CN110246223A - A kind of tunnel modeling method and device - Google Patents
A kind of tunnel modeling method and device Download PDFInfo
- Publication number
- CN110246223A CN110246223A CN201910478093.1A CN201910478093A CN110246223A CN 110246223 A CN110246223 A CN 110246223A CN 201910478093 A CN201910478093 A CN 201910478093A CN 110246223 A CN110246223 A CN 110246223A
- Authority
- CN
- China
- Prior art keywords
- tunnel
- section
- jurisdiction
- point cloud
- image
- 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
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
- G06T17/20—Finite element generation, e.g. wire-frame surface description, tesselation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
A kind of tunnel modeling method and device, which comprises according to the tunnel three dimensional point cloud of acquisition, generate tunnel reflected image, and establish the mapping relations of the tunnel three dimensional point cloud Yu the tunnel reflected image;It identifies all vertical boundaries in the tunnel reflected image, and the tunnel reflected image is divided into one group of monocycle section of jurisdiction sub-image;One group of monocycle section of jurisdiction sub-image is identified, obtains the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image respectively;Obtain the corresponding point cloud data collection of each section of jurisdiction block in the tunnel three dimensional point cloud;The corresponding point cloud data collection of each section of jurisdiction block is optimized into matching with corresponding theoretical model in preset section of jurisdiction block theoretical model library, constructs tunnel overall model.Above scheme can be improved the efficiency and accuracy modeled to tunnel.
Description
Technical field
The present invention relates to technical field of rail traffic more particularly to a kind of tunnel modeling method and devices.
Background technique
With the continuous development in city, to mitigate traffic above-ground pressure, subway, which becomes, possesses the public of high carrying capacity
The chief component of traffic system.However, on the one hand, more and more buildings result in the change in tunnel near subway tunnel
Shape may cause to seriously affect to its safety and stability;On the other hand, subway is as city major public transportation tool,
Such as cave-in accident, the traffic system in entire city will be jeopardized.Therefore, just to the routine testing of entire tunnel system and maintenance
Seem extremely important.
When detecting to the deformation of subway tunnel, for conventional method based on manual inspection, accuracy and efficiency are lower, need
Consume a large amount of human and material resources, financial resources etc..In contrast to this, the application of three-dimensional laser scanning technique substantially increases measurement data
The efficiency and precision of acquisition have strong real-time, and data precision is high, and untouchable, initiative is strong, and energy all weather operations etc. is excellent
Point, meanwhile, the extensive three dimensional point cloud obtained not only includes tunnel coordinate information, is also believed comprising laser reflectivity
Breath, can accurately reflect tunnel surface disease state.Currently, the mainstream that three dimension laser scanning surreying has become tunnel structure detection becomes
Gesture.
Tunnel point cloud data based on acquisition, is established by Data Management Analysis and realizes rebuilding tunnel, and building tunnel is accurate
Threedimensional model is equally of great significance.As railway systems by leaps and bounds develop, to railway tunnel detection, maintenance etc.
It made higher requirement, hardly possible meets the need to various aspects to traditional manifestation mode based on two-dimensional map and mock-up
It asks, people prefer to the actual conditions for reflecting railway tunnel by more intuitive form, to provide for Measuring error personnel
More intuitive decision-making foundation.So three-dimensional visialization of tunnel technology is paid more and more attention, it is possible to provide more intuitive three-dimensional visible
Change tunnel geology model, more comprehensively analysis tunnel deformed state.Meanwhile tunnel three-dimensional visualization model can pass through finite element fraction
Analysis, accurately reflects tunnel structure stress condition, so as to navigate to tunnel stress abnormal position, realizes precisely maintenance, can also be with
Anticipation is made in the deformation and disease being likely to occur by stress condition to tunnel, so that tunnel is effectively prevent to be abnormal deformation,
It is of great immediate significance to the healthy and safe condition managing in tunnel.
However, the data processing of the tunnel three-dimensional point cloud based on acquisition at this stage and rapid modeling technology are still in a step
Section, wherein generally calculated in section deformation analytic process by circle fitting or ellipse fitting to complete data.Due to outer masterpiece
With the influence of noise, abnormal point etc. in the irregular deformation and point cloud data for leading to tunnel cross-section, above by circle fitting or
The method of person's ellipse fitting is difficult to carry out accurate profile data processing and calculating and further tunnel Holistic modeling, this is to tunnel
The deformation analysis of road structure produces very big interference, has seriously affected the efficiency and accuracy of tunnel analysis.
Summary of the invention
What the embodiment of the present invention solved is lower to the modeling accuracy of tunnel progress.
In order to solve the above technical problems, the embodiment of the invention provides a kind of tunnel modeling methods, comprising: according to acquisition
Tunnel three dimensional point cloud generates tunnel reflected image, and establishes the tunnel three dimensional point cloud and the tunnel back projection
The mapping relations of picture;Identify all vertical boundaries in the tunnel reflected image, and according to all vertical boundaries by institute
It states tunnel reflected image and is divided into one group of monocycle section of jurisdiction sub-image;One group of monocycle section of jurisdiction sub-image is identified, respectively
Obtain the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image;It is anti-according to the tunnel three dimensional point cloud and the tunnel
The mapping relations of projection picture, all vertical boundaries in the tunnel reflected image and each section of jurisdiction Zhang Danhuan sub-image
Corresponding section of jurisdiction block boundary obtains the corresponding point cloud data collection of each section of jurisdiction block in the tunnel three dimensional point cloud;It will be described each
Block corresponding point cloud data collection in section of jurisdiction optimizes matching, structure with corresponding theoretical model in preset section of jurisdiction block theoretical model library
Built tunnel overall model.
Optionally, the tunnel three-dimensional point cloud model according to acquisition generates tunnel reflected image, comprising: according to described
Tunnel three dimensional point cloud extracts tunnel central axes;The tunnel three dimensional point cloud is unfolded according to the tunnel central axes, it is raw
At the tunnel reflected image corresponding with the tunnel three dimensional point cloud.
Optionally, described according to the tunnel three dimensional point cloud, extract tunnel central axes, comprising: according to the tunnel
The Theoretical Design axis of three dimensional point cloud intercepts one group of equidistant tunnel cross sectional from the tunnel three dimensional point cloud;
Using RANSAC approximating method, the kernel of section of each tunnel cross sectional is extracted respectively;To each described tunnel cross sectional
Kernel of section carry out conic fitting, obtain initial central axes;The initial central axes are updated, until obtain
When the distance between the central axes of central axes and (i-1)-th update that i-th updates deviation is less than preset threshold, more by i-th
New central axes are as the tunnel central axes;It carries out jth time to the initial central axes using following steps to update: according to the
The central axes of j-1 update, intercept one group of equidistant tunnel cross sectional again from the tunnel three dimensional point cloud;Using
RANSAC approximating method extracts each of the one group of equidistant tunnel cross sectional intercepted again tunnel cross sectional respectively
Kernel of section;The kernel of section application conic fitting algorithm of this each obtained tunnel cross sectional is fitted
It extracts, obtains the central axes that jth time updates;1≤j≤i;I, j is positive integer.
Optionally, described that the tunnel three dimensional point cloud is unfolded according to the tunnel central axes, comprising: to be directed to the tunnel
Point p in road three dimensional point cloudi, obtain on the tunnel central axes with the point piApart from nearest point ci;With the point ci
For origin, with the point ciIt is coordinate that unit on the tunnel central axes, which cuts arrow T, per unit system arrow N and unit pair method arrow B,
Axis constructs part Frenet coordinate system;Parametrization equation is constructed according to the part Frenet coordinate system;Using the parametrization
Each of described tunnel three dimensional point cloud point is carried out parametric maps, by the tunnel three dimensional point cloud by equation
Each of point be mapped in the corresponding parameter plane of the parametrization equation.
