit of centerline end point as knot to be matched
Structure unit;
B), the above bridge floor is that structural unit to be matched is grouped by boundary;
C), every group of center line is chosen respectively forms line pair;
If d), line is less than the threshold value Th of preset elevation difference to the elevation difference of endpoint3, then the line to for preliminary screening go out
Matching scheme;
E), the line in the matching scheme gone out to preliminary screening is fitted the two-dimensional curve carried out in X/Y plane, with fitting coefficient R2It is right
Fitting result is evaluated, to determine Optimum Matching scheme.
7. the method according to claim 6 for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data,
It is characterized in that: utilizing fitting coefficient R2The method evaluated the fitting of structural unit matched center line is as follows: statistical fit
Total sum of squares SST, residuals squares SSR and the fitting coefficient R of curve2, the actual value y of match point yiAnd average valueDifference
Square the sum of be total sum of squares, the estimated value y ' of the y value of every bitiWith actual value yiDifference the sum of square be residual sum of squares (RSS),
Fitting coefficient R2Between 0-1, fitting coefficient R2Closer to 1, fitting effect is better, total sum of squares SST, residuals squares SSR,
Fitting coefficient R2Calculation formula it is as follows:
R2=1-SSR/SST.
8. the method according to claim 6 for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data,
It is characterized in that: the threshold value Th of elevation difference3Pass through formula Th3=D × i, which is calculated, to be obtained, and wherein D is adjacent fault barrier structure list
Maximum distance between first centerline end point, i are the bridge floor maximum longitudinal grade gradient.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510511068.0A CN105787921B (en) | 2015-08-19 | 2015-08-19 | The method for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510511068.0A CN105787921B (en) | 2015-08-19 | 2015-08-19 | The method for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105787921A CN105787921A (en) | 2016-07-20 |
CN105787921B true CN105787921B (en) | 2019-01-11 |
Family
ID=56390108
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510511068.0A Expired - Fee Related CN105787921B (en) | 2015-08-19 | 2015-08-19 | The method for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105787921B (en) |
Families Citing this family (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106324581B (en) * | 2016-08-25 | 2018-11-20 | 辽宁工程技术大学 | A kind of airborne LIDAR building analyte detection method based on volume elements |
CN106920278B (en) * | 2017-03-03 | 2020-03-17 | 合肥工业大学 | Flyover three-dimensional modeling method based on Reeb graph |
CN109839119B (en) * | 2017-11-24 | 2021-08-10 | 阿里巴巴(中国)有限公司 | Method and device for acquiring bridge floor area of bridge of cross-road bridge |
CN108919295A (en) * | 2018-05-15 | 2018-11-30 | 国网通用航空有限公司 | Airborne LiDAR point cloud road information extracting method and device |
CN110555905B (en) * | 2018-06-04 | 2023-05-23 | 百度在线网络技术(北京)有限公司 | Method, device, computer equipment and storage medium for generating three-dimensional overpass model |
CN109242786B (en) * | 2018-08-15 | 2020-09-22 | 东华理工大学 | Automatic morphological filtering method suitable for urban area |
CN110108236B (en) * | 2018-10-07 | 2020-05-05 | 东南大学 | Rapid visual measurement system and method for line reconstruction size of high-temperature forging |
CN111369491B (en) * | 2018-12-25 | 2023-06-30 | 宁波舜宇光电信息有限公司 | Image stain detection method, device, system and storage medium |
CN111501496B (en) * | 2019-01-31 | 2022-02-01 | 阿里巴巴集团控股有限公司 | Method and device for determining road elevation |
CN110415331B (en) * | 2019-06-14 | 2022-08-19 | 中国电建集团昆明勘测设计研究院有限公司 | Method for rapidly detecting, judging and repairing contour or hole based on point cloud data |
CN110362872B (en) * | 2019-06-18 | 2023-02-03 | 中建一局集团建设发展有限公司 | Three-direction positioning high-precision control method for hanging grandstand |
CN112099025B (en) * | 2020-08-20 | 2024-04-02 | 杭州飞步科技有限公司 | Method, device, equipment and storage medium for positioning vehicle under bridge crane |
CN112288854B (en) * | 2020-10-29 | 2022-11-18 | 武汉大学 | Construction method of three-dimensional model of overpass |