Optionally, described that the tunnel three dimensional point cloud is unfolded according to the tunnel central axes, it generates and the tunnel
The corresponding tunnel reflected image of three dimensional point cloud, comprising: plane based on the parameter, according to the scanning of laser scanner
Frequency constructs coordinate grid;According to the parametrization point set for falling into each cell of coordinate grid, each unit is calculated
The pixel value of lattice;According to the pixel value of each cell, generate corresponding with the tunnel three dimensional point cloud described
Tunnel reflected image.
Optionally, described to obtain the corresponding point cloud data collection of each section of jurisdiction block in the tunnel three dimensional point cloud, comprising: root
According to all vertical boundaries and the parametrization equation in the tunnel reflected image, by the institute in the tunnel reflected image
There is the pixel reverse Mapping of vertical boundary to three-dimensional point cloud space, and the three-dimensional point cloud that mapping obtains is divided into one group of monocycle pipe
Piece point cloud subset;It, will according to the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image and the parametrization equation
The pixel-map of the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image is corresponding to the monocycle section of jurisdiction point cloud subset
The monocycle section of jurisdiction subspace Dian Yun, by monocycle section of jurisdiction point cloud segmentation at the corresponding point cloud data collection of each section of jurisdiction block.
Optionally, all vertical boundaries in the identification tunnel reflected image, comprising: mentioned using the boundary Canny
Algorithm is taken to identify all boundaries in the tunnel reflected image;Using Hough transformation algorithm, examined by preset slope
It surveys section and extracts all vertical boundaries in the tunnel reflected image.
Optionally, before identifying all vertical boundaries in the tunnel reflected image, further includes: anti-to the tunnel
Projection picture is sharpened processing.
Optionally, described to being identified in the sub-image of one group of monocycle section of jurisdiction, each section of jurisdiction Zhang Danhuan is obtained respectively
The corresponding section of jurisdiction block boundary of sub-image, comprising: use Hough change algorithm, extracted by preset slope detection section every
It binds in one monocycle section of jurisdiction sub-image the tilt boundary of block;According to inclining for the block that binds in each section of jurisdiction Zhang Danhuan sub-image
Bevel edge circle, design optimization match energy equation;According to the Optimized Matching energy equation, each section of jurisdiction Zhang Danhuan sub-image is extracted
Corresponding section of jurisdiction block boundary.
Optionally, the tilt boundary according to the block that binds in each section of jurisdiction Zhang Danhuan sub-image, design optimization
With algorithm, comprising: according to Structural Test of Tunnel Segments arrangement mode, construct monocycle tunnel segment structure theory template;According to the block boundary that binds
Theoretical position information, design template matching operator simultaneously drive the monocycle tunnel segment structure theory template, to each Zhang Dan
Endless tube piece sub-image carries out mobile Optimized Matching, obtains the Optimized Matching energy side based on the monocycle tunnel segment structure theory template
Journey.
Optionally, described by the corresponding point cloud data collection of each section of jurisdiction block and phase in preset section of jurisdiction block theoretical model library
The theoretical model answered optimizes matching, comprising: the relevant energy side of corresponding to j-th of the section of jurisdiction block point cloud data collection of design
Journey;It, will be in the corresponding point cloud data collection of j-th of section of jurisdiction block and the section of jurisdiction block theoretical model library according to the energy equation
K-th of theoretical model optimize matching;Wherein, the section of jurisdiction block theoretical model library is set according to the theory of all kinds of section of jurisdiction blocks
Meter structure constructs in advance.
Optionally, the relevant energy equation of corresponding to j-th of the section of jurisdiction block point cloud data collection of the design, comprising: use
ICP method is initially registered block corresponding point cloud data collection in j-th of section of jurisdiction with k-th of theoretical model;According to
The Euclidean distance error of j-th of section of jurisdiction block corresponding point cloud data collection and k-th of theoretical model, designs the energy
Equation.
Optionally, described will be in the corresponding point cloud data collection of j-th of section of jurisdiction block and the section of jurisdiction block theoretical model library
K-th of theoretical model optimize matching, comprising: according to the energy equation, solve the corresponding point of j-th of section of jurisdiction block
Cloud data set is matched with the optimal location of k-th of theoretical model, until the transition matrix convergence in the energy equation.
The embodiment of the invention also provides a kind of tunnel modeling devices, comprising: generation unit, for the tunnel according to acquisition
Three dimensional point cloud generates tunnel reflected image, and establishes the tunnel three dimensional point cloud and the tunnel reflected image
Mapping relations;Recognition unit, for identification all vertical boundaries in the tunnel reflected image;Cutting unit is used for basis
The tunnel reflected image is divided into one group of monocycle section of jurisdiction sub-image by all vertical boundaries;Section of jurisdiction block boundary obtains single
Member obtains the corresponding pipe of each section of jurisdiction Zhang Danhuan sub-image for identifying to one group of monocycle section of jurisdiction sub-image respectively
Piece block boundary;Point cloud data collection acquiring unit, for according to the tunnel three dimensional point cloud and the tunnel reflected image
Mapping relations, all vertical boundaries in the tunnel reflected image and the corresponding pipe of each section of jurisdiction Zhang Danhuan sub-image
Piece block boundary obtains the corresponding point cloud data collection of each section of jurisdiction block in the tunnel three dimensional point cloud;Construction unit is used for institute
The corresponding point cloud data collection of each section of jurisdiction block and corresponding theoretical model in preset section of jurisdiction block theoretical model library is stated to optimize
Match, constructs tunnel overall model.
Optionally, the generation unit, for extracting tunnel central axes according to the tunnel three dimensional point cloud;According to
The tunnel three dimensional point cloud is unfolded in the tunnel central axes, generates the tunnel corresponding with the tunnel three dimensional point cloud
Road reflected image.
Optionally, the generation unit, for the Theoretical Design axis according to the tunnel three dimensional point cloud, from described
One group of equidistant tunnel cross sectional is intercepted in tunnel three dimensional point cloud;Using RANSAC approximating method, each is extracted respectively
The kernel of section of tunnel cross sectional;Conic fitting is carried out to the kernel of section of each tunnel cross sectional, is obtained just
Beginning central axes;The initial central axes are updated, until what the central axes that obtained i-th updates were updated with (i-1)-th time
When the distance between central axes deviation is less than preset threshold, the central axes that i-th is updated are as the tunnel central axes;Using
Following steps carry out jth time to the initial central axes and update: the central axes updated according to jth -1 time, three-dimensional from the tunnel
Again one group of equidistant tunnel cross sectional is intercepted in point cloud data;Using RANSAC approximating method, extraction is described respectively cuts again
The kernel of section of each of the one group of equidistant tunnel cross sectional taken tunnel cross sectional;Each tunnel obtained to this
The kernel of section application conic fitting algorithm of cross section is fitted extraction, obtains the central axes that jth time updates;1≤j≤
i;I, j is positive integer.
Optionally, the generation unit, for for the point p in the tunnel three dimensional point cloudi, obtain the tunnel
On central axes with the point piApart from nearest point ci;With the point ciFor origin, with the point ciOn the tunnel central axes
Unit cut arrow T, per unit system arrow N and unit pair method arrow B be reference axis, construct part Frenet coordinate system;According to the part
Frenet coordinate system building parametrization equation;It, will be each in the tunnel three dimensional point cloud using the parametrization equation
A point carries out parametric maps, and it is corresponding that each of described tunnel three dimensional point cloud point is mapped to the parametrization equation
Parameter plane on.
Optionally, the generation unit is constructed for plane based on the parameter according to the scan frequency of laser scanner
Coordinate grid;According to the parametrization point set for falling into each cell of coordinate grid, the pixel of each cell is calculated
Value;According to the pixel value of each cell, the tunnel reflection corresponding with the tunnel three dimensional point cloud is generated
Image.
Optionally, the point cloud data collection acquiring unit, for according to all vertical edges in the tunnel reflected image
Boundary and the parametrization equation, by the pixel reverse Mapping of all vertical boundaries in the tunnel reflected image to three-dimensional point
Cloud space, and the three-dimensional point cloud that mapping obtains is divided into one group of monocycle section of jurisdiction point cloud subset;It is managed according to each Zhang Danhuan
The corresponding section of jurisdiction block boundary of piece sub-image and the parametrization equation, by the corresponding pipe of each section of jurisdiction Zhang Danhuan sub-image
The pixel-map of piece block boundary is to the corresponding monocycle section of jurisdiction subspace Dian Yun of the monocycle section of jurisdiction point cloud subset, by the monocycle pipe
Piece point cloud segmentation is at the corresponding point cloud data collection of each section of jurisdiction block.
Optionally, the recognition unit, for being identified in the tunnel reflected image using Canny boundary extraction algorithm
All boundaries;Using Hough transformation algorithm, extracted in the tunnel reflected image by preset slope detection section
All vertical boundaries.
Optionally, the recognition unit is also used to before identifying all vertical boundaries in the tunnel reflected image,
Processing is sharpened to the tunnel reflected image.
Optionally, the section of jurisdiction block boundary acquiring unit passes through preset slope for using Hough change algorithm
Detection interval extracts the tilt boundary for the block that binds in the sub-image of each section of jurisdiction Zhang Danhuan;According to each sub- shadow in the section of jurisdiction Zhang Danhuan
It binds as in the tilt boundary of block, design optimization matches energy equation;According to the Optimized Matching energy equation, each is extracted
The corresponding section of jurisdiction block boundary of monocycle section of jurisdiction sub-image.
Optionally, the section of jurisdiction block boundary acquiring unit, for constructing monocycle pipe according to Structural Test of Tunnel Segments arrangement mode
Chip architecture theory template;According to the theoretical position information for the block boundary that binds, design template matching operator simultaneously drives the monocycle pipe
Chip architecture theory template carries out mobile Optimized Matching to each section of jurisdiction Zhang Danhuan sub-image, obtains based on the monocycle pipe
The Optimized Matching energy equation of chip architecture theory template.
Optionally, the construction unit, for designing the relevant energy side of corresponding to j-th of section of jurisdiction block point cloud data collection
Journey;It, will be in the corresponding point cloud data collection of j-th of section of jurisdiction block and the section of jurisdiction block theoretical model library according to the energy equation
K-th of theoretical model optimize matching;Wherein, the section of jurisdiction block theoretical model library is set according to the theory of all kinds of section of jurisdiction blocks
Meter structure constructs in advance.
Optionally, the construction unit, for using ICP method, point cloud data collection corresponding to j-th of section of jurisdiction block
It is initially registered with k-th of theoretical model;According to the corresponding point cloud data collection of j-th of section of jurisdiction block and the kth
The Euclidean distance error of a theoretical model, designs the energy equation.
Optionally, the construction unit, for solving the corresponding point of j-th of section of jurisdiction block according to the energy equation
Cloud data set is matched with the optimal location of k-th of theoretical model, until the transition matrix convergence in the energy equation.
Compared with prior art, the technical solution of the embodiment of the present invention has the advantage that
According to the three dimensional point cloud got, tunnel reflected image is generated.Tunnel reflected image is divided into one group of list
Endless tube piece sub-image, and identify the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image, so as to obtain each
The corresponding point cloud data collection of section of jurisdiction block.The corresponding point cloud data collection of each obtained section of jurisdiction block and preset section of jurisdiction block are managed
It is matched by theoretical model corresponding in model library, so as to obtain each corresponding theoretical model of section of jurisdiction block, in turn
Tunnel overall model can be constructed.By the way that tunnel reflected image is divided, it may be possible to obtain accurate profile data;By obtaining
The corresponding theoretical model of section of jurisdiction block is taken, tunnel overall model can be constructed more accurately.
Detailed description of the invention
Fig. 1 is the flow chart of one of embodiment of the present invention tunnel modeling method;
Fig. 2 is the schematic diagram of one of embodiment of the present invention tunnel live image;
Fig. 3 is the corresponding tunnel three dimensional point cloud schematic diagram of Fig. 2;
Fig. 4 is one of embodiment of the present invention tunnel three dimensional point cloud schematic diagram;
Fig. 5 is the corresponding tunnel reflected image schematic diagram of Fig. 4;
Fig. 6 is the structural schematic diagram of one of embodiment of the present invention tunnel modeling device.
Specific embodiment
In the prior art, the data processing of the tunnel three-dimensional point cloud based on acquisition and rapid modeling technology are still in a step
Section, wherein generally calculated in section deformation analytic process by circle fitting or ellipse fitting to complete data.Due to outer masterpiece
With the influence of noise, abnormal point etc. in the irregular deformation and point cloud data for leading to tunnel cross-section, above by circle fitting or
The method of person's ellipse fitting is difficult to carry out accurate profile data processing and calculating and further tunnel Holistic modeling, this is to tunnel
The deformation analysis of road structure produces very big interference, has seriously affected the efficiency and accuracy of tunnel analysis.
In embodiments of the present invention, according to the three dimensional point cloud got, tunnel reflected image is generated.Tunnel is reflected
Image Segmentation identifies the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image at one group of monocycle section of jurisdiction sub-image, from
And the corresponding point cloud data collection of available each section of jurisdiction block.By the corresponding point cloud data collection of each obtained section of jurisdiction block
It is matched with corresponding theoretical model in preset section of jurisdiction block theoretical model library, it is corresponding so as to obtain each section of jurisdiction block
Theoretical model, and then tunnel overall model can be constructed.By the way that tunnel reflected image is divided, it may be possible to obtain accurate
Profile data;By obtaining the corresponding theoretical model of section of jurisdiction block, tunnel overall model can be constructed more accurately.
It is understandable to enable the above objects, features, and advantages of the embodiment of the present invention to become apparent, it is right with reference to the accompanying drawing
Specific embodiments of the present invention are described in detail.
The embodiment of the invention provides a kind of tunnel modeling methods, referring to Fig.1, carry out specifically below by way of specific steps
It is bright.
Step S101 generates tunnel reflected image according to the tunnel three dimensional point cloud of acquisition, and establishes the tunnel three
Tie up the mapping relations of point cloud data and the tunnel reflected image.
In specific implementation, tunnel live image can first be obtained.According to tunnel live image, tunnel three-dimensional point cloud is extracted
Data.
Referring to Fig. 2, the schematic diagram of one of embodiment of the present invention tunnel live image is given.By in Fig. 2
Tunnel live image is handled, and extracts corresponding tunnel three dimensional point cloud schematic diagram referring to Fig. 3.
In specific implementation, tunnel central axes can be extracted according to tunnel three dimensional point cloud.Getting tunnel axis
After line, the tunnel three dimensional point cloud got can be unfolded according to tunnel central axes, to obtain and tunnel three-dimensional
The corresponding tunnel reflected image of point cloud data.After obtaining tunnel reflected image, can establish tunnel three dimensional point cloud with
The corresponding mapping relations of tunnel reflected image.
In specific implementation, can be according to the corresponding Theoretical Design axis of tunnel three dimensional point cloud, one group of interception is equidistant
From tunnel cross sectional.For one group of equidistant tunnel cross sectional of interception, using RANSAC approximating method, extract respectively every
The corresponding kernel of section of one tunnel cross sectional.It, can be by institute after obtaining the corresponding kernel of section of each tunnel cross sectional
There is the kernel of section of tunnel cross sectional to carry out conic section set, obtains initial central axes.
For example, 6 equidistant tunnel cross sectionals of interception are as one group of tunnel cross sectional.Using random sampling consistency
(random sample consensus, RANSAC) approximating method, obtains respectively in the corresponding section of each tunnel cross sectional
The heart, namely obtain 6 kernel of section.Obtain 6 kernel of section are subjected to conic fitting, obtain initial central axes.