CN112365500B (en) * | 2021-01-12 | 2022-02-08 | 深兰人工智能芯片研究院(江苏)有限公司 | Contour data completion method and device, electronic equipment and storage medium |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103810489A (en) * | 2013-12-23 | 2014-05-21 | 西安电子科技大学 | LiDAR point cloud data overwater bridge extraction method based on irregular triangulated network |
CN104376595A (en) * | 2014-11-28 | 2015-02-25 | 史文中 | Method for generating three-dimensional roads on basis of airborne LiDAR (light detection and ranging) and GIS (geographic information system) cooperation |
-
2015
- 2015-08-19 CN CN201510511068.0A patent/CN105787921B/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103810489A (en) * | 2013-12-23 | 2014-05-21 | 西安电子科技大学 | LiDAR point cloud data overwater bridge extraction method based on irregular triangulated network |
CN104376595A (en) * | 2014-11-28 | 2015-02-25 | 史文中 | Method for generating three-dimensional roads on basis of airborne LiDAR (light detection and ranging) and GIS (geographic information system) cooperation |
Non-Patent Citations (2)
Title |
---|
Semi-automated extraction and delineation of 3D roads of street scene from mobile laser scanning point clouds;Bisheng Yang等;《ISPRS Journal of Photogrammetry and Remote Sensing》;20130316;第80-93页 |
集成激光雷达数据和遥感影像的立交桥自动检测方法;李卉 等;《测绘学报》;20120630;第41卷(第3期);第428-432页 |
Also Published As
Publication number | Publication date |
---|---|
CN105787921A (en) | 2016-07-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105787921B (en) | The method for rebuilding Large and Complicated Interchange bridge threedimensional model using on-board LiDAR data | |
CN106204705B (en) | A kind of 3D point cloud dividing method based on multi-line laser radar | |
CN110570428B (en) | Method and system for dividing building roof sheet from large-scale image dense matching point cloud | |
CN106815847B (en) | Trees dividing method and single tree extracting method based on laser radar point cloud | |
CN104376595B (en) | A kind of three-dimensional road generation method cooperateed with based on airborne LiDAR and GIS | |
CN105513127B (en) | Shaft regularization three-dimensional modeling method and system based on density peaks cluster | |
CN105894044B (en) | A kind of individual plant tree point cloud extraction method based on Vehicle-borne Laser Scanning data | |
CN103390169B (en) | A kind of city terrain classification method of Vehicle-borne Laser Scanning cloud data | |
Cheng et al. | Three-dimensional reconstruction of large multilayer interchange bridge using airborne LiDAR data | |
CN107679229A (en) | The synthetical collection and analysis method of city three-dimensional building high-precision spatial big data | |
CN110543885B (en) | Method for interactively extracting high-resolution remote sensing image road and generating road network | |
CN108898605A (en) | A kind of grating map dividing method based on figure | |
CN106918311A (en) | Isolated tree crown mapping area automatic calculating method based on vehicle-mounted laser cloud data | |
CN105913469B (en) | TF/TA2 path planning methods based on skeleton drawing | |
CN105258704A (en) | Multi-scale space-time hot point path detection method based on rapid road network modeling | |
CN102096072B (en) | Method for automatically measuring urban parts | |
CN102881039B (en) | Based on the tree three-dimensional vector model construction method of laser three-dimensional scanning data | |
CN104484882A (en) | City region power line detection method based on airborne LiDAR (light detection and ranging) data | |
CN104899357A (en) | Topology data extraction method based on AutoCAD plane intersection engineering drawing | |
CN105701862A (en) | Ground object key point extraction method based on point cloud | |
CN105809194A (en) | Method for translating SAR image into optical image | |
CN107657636A (en) | A kind of method that route topography figure elevational point is automatically extracted based on mobile lidar data | |
CN112561944A (en) | Lane line extraction method based on vehicle-mounted laser point cloud | |
CN101783016A (en) | Crown appearance extract method based on shape analysis | |
CN108765478A (en) | It is a kind of to build the density clustering algorithm that separating monomer is built in point cloud |
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 |
Granted publication date: 20190111 Termination date: 20190819 |
|
CF01 | Termination of patent right due to non-payment of annual fee |