It, can be according to initial central axes, the progress that cuts section corresponding to tunnel cross sectional after obtaining initial central axes
Iterative fitting, until obtained curve convergence, using convergent curve as final tunnel central axes.
In specific implementation, after obtaining initial central axes, initial central axes can be updated, until obtain
When the distance between the central axes of central axes and (i-1)-th update that i-th updates deviation is less than preset threshold, more by i-th
New central axes are as finally determining tunnel central axes.
In embodiments of the present invention, it when carrying out jth time update to initial central axes, can be updated according to jth -1 time
Central axes intercept one group of equidistant tunnel cross sectional again from tunnel three dimensional point cloud;Later, using the fitting side RANSAC
Method extracts the kernel of section of each of the one group of equidistant tunnel cross sectional intercepted again tunnel cross sectional respectively;To this
The interface center applications conic fitting algorithm of secondary each obtained tunnel cross sectional is fitted extraction, obtains jth time
The central axes of update;1≤j≤i;I, j is positive integer.
Obtained central axes are iterated using the above method, until the distance between central axes updated twice deviation
Less than preset threshold.At this point, the central axes that last time is updated are as determining tunnel central axes.
In embodiments of the present invention, first time update is carried out to initial central axes, obtained central axes are to update for the first time
Central axes;The central axes updated to first time are updated, and obtained central axes are second of central axes updated, with such
It pushes away.
In specific implementation, after obtaining tunnel central axes, tunnel three-dimensional point cloud can be unfolded according to tunnel central axes
Data, to generate tunnel corresponding with tunnel three dimensional point cloud reflected image.
In embodiments of the present invention, for any point p on tunnel three dimensional point cloudi=(xi,yi,zi), it can obtain
Take on tunnel central axes with point piApart from nearest point ci.With point ciFor origin, point ciUnit on tunnel central axes cuts arrow
T, per unit system arrow N and unit pair method arrow B is reference axis, constructs part Frenet coordinate system.According to local Frenet coordinate system structure
Parametrization equation is built, by point piIt is mapped to parameter plane qi=(ui,vi), constructed parametrization equation are as follows:
Wherein, D is tunnel theory design radial, sign (ui) it is to seek uiSIN function.
Using above-mentioned parameter equation, each of tunnel three dimensional point cloud point can be parameterized accordingly
Mapping, so as to which each of tunnel three dimensional point cloud point is mapped in the corresponding parameter plane of parametrization equation,
To which tunnel three dimensional point cloud is deployed in parameter plane.
In specific implementation, it is based on parameter plane, coordinate grid can be constructed according to the scan frequency of laser scanner,
The resolution ratio of coordinate grid and the resolution ratio of laser scanner are related.It in practical applications, can be by the resolution ratio of coordinate grid
It is set as 0.002.According to the parametrization point set for falling into each cell in coordinate grid, the pixel of each cell is calculated
Value.After obtaining the pixel value of each cell, that is, produce the corresponding tunnel reflected image of tunnel three dimensional point cloud.
In embodiments of the present invention, the pixel value of following formula computing unit lattice can be used:
Wherein, AP is parametrization point set, | AP | for the gesture of parametrization point set AP.
By above-mentioned parameter mapping method, the corresponding tunnel reflected image of tunnel three dimensional point cloud can be obtained.
Referring to Fig. 4, the schematic diagram of one of embodiment of the present invention tunnel three dimensional point cloud is given, wherein a1、b1、
c1、d1、f1For the section of jurisdiction block number of the first endless tube piece;a2、b2、c2、d2、f2For the section of jurisdiction block number of the second endless tube piece.Reference Fig. 5,
It is the schematic diagram of the corresponding tunnel reflected image of Fig. 4.
Step S102 identifies all vertical boundaries in the tunnel reflected image, and the tunnel reflected image is divided
It is cut into one group of monocycle section of jurisdiction sub-image.
In specific implementation, all sides in the tunnel reflected image can be identified using Canny boundary extraction algorithm
Boundary.Later, it using Hough transformation algorithm, is extracted by preset slope detection section all vertical in tunnel reflected image
Straight boundary.According to obtained vertical boundary, tunnel reflected image is divided into one group of monocycle section of jurisdiction sub-image, realizes tunnel reflection
The first time of image is divided.
In practical applications, slope section can be set according to actual application scenarios.In embodiments of the present invention, may be used
To set [80~100] for slope detection section, slope is also in the boundary within section [80~100] as vertical
Boundary.
It in embodiments of the present invention, can also be to tunnel before all vertical boundaries in identification tunnel reflected image
Reflected image is sharpened processing, vertical so as to more accurately identify to enhance the boundary in tunnel reflected image
Boundary.
Step S103 identifies one group of monocycle section of jurisdiction sub-image, obtains each sub- shadow in the section of jurisdiction Zhang Danhuan respectively
As corresponding section of jurisdiction block boundary.
After obtaining one group of monocycle section of jurisdiction sub-image, each of these monocycle section of jurisdiction sub-image, identification can be directed to
Its corresponding all section of jurisdiction block boundary out.
In specific implementation, Hough transformation algorithm can be used, by preset slope detection section, is extracted each
It binds in the sub-image of the section of jurisdiction Zhang Danhuan the tilt boundary of block.In practical applications it is found that each section of jurisdiction Zhang Danhuan sub-image is corresponding
Tilt boundary be 2.Therefore, in embodiments of the present invention, corresponding two slope detection sections can be set, to obtain 2
Tilt boundary.
In specific implementation, two slope detection sections of setting can be respectively [40~50] and [- 50~-40].
It is understood that slope detection section may be arranged as other values, can be set according to actual application scenarios.
In specific implementation, corresponding 2 tilt boundaries of each section of jurisdiction Zhang Danhuan sub-image are being obtained, can be related to corresponding to
Optimized Matching algorithm.According to designed Optimized Matching algorithm, the section of jurisdiction block of each section of jurisdiction Zhang Danhuan sub-image can be extracted
Boundary.
In embodiments of the present invention, monocycle tunnel segment structure theory mould can be constructed according to Structural Test of Tunnel Segments arrangement mode
Plate.Theoretical position information based on the block boundary that binds, is related to template matching operator.It is a n × n by template matching operator definitions
Matrix, definite equation is as follows:
Wherein, n × n is the sum of all pixels of template matching operator, aijFor the i-th row in template matching operator, jth column pixel
Pixel value, B are the combination for the block boundary pixel that binds in template matching operator, and N is the non-block boundary picture that binds in template matching operator
The set of element.Monocycle tunnel segment structure theory template is driven to optimize matching by the template matching operator of design, design is based on
The Optimized Matching energy equation of monocycle tunnel segment structure theory template are as follows:
Wherein, KdesignFor template matching operator matrix, KhFor one and K in the sub-image of monocycle section of jurisdictiondesignIt is identical big
The matrix of zonule.By solving above-mentioned energy equation, each section of jurisdiction Zhang Danhuan sub-image corresponding section of jurisdiction block side can be extracted
Boundary.
Step S104 obtains the corresponding point cloud data collection of each section of jurisdiction block in the tunnel three dimensional point cloud.
It in specific implementation, can be anti-according to tunnel three dimensional point cloud and the mapping relations of tunnel reflected image, tunnel
All vertical boundaries and the corresponding section of jurisdiction boundary of each section of jurisdiction Zhang Danhuan sub-image in projection picture obtain tunnel three-dimensional point cloud
The corresponding point cloud data collection of each section of jurisdiction block in data.
In embodiments of the present invention, can be according to all vertical boundaries in tunnel reflected image, and equation is parameterized,
By the corresponding pixel-map of all vertical boundaries in tunnel reflected image to three-dimensional point cloud space, and the three-dimensional that mapping is obtained
Point cloud segmentation is at one group of monocycle section of jurisdiction point cloud subset.According to the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image and ginseng
Numberization equation, by the pixel-map of the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image to monocycle section of jurisdiction point cloud subset pair
The monocycle section of jurisdiction subspace Dian Yun answered, thus by monocycle section of jurisdiction point cloud segmentation at one group of section of jurisdiction block point cloud data collection.
Step S105, the corresponding point cloud data collection of each section of jurisdiction block is corresponding to preset section of jurisdiction block theoretical model library
Theoretical model optimize matching, construct tunnel overall model.
In specific implementation, section of jurisdiction block theoretical model can be constructed previously according to the Theoretical Design structure of all kinds of section of jurisdiction blocks
Library.For the corresponding point cloud data collection of j-th of section of jurisdiction block, corresponding energy equation is designed.Using designed energy
Equation optimizes the corresponding point cloud data collection of j-th of section of jurisdiction block with k-th of theoretical model and matches, and k-th of theoretical model is
Theoretical model corresponding with j-th of section of jurisdiction block in the block theoretical model library of section of jurisdiction.The corresponding point cloud data collection of all section of jurisdiction blocks is traversed,
It is final to realize tunnel Holistic modeling to find each corresponding theoretical model of section of jurisdiction block.
It in embodiments of the present invention, can be with for the corresponding point cloud data collection of j-th of section of jurisdiction block and k-th of theoretical model
The corresponding mass center of j-th of section of jurisdiction block, the corresponding mass center of k-th of theoretical model are first calculated separately, it, will by translation rotation matrix M
The mass center overlapping corresponding with k-th of theoretical model of the corresponding mass center of j-th of section of jurisdiction block, and using ICP method to j-th of section of jurisdiction block
Corresponding point cloud data collection is initially registered with k-th of theoretical model.According to the corresponding point cloud data collection of j-th of section of jurisdiction block with
The Euclidean distance error of k-th of theoretical model, design energy equation simultaneously solve.
In embodiments of the present invention, energy equation can be expressed from the next:
Wherein, { S, R, T } is the corresponding transition matrix of energy equation, and S is the change of scale matrix of k-th of theoretical model, R
For the spin matrix of k-th of theoretical model, T is the translation matrix of k-th of theoretical model,The point p concentrated for the corresponding point cloud data of j-th of section of jurisdiction blockiWith k-th of theory
Euclidean distance between model Ψ.By solving above formula, iteration optimization { S, R, T } matrix completes Optimized Matching mistake to restraining
Journey.
It can be seen that generating tunnel reflected image according to the three dimensional point cloud got.Tunnel reflected image is divided
At one group of monocycle section of jurisdiction sub-image, and identify the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image, so as to
To the corresponding point cloud data collection of each section of jurisdiction block.By the corresponding point cloud data collection of each obtained section of jurisdiction block with it is preset
Corresponding theoretical model is matched in the block theoretical model library of section of jurisdiction, so as to obtain the corresponding theoretical mould of each section of jurisdiction block
Type, and then tunnel overall model can be constructed.By the way that tunnel reflected image is divided, it may be possible to obtain accurate section number
According to;By obtaining the corresponding theoretical model of section of jurisdiction block, tunnel overall model can be constructed more accurately.
Referring to Fig. 6, the embodiment of the invention also provides a kind of tunnel modeling devices 60, comprising:
Generation unit 601 generates tunnel reflected image for the tunnel three dimensional point cloud according to acquisition, and establishes institute
State the mapping relations of tunnel three dimensional point cloud Yu the tunnel reflected image;
Recognition unit 602, for identification all vertical boundaries in the tunnel reflected image;
Cutting unit 603, for the tunnel reflected image to be divided into one group of monocycle according to all vertical boundaries
Section of jurisdiction sub-image;
Section of jurisdiction block boundary acquiring unit 604 obtains every respectively for identifying to one group of monocycle section of jurisdiction sub-image
The corresponding section of jurisdiction block boundary of one monocycle section of jurisdiction sub-image;
Point cloud data collection acquiring unit 605, for according to the tunnel three dimensional point cloud and the tunnel reflected image
Mapping relations, all vertical boundaries in the tunnel reflected image and each section of jurisdiction Zhang Danhuan sub-image it is corresponding
Section of jurisdiction block boundary obtains the corresponding point cloud data collection of each section of jurisdiction block in the tunnel three dimensional point cloud;
Construction unit 606 is used for each corresponding point cloud data collection of section of jurisdiction block and preset section of jurisdiction block theoretical model
Corresponding theoretical model optimizes matching in library, constructs tunnel overall model.
In specific implementation, the generation unit 601 can be used for extracting tunnel according to the tunnel three dimensional point cloud
Road central axes;The tunnel three dimensional point cloud is unfolded according to the tunnel central axes, generates and the tunnel three-dimensional point cloud number
According to the corresponding tunnel reflected image.
In specific implementation, the generation unit 601 can be used for being set according to the theory of the tunnel three dimensional point cloud
Meter shaft line intercepts one group of equidistant tunnel cross sectional from the tunnel three dimensional point cloud;Using RANSAC approximating method, divide
Indescribably take the kernel of section of each tunnel cross sectional;Conic section is carried out to the kernel of section of each tunnel cross sectional
Fitting, obtains initial central axes;The initial central axes are updated, until the central axes that update of obtained i-th and the
When the distance between the central axes of i-1 update deviation is less than preset threshold, the central axes that i-th is updated are as the tunnel
Central axes;Jth time is carried out to the initial central axes using following steps to update: the central axes updated according to jth -1 time, from institute
It states and intercepts one group of equidistant tunnel cross sectional in tunnel three dimensional point cloud again;Using RANSAC approximating method, extract respectively
Again the kernel of section of each of the equidistant tunnel cross sectional of one intercepted group tunnel cross sectional;This is obtained
The kernel of section application conic fitting algorithm of each tunnel cross sectional is fitted extraction, obtains in jth time update
Axis;1≤j≤i;I, j is positive integer.
In specific implementation, the generation unit 601 can be used for for the point in the tunnel three dimensional point cloud
pi, obtain on the tunnel central axes with the point piApart from nearest point ci;With the point ciFor origin, with the point ciInstitute
Stating the unit on tunnel central axes and cutting arrow T, per unit system arrow N and unit pair method arrow B is reference axis, constructs part Frenet coordinate
System;Parametrization equation is constructed according to the part Frenet coordinate system;Using the parametrization equation, by the tunnel three-dimensional point
Each of cloud data point carries out parametric maps, each of described tunnel three dimensional point cloud point is mapped to described
It parameterizes in the corresponding parameter plane of equation.
In specific implementation, the generation unit 601, can be used for plane based on the parameter, according to laser scanner
Scan frequency construct coordinate grid;According to the parametrization point set for falling into each cell of coordinate grid, calculate each
The pixel value of a cell;According to the pixel value of each cell, generate corresponding with the tunnel three dimensional point cloud
The tunnel reflected image.
In specific implementation, the point cloud data collection acquiring unit 605, can be used for according in the tunnel reflected image
All vertical boundaries and the parametrization equation, the pixel of all vertical boundaries in the tunnel reflected image is reverse
It is mapped to three-dimensional point cloud space, and the three-dimensional point cloud that mapping obtains is divided into one group of monocycle section of jurisdiction point cloud subset;According to described
The corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image and the parametrization equation, by each section of jurisdiction Zhang Danhuan
The pixel-map of the corresponding section of jurisdiction block boundary of image to the corresponding monocycle section of jurisdiction subspace Dian Yun of the monocycle section of jurisdiction point cloud subset,
By monocycle section of jurisdiction point cloud segmentation at the corresponding point cloud data collection of each section of jurisdiction block.
In specific implementation, the recognition unit 602 can be used for identifying the tunnel using Canny boundary extraction algorithm
All boundaries in road reflected image;Using Hough transformation algorithm, the tunnel is extracted by preset slope detection section
All vertical boundaries in road reflected image.
In specific implementation, the recognition unit 602 can be also used for identifying owning in the tunnel reflected image
Before vertical boundary, processing is sharpened to the tunnel reflected image.
In specific implementation, the section of jurisdiction block boundary acquiring unit 604 can be used for passing through using Hough change algorithm
Extract the tilt boundary for the block that binds in the sub-image of each section of jurisdiction Zhang Danhuan in preset slope detection section;According to described each
It binds in the sub-image of the section of jurisdiction Zhang Danhuan the tilt boundary of block, design optimization matches energy equation;According to the Optimized Matching energy
Equation extracts the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image.
In specific implementation, the section of jurisdiction block boundary acquiring unit 604, can be used for according to Structural Test of Tunnel Segments arrangement side
Formula constructs monocycle tunnel segment structure theory template;According to the theoretical position information for the block boundary that binds, design template matching operator is simultaneously driven
The monocycle tunnel segment structure theory template is moved, mobile Optimized Matching is carried out to each section of jurisdiction Zhang Danhuan sub-image, obtains base
In the Optimized Matching energy equation of the monocycle tunnel segment structure theory template.
In specific implementation, the construction unit 606 can be used for designing point cloud data corresponding with j-th of section of jurisdiction block
Collect relevant energy equation;According to the energy equation, by the corresponding point cloud data collection of j-th of section of jurisdiction block and the section of jurisdiction
K-th of theoretical model in block theoretical model library optimizes matching;Wherein, the section of jurisdiction block theoretical model library is according to all kinds of
What the Theoretical Design structure of section of jurisdiction block constructed in advance.
In specific implementation, the construction unit 606 can be used for using ICP method, to j-th of section of jurisdiction block pair
The point cloud data collection answered is initially registered with k-th of theoretical model;According to corresponding cloud number of j-th of section of jurisdiction block
According to the Euclidean distance error of collection and k-th of theoretical model, the energy equation is designed.
In specific implementation, the construction unit 606 can be used for solving j-th of pipe according to the energy equation
The corresponding point cloud data collection of tile is matched with the optimal location of k-th of theoretical model, until turning in the energy equation
Change matrix convergence.
Those of ordinary skill in the art will appreciate that all or part of the steps in the various methods of above-described embodiment is can
It is completed with instructing relevant hardware by program, which can be stored in a computer readable storage medium, storage
Medium may include: ROM, RAM, disk or CD etc..
Although present disclosure is as above, present invention is not limited to this.Anyone skilled in the art are not departing from this
It in the spirit and scope of invention, can make various changes or modifications, therefore protection scope of the present invention should be with claim institute
Subject to the range of restriction.
Claims (26)
1. a kind of tunnel modeling method characterized by comprising
According to the tunnel three dimensional point cloud of acquisition, generate tunnel reflected image, and establish the tunnel three dimensional point cloud with
The mapping relations of the tunnel reflected image;
It identifies all vertical boundaries in the tunnel reflected image, and is reflected the tunnel according to all vertical boundaries
Image Segmentation is at one group of monocycle section of jurisdiction sub-image;
One group of monocycle section of jurisdiction sub-image is identified, obtains the corresponding section of jurisdiction block of each section of jurisdiction Zhang Danhuan sub-image respectively
Boundary;
According in the mapping relations of the tunnel three dimensional point cloud and the tunnel reflected image, the tunnel reflected image
All vertical boundaries and the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image, obtain the tunnel three-dimensional point cloud
The corresponding point cloud data collection of each section of jurisdiction block in data;
By the corresponding point cloud data collection of each section of jurisdiction block and corresponding theoretical model in preset section of jurisdiction block theoretical model library into
Row Optimized Matching constructs tunnel overall model.
2. tunnel modeling method as described in claim 1, which is characterized in that the tunnel three-dimensional point cloud mould according to acquisition
Type generates tunnel reflected image, comprising:
According to the tunnel three dimensional point cloud, tunnel central axes are extracted;
The tunnel three dimensional point cloud is unfolded according to the tunnel central axes, generates corresponding with the tunnel three dimensional point cloud
The tunnel reflected image.
3. tunnel modeling method as claimed in claim 2, which is characterized in that it is described according to the tunnel three dimensional point cloud,
Extract tunnel central axes, comprising:
According to the Theoretical Design axis of the tunnel three dimensional point cloud, one group etc. is intercepted from the tunnel three dimensional point cloud
Apart from tunnel cross sectional;
Using RANSAC approximating method, the kernel of section of each tunnel cross sectional is extracted respectively;
Conic fitting is carried out to the kernel of section of each tunnel cross sectional, obtains initial central axes;
The initial central axes are updated, until central axes and the central axes of (i-1)-th update that obtained i-th updates
The distance between deviation be less than preset threshold when, using i-th update central axes as the tunnel central axes;
Jth time is carried out to the initial central axes using following steps to update: the central axes updated according to jth -1 time, from described
Again one group of equidistant tunnel cross sectional is intercepted in tunnel three dimensional point cloud;Using RANSAC approximating method, institute is extracted respectively
State the kernel of section of each of the one group of equidistant tunnel cross sectional intercepted again tunnel cross sectional;This is obtained every
The kernel of section application conic fitting algorithm of one tunnel cross sectional is fitted extraction, obtains the axis that jth time updates
Line;
1≤j≤i;I, j is positive integer.
4. tunnel modeling method as claimed in claim 2, which is characterized in that described according to the expansion of the tunnel central axes
Tunnel three dimensional point cloud, comprising:
For the point p in the tunnel three dimensional point cloudi, obtain on the tunnel central axes with the point piDistance is nearest
Point ci;
With the point ciFor origin, with the point ciUnit on the tunnel central axes cuts arrow T, per unit system arrow N and unit pair
Method swears that B is reference axis, constructs part Frenet coordinate system;
Parametrization equation is constructed according to the part Frenet coordinate system;
Using the parametrization equation, each of described tunnel three dimensional point cloud point is subjected to parametric maps, by institute
Each of tunnel three dimensional point cloud point is stated to be mapped in the corresponding parameter plane of the parametrization equation.
5. tunnel modeling method as claimed in claim 4, which is characterized in that described according to the expansion of the tunnel central axes
Tunnel three dimensional point cloud generates the tunnel reflected image corresponding with the tunnel three dimensional point cloud, comprising:
Plane based on the parameter constructs coordinate grid according to the scan frequency of laser scanner;
According to the parametrization point set for falling into each cell of coordinate grid, the pixel value of each cell is calculated;
According to the pixel value of each cell, the tunnel reflection corresponding with the tunnel three dimensional point cloud is generated
Image.
6. tunnel modeling method as claimed in claim 4, which is characterized in that described to obtain in the tunnel three dimensional point cloud
The corresponding point cloud data collection of each section of jurisdiction block, comprising:
According to all vertical boundaries and the parametrization equation in the tunnel reflected image, by the tunnel reflected image
In the pixel reverse Mappings of all vertical boundaries be divided into one group to three-dimensional point cloud space, and by the obtained three-dimensional point cloud of mapping
Monocycle section of jurisdiction point cloud subset;
It, will be described each according to the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image and the parametrization equation
The pixel-map of the corresponding section of jurisdiction block boundary of the section of jurisdiction Zhang Danhuan sub-image is to the corresponding monocycle pipe of the monocycle section of jurisdiction point cloud subset
Piece point cloud subspace, by monocycle section of jurisdiction point cloud segmentation at the corresponding point cloud data collection of each section of jurisdiction block.
7. tunnel modeling method as described in claim 1, which is characterized in that the institute in the identification tunnel reflected image
There is vertical boundary, comprising:
All boundaries in the tunnel reflected image are identified using Canny boundary extraction algorithm;
Using Hough transformation algorithm, extracted by preset slope detection section all vertical in the tunnel reflected image
Straight boundary.
8. tunnel modeling method as claimed in claim 7, which is characterized in that all in the tunnel reflected image identifying
Before vertical boundary, further includes:
Processing is sharpened to the tunnel reflected image.
9. tunnel modeling method as described in claim 1, which is characterized in that described in the sub-image of one group of monocycle section of jurisdiction
It is identified, obtains the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image respectively, comprising:
Using Hough change algorithm, is extracted in the sub-image of each section of jurisdiction Zhang Danhuan and bound by preset slope detection section
The tilt boundary of block;
According to the tilt boundary for the block that binds in each section of jurisdiction Zhang Danhuan sub-image, design optimization matches energy equation;
According to the Optimized Matching energy equation, the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image is extracted.
10. tunnel modeling method as claimed in claim 9, which is characterized in that described according to each section of jurisdiction Zhang Danhuan
It binds in image the tilt boundary of block, design optimization matching algorithm, comprising:
According to Structural Test of Tunnel Segments arrangement mode, monocycle tunnel segment structure theory template is constructed;
According to the theoretical position information for the block boundary that binds, design template matching operator simultaneously drives the monocycle tunnel segment structure theory mould
Plate carries out mobile Optimized Matching to each section of jurisdiction Zhang Danhuan sub-image, obtains based on the monocycle tunnel segment structure theory mould
The Optimized Matching energy equation of plate.
11. tunnel modeling method as described in claim 1, which is characterized in that described by described corresponding cloud of each section of jurisdiction block
Data set optimizes matching with corresponding theoretical model in preset section of jurisdiction block theoretical model library, comprising:
Design the relevant energy equation of corresponding to j-th of section of jurisdiction block point cloud data collection;
According to the energy equation, by the corresponding point cloud data collection of j-th of section of jurisdiction block and the section of jurisdiction block theoretical model library
In k-th of theoretical model optimize matching;
Wherein, the section of jurisdiction block theoretical model library is constructed in advance according to the Theoretical Design structure of all kinds of section of jurisdiction blocks.
12. tunnel modeling method as claimed in claim 11, which is characterized in that the design is corresponding with j-th of section of jurisdiction block
The relevant energy equation of point cloud data collection, comprising:
Using ICP method, the corresponding point cloud data collection of j-th of section of jurisdiction block and k-th of theoretical model are carried out initial
Registration;
According to the Euclidean distance error of j-th of section of jurisdiction block corresponding point cloud data collection and k-th of theoretical model, design
The energy equation.
13. tunnel modeling method as claimed in claim 12, which is characterized in that described that j-th of section of jurisdiction block is corresponding
Point cloud data collection is optimized with k-th of theoretical model in the section of jurisdiction block theoretical model library and is matched, comprising:
According to the energy equation, the corresponding point cloud data collection of j-th of section of jurisdiction block and k-th of theoretical model are solved
Optimal location matching, until the transition matrix convergence in the energy equation.
14. a kind of tunnel modeling device characterized by comprising
Generation unit generates tunnel reflected image for the tunnel three dimensional point cloud according to acquisition, and establishes the tunnel three
Tie up the mapping relations of point cloud data and the tunnel reflected image;
Recognition unit, for identification all vertical boundaries in the tunnel reflected image;
Cutting unit, for the tunnel reflected image to be divided into one group of sub- shadow in monocycle section of jurisdiction according to all vertical boundaries
Picture;
Section of jurisdiction block boundary acquiring unit obtains each Zhang Dan for identifying to one group of monocycle section of jurisdiction sub-image respectively
The corresponding section of jurisdiction block boundary of endless tube piece sub-image;
Point cloud data collection acquiring unit, for being closed according to the mapping of the tunnel three dimensional point cloud and the tunnel reflected image
System, all vertical boundaries in the tunnel reflected image and each section of jurisdiction Zhang Danhuan sub-image corresponding section of jurisdiction block side
Boundary obtains the corresponding point cloud data collection of each section of jurisdiction block in the tunnel three dimensional point cloud;
Construction unit, for the corresponding point cloud data collection of each section of jurisdiction block is corresponding to preset section of jurisdiction block theoretical model library
Theoretical model optimize matching, construct tunnel overall model.
15. tunnel modeling device as claimed in claim 14, which is characterized in that the generation unit, for according to the tunnel
Road three dimensional point cloud extracts tunnel central axes;The tunnel three dimensional point cloud is unfolded according to the tunnel central axes, generates
The tunnel reflected image corresponding with the tunnel three dimensional point cloud.
16. tunnel modeling device as claimed in claim 15, which is characterized in that the generation unit, for according to the tunnel
It is transversal to intercept one group of equidistant tunnel from the tunnel three dimensional point cloud for the Theoretical Design axis of road three dimensional point cloud
Face;Using RANSAC approximating method, the kernel of section of each tunnel cross sectional is extracted respectively;It is transversal to each described tunnel
The kernel of section in face carries out conic fitting, obtains initial central axes;The initial central axes are updated, until obtaining
The central axes that update of i-th and (i-1)-th update the distance between central axes deviation less than preset threshold when, by i-th
The central axes of update are as the tunnel central axes;Jth time is carried out to the initial central axes using following steps to update: according to
The central axes of jth -1 time update, intercept one group of equidistant tunnel cross sectional again from the tunnel three dimensional point cloud;Using
RANSAC approximating method extracts each of the one group of equidistant tunnel cross sectional intercepted again tunnel cross sectional respectively
Kernel of section;The kernel of section application conic fitting algorithm of this each obtained tunnel cross sectional is fitted
It extracts, obtains the central axes that jth time updates;1≤j≤i;I, j is positive integer.
17. tunnel modeling device as claimed in claim 14, which is characterized in that the generation unit, for being directed to the tunnel
Point p in road three dimensional point cloudi, obtain on the tunnel central axes with the point piApart from nearest point ci;With the point ci
For origin, with the point ciIt is coordinate that unit on the tunnel central axes, which cuts arrow T, per unit system arrow N and unit pair method arrow B,
Axis constructs part Frenet coordinate system;Parametrization equation is constructed according to the part Frenet coordinate system;Using the parametrization
Each of described tunnel three dimensional point cloud point is carried out parametric maps, by the tunnel three dimensional point cloud by equation
Each of point be mapped in the corresponding parameter plane of the parametrization equation.
18. tunnel modeling device as claimed in claim 17, which is characterized in that the generation unit, for being based on the ginseng
Number plane constructs coordinate grid according to the scan frequency of laser scanner;According to falling into each cell of the coordinate grid
Parametrization point set, calculate the pixel value of each cell;According to the pixel value of each cell, generate with it is described
The corresponding tunnel reflected image of tunnel three dimensional point cloud.
19. tunnel modeling method as claimed in claim 17, which is characterized in that the point cloud data collection acquiring unit is used for
It, will be in the tunnel reflected image according to all vertical boundaries and the parametrization equation in the tunnel reflected image
The pixel reverse Mapping of all vertical boundaries is divided into one group of monocycle to three-dimensional point cloud space, and by the three-dimensional point cloud that mapping obtains
Section of jurisdiction point cloud subset;According to the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image and the parametrization equation,
By the pixel-map of the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image to the monocycle section of jurisdiction point cloud subset pair
The monocycle section of jurisdiction subspace Dian Yun answered, by monocycle section of jurisdiction point cloud segmentation at the corresponding point cloud data collection of each section of jurisdiction block.
20. tunnel modeling device as claimed in claim 14, which is characterized in that the recognition unit, for using the side Canny
Boundary's extraction algorithm identifies all boundaries in the tunnel reflected image;Using Hough transformation algorithm, by preset oblique
Rate detection interval extracts all vertical boundaries in the tunnel reflected image.
21. tunnel modeling device as claimed in claim 20, which is characterized in that the recognition unit is also used in identification institute
Before stating all vertical boundaries in tunnel reflected image, processing is sharpened to the tunnel reflected image.
22. tunnel modeling device as claimed in claim 14, which is characterized in that the section of jurisdiction block boundary acquiring unit is used for
Using Hough change algorithm, the block that binds in the sub-image of each section of jurisdiction Zhang Danhuan is extracted by preset slope detection section
Tilt boundary;According to the tilt boundary for the block that binds in each section of jurisdiction Zhang Danhuan sub-image, design optimization matches energy equation;
According to the Optimized Matching energy equation, the corresponding section of jurisdiction block boundary of each section of jurisdiction Zhang Danhuan sub-image is extracted.
23. tunnel modeling device as claimed in claim 22, which is characterized in that the section of jurisdiction block boundary acquiring unit is used for
According to Structural Test of Tunnel Segments arrangement mode, monocycle tunnel segment structure theory template is constructed;Believed according to the theoretical position for the block boundary that binds
Breath, design template matching operator simultaneously drives the monocycle tunnel segment structure theory template, to each section of jurisdiction Zhang Danhuan sub-image
Mobile Optimized Matching is carried out, the Optimized Matching energy equation based on the monocycle tunnel segment structure theory template is obtained.
24. tunnel modeling device as claimed in claim 14, which is characterized in that the construction unit, for designing and j-th
The relevant energy equation of the corresponding point cloud data collection of section of jurisdiction block;It is according to the energy equation, j-th of section of jurisdiction block is corresponding
Point cloud data collection is optimized with k-th of theoretical model in the section of jurisdiction block theoretical model library and is matched;Wherein, the section of jurisdiction block
Theoretical model library is constructed in advance according to the Theoretical Design structure of all kinds of section of jurisdiction blocks.
25. tunnel modeling device as claimed in claim 24, which is characterized in that the construction unit, for using the side ICP
Method is initially registered block corresponding point cloud data collection in j-th of section of jurisdiction with k-th of theoretical model;According to described
The Euclidean distance error of j-th of section of jurisdiction block corresponding point cloud data collection and k-th of theoretical model, designs the energy side
Journey.
26. tunnel modeling device as claimed in claim 25, which is characterized in that the construction unit, for according to the energy
Equation is measured, the corresponding point cloud data collection of j-th of section of jurisdiction block is solved and is matched with the optimal location of k-th of theoretical model,
Until the transition matrix convergence in the energy equation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910478093.1A CN110246223B (en) | 2019-06-03 | 2019-06-03 | Tunnel modeling method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910478093.1A CN110246223B (en) | 2019-06-03 | 2019-06-03 | Tunnel modeling method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110246223A true CN110246223A (en) | 2019-09-17 |
CN110246223B CN110246223B (en) | 2022-12-13 |
Family
ID=67885826
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910478093.1A Active CN110246223B (en) | 2019-06-03 | 2019-06-03 | Tunnel modeling method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110246223B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110889898A (en) * | 2019-12-11 | 2020-03-17 | 南京航空航天大学 | Modeling method suitable for appearance of single aviation conduit |
CN111710027A (en) * | 2020-05-25 | 2020-09-25 | 南京林业大学 | Tunnel three-dimensional geometric reconstruction method considering data-driven segment segmentation and model-driven segment assembly |
CN113838207A (en) * | 2021-11-25 | 2021-12-24 | 腾讯科技(深圳)有限公司 | Map data processing method and device, readable medium and electronic equipment |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108230442A (en) * | 2018-01-24 | 2018-06-29 | 上海岩土工程勘察设计研究院有限公司 | A kind of shield tunnel three-dimensional emulation method |
CN109373921A (en) * | 2018-10-26 | 2019-02-22 | 南京航空航天大学 | A kind of tunnel monitoring method and device |
-
2019
- 2019-06-03 CN CN201910478093.1A patent/CN110246223B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108230442A (en) * | 2018-01-24 | 2018-06-29 | 上海岩土工程勘察设计研究院有限公司 | A kind of shield tunnel three-dimensional emulation method |
CN109373921A (en) * | 2018-10-26 | 2019-02-22 | 南京航空航天大学 | A kind of tunnel monitoring method and device |
Non-Patent Citations (1)
Title |
---|
朱宁宁: "三维激光扫描在地铁隧道形变监测中的应用", 《测绘工程》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110889898A (en) * | 2019-12-11 | 2020-03-17 | 南京航空航天大学 | Modeling method suitable for appearance of single aviation conduit |
CN111710027A (en) * | 2020-05-25 | 2020-09-25 | 南京林业大学 | Tunnel three-dimensional geometric reconstruction method considering data-driven segment segmentation and model-driven segment assembly |
CN111710027B (en) * | 2020-05-25 | 2021-05-04 | 南京林业大学 | Tunnel three-dimensional geometric reconstruction method |
CN113838207A (en) * | 2021-11-25 | 2021-12-24 | 腾讯科技(深圳)有限公司 | Map data processing method and device, readable medium and electronic equipment |
Also Published As
Publication number | Publication date |
---|---|
CN110246223B (en) | 2022-12-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Xie et al. | Development of a 3D modeling algorithm for tunnel deformation monitoring based on terrestrial laser scanning | |
Xu et al. | 3D reconstruction and measurement of surface defects in prefabricated elements using point clouds | |
Bosché et al. | The value of integrating Scan-to-BIM and Scan-vs-BIM techniques for construction monitoring using laser scanning and BIM: The case of cylindrical MEP components | |
Yan et al. | Automated extraction of structural elements in steel girder bridges from laser point clouds | |
CN103605135B (en) | A kind of road feature extraction method based on section subdivision | |
CN106127857B (en) | The on-board LiDAR data modeling method of integrated data driving and model-driven | |
CN103020342B (en) | Method for extracting contour and corner of building from ground LiDAR data | |
Li et al. | Terrestrial laser scanning assisted flatness quality assessment for two different types of concrete surfaces | |
CN110030951A (en) | A kind of diversion vertical shaft defect inspection method based on three-dimensional laser scanning technique | |
Mizoguchi et al. | Quantitative scaling evaluation of concrete structures based on terrestrial laser scanning | |
CN110246223A (en) | A kind of tunnel modeling method and device | |
CN103729853B (en) | High score remote sensing image building under three-dimension GIS auxiliary damages detection method | |
CN103324916B (en) | Vehicle-mounted and aviation LiDAR data method for registering based on building profile | |
Yin et al. | Application of 3D laser scanning technology for image data processing in the protection of ancient building sites through deep learning | |
CN103020966B (en) | A kind of aviation based on contour of building constraint and ground LiDAR data autoegistration method | |
CN104021556A (en) | Heterological remote-sensing image registration method based on geometric structure similarity | |
CN109682303A (en) | Curve PC track girder precision three-dimensional detection method based on BIM technology | |
Hammoudi et al. | Extracting wire-frame models of street facades from 3D point clouds and the corresponding cadastral map | |
Duan et al. | Reconstruction of shield tunnel lining using point cloud | |
Zhao et al. | Automated recognition and measurement based on three-dimensional point clouds to connect precast concrete components | |
CN108897937B (en) | Method for automatically converting CAD (computer aided design) data of civil aviation airport into DEM (digital elevation model) data | |
Ostrowski et al. | Analysis of 3D building models accuracy based on the airborne laser scanning point clouds | |
Chen et al. | A novel image-based approach for interactive characterization of rock fracture spacing in a tunnel face | |
Hou et al. | Review of GPR activities in Civil Infrastructures: Data analysis and applications | |
Luo et al. | 3D deformation monitoring method for temporary structures based on multi-thread LiDAR point cloud |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |