CN110363849A - A kind of interior three-dimensional modeling method and system - Google Patents
A kind of interior three-dimensional modeling method and system Download PDFInfo
- Publication number
- CN110363849A CN110363849A CN201810321028.3A CN201810321028A CN110363849A CN 110363849 A CN110363849 A CN 110363849A CN 201810321028 A CN201810321028 A CN 201810321028A CN 110363849 A CN110363849 A CN 110363849A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- plane
- point
- cloud data
- metope
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 104
- 238000012545 processing Methods 0.000 claims abstract description 32
- 238000013499 data model Methods 0.000 claims description 63
- 238000004422 calculation algorithm Methods 0.000 claims description 42
- 238000004590 computer program Methods 0.000 claims description 20
- 238000000605 extraction Methods 0.000 claims description 16
- 238000013459 approach Methods 0.000 claims description 13
- 238000001914 filtration Methods 0.000 claims description 13
- 238000005070 sampling Methods 0.000 claims description 12
- 238000003860 storage Methods 0.000 claims description 10
- 239000000284 extract Substances 0.000 claims description 9
- 238000005259 measurement Methods 0.000 claims description 7
- 241001269238 Data Species 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 6
- 238000005520 cutting process Methods 0.000 claims description 4
- 241000700647 Variola virus Species 0.000 claims description 2
- 230000001815 facial effect Effects 0.000 claims description 2
- 230000008569 process Effects 0.000 description 37
- 230000006870 function Effects 0.000 description 13
- 238000010586 diagram Methods 0.000 description 10
- 239000011159 matrix material Substances 0.000 description 10
- 230000009466 transformation Effects 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 5
- 230000011218 segmentation Effects 0.000 description 5
- 238000012800 visualization Methods 0.000 description 4
- 241000233838 Commelina Species 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000013519 translation Methods 0.000 description 3
- 238000011282 treatment Methods 0.000 description 3
- HUTDUHSNJYTCAR-UHFFFAOYSA-N ancymidol Chemical compound C1=CC(OC)=CC=C1C(O)(C=1C=NC=NC=1)C1CC1 HUTDUHSNJYTCAR-UHFFFAOYSA-N 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 238000012217 deletion Methods 0.000 description 2
- 230000037430 deletion Effects 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000012804 iterative process Methods 0.000 description 2
- 230000033001 locomotion Effects 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 239000002184 metal Substances 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 244000062793 Sorghum vulgare Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007850 degeneration Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 235000019713 millet Nutrition 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000011946 reduction process Methods 0.000 description 1
- 230000003716 rejuvenation Effects 0.000 description 1
- 238000012732 spatial analysis Methods 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/05—Geographic models
-
- 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
- G06T17/10—Constructive solid geometry [CSG] using solid primitives, e.g. cylinders, cubes
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Computer Graphics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Remote Sensing (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Processing Or Creating Images (AREA)
Abstract
The embodiment of the invention provides a kind of indoor three-dimensional modeling method and systems.The embodiment of the present invention is registrated using collected point cloud data, without demarcating to scene, is simplified data acquisition, is expanded the application range of indoor three-dimensional modeling method.In addition, the embodiment of the present invention also passes through inverted image delete processing, the laser reflection deleted in laser point cloud data acquisition influences, and improves the precision of reconstruction model.In addition, the indoor threedimensional model that the embodiment of the present invention is obtained based on reconstruction, may be implemented to accurately measure real indoor environment.
Description
Technical field
The present invention relates to map structuring technical fields, and in particular to a kind of interior three-dimensional modeling method and system.
Background technique
With the rapid development of computer technology, wireless location technology, earth observation system data and information system and mobile interchange technology, it is based on
The service of position becomes a reality and is widely applied in practice.Indoors in environment, as airport hall, exhibition room, warehouse,
In the environment such as building, supermarket, library, underground parking, railway station and subway station, it is often necessary to determine mobile terminal or its hold
The location information of the person of having, facility and article indoors, and corresponding add is provided and is such as navigated, search inquiry etc. is based on indoor position
The application service set.Therefore, work of the building of indoor three-dimensional (3D) model as a kind of basis, is increasingly paid attention to.
However, due to interior architecture enormous amount and indoor environment is complicated and changeable, such as supermarket, the periodically-varied of exhibition room finishing layout,
Stern challenge is proposed to the timeliness of the map rejuvenation of indoor location service.
Currently, during constructing indoor map, in the original CAD diagram paper of the building of no indoor map to be built
In the case where, the indoor ground that needs staff to measure indoor each room wall, door and window equidimension, and then built
Figure.The building process of this indoor map needs the staff for largely having professional drawing ability to carry out in person to indoor environment
Mapping, it is high to staff Capability Requirement and the amount of labour is big.
To solve the above problems, the prior art proposes the technology using laser point cloud technology building indoor model, one
Determine solve the problems, such as artificial room's modeling in degree.Then, it is above-mentioned in the prior art, in collection room when point cloud data, need
The motion profile for acquiring equipment forms a closed loop indoors, otherwise will affect the precision of Registration of Measuring Data, therefore limit
State the use scope of technology.
Summary of the invention
Technical problems to be solved of the embodiment of the present invention are to provide a kind of indoor three-dimensional modeling method and system, utilize a cloud
The acquisition position of data is registrated, and is carried out data acquisition according to closed loop path without acquiring equipment, is simplified data and acquired
Journey expands the application range of indoor three-dimensional modeling method.
In order to solve the above technical problems, interior three-dimensional modeling method provided in an embodiment of the present invention, comprising:
The point cloud data of the target interior space is obtained, and the point cloud data is registrated, generates point cloud data model,
Generate point cloud data model;
Basic geometric primitive is extracted from the point cloud data model;
Curve reestablishing is carried out to basic geometric primitive, the geometric primitive after being rebuild, and to the geometric primitive after reconstruction
Spliced, obtains the threedimensional model of the target interior space.
Preferably, in the above method, the point cloud data is registrated, comprising:
It is adjacent to position since collected first frame point cloud data using undistinguishable closest approach iteration ICP algorithm
Two frame point cloud datas be registrated and merged, and finally obtain the point cloud data model being registrated.
Preferably, in the above method, basic geometric primitive is extracted from the point cloud data model, comprising:
Using the consistent RANSAC algorithm of random sampling, curved surface knowledge is carried out to the point cloud data in the point cloud data model
Not, the type of curved surface is identified;
According to the type of curved surface, using corresponding the geometric parameter equation, extraction obtains the basic geometric primitive.
Preferably, in the above method, basic geometric primitive is extracted from the point cloud data model, further includes:
Basic geometric primitive in the point cloud data model is calculated, determines the corresponding ground point of floor
Cloud;
The normal vector of the ground point cloud is calculated, and determines the central point of the ground point cloud;
To the basic geometric primitive of each non-horizontal directions, each point of the basic geometric primitive is calculated in described
Whether the subpoint in the normal direction of heart point is less than preset threshold at a distance from the central point, and according to the distance, and determining should
Whether point is inverted image point;
Delete the inverted image point in the basic geometric primitive.
Preferably, in the above method, basic geometric primitive is extracted from the point cloud data model, further includes:
Plane in the point cloud data model is traversed, the angle between the normal vector of any two plane is calculated
And preset coordinate origin is to the distance difference of any two plane;
When the angle is less than the first pre-determined threshold and the distance difference is less than the second pre-determined threshold, by this any two
A plane merges into a plane.
Preferably, in the above method, basic geometric primitive is extracted from the point cloud data model, further includes:
Metope plane is extracted from the point cloud data model, and extracts from the metope plane that there are partial dots
First metope plane of cloud missing, the metope plane are vertical with the floor;
According to the space of the floor and the first metope plane, the direction of the first metope plane is calculated;
According to the direction of the first metope plane, four vertex of the first metope plane are calculated;
Using four vertex of the first metope plane and direction, constructs one and completely refer to metope point cloud;
Subtract each other described with reference to metope point cloud with corresponding cloud of the first metope plane, it is corresponding to obtain door opening plane
Door opening point cloud.
Preferably, in the above method, curve reestablishing is carried out to the basic geometric primitive, comprising:
Surface fitting is carried out to the point cloud data of the basic geometric primitive, it is corresponding heavy to obtain the basic geometric primitive
Geometric primitive after building.
Preferably, in the above method, the geometric primitive after the reconstruction is spliced, obtains the target interior space
Threedimensional model the step of, comprising:
According to the direction of the normal vector of plane, isolated from the point cloud of the geometric primitive after each reconstruction floor and
Corresponding first point of ceiling plane is converged conjunction and the corresponding second point of metope plane converges conjunction;
According to the height of point, the point in closing is converged by described first point and is divided to ceiling point set or ground point set,
And is carried out by plane fitting, obtains the target interior space for the point in the ceiling point set and ground point set respectively
Ceiling plane and floor;
Determine the target interior space the first principal direction in the vertical direction and the second main side in the horizontal direction
To with third principal direction;
According to corresponding cloud of first principal direction and each metope plane, the wall of the target interior space is generated
Face;
By the metope of the target interior space, respectively with the ceiling plane and floor of the target interior space
Mutually splice, and is inserted into the corresponding rectangle of door opening plane, obtains the geometric primitive of the target interior space and carries out triangle gridding
Filtering, generates the threedimensional model of the target interior space.
Preferably, in the above method, determine the target interior space the first principal direction in the vertical direction and
The second principal direction and third principal direction of horizontal direction, comprising:
According to the direction perpendicular with floor, the first principal direction on vertical direction is obtained;
The direction of all metope planes is clustered, multiple cluster centres are obtained, select wherein group's number it is maximum and
Orthogonal two cluster centres in direction, by the direction of two cluster centres, as the second principal direction in horizontal direction
With third principal direction.
Preferably, it in the above method, according to corresponding cloud of first principal direction and each metope plane, generates
Before the metope of the target interior space, first principal direction, the second principal direction and third principal direction are further utilized, it is right
Each plane is corrected.
Preferably, the above method further include:
Receive the inquiry operation that user is directed to threedimensional model input;
Respond the inquiry operation and calculate and export corresponding query result, the query result include space coordinate, away from
From, area or volume.
The embodiment of the invention also provides a kind of indoor 3 d modeling systems, comprising:
Point cloud data acquisition module, for obtaining the point cloud data of the target interior space, and and to the point cloud data into
Row registration, generates point cloud data model;
Point cloud data processing module, for extracting basic geometric primitive from the point cloud data model;
Three-dimensional modeling module, for carrying out curve reestablishing to the basic geometric primitive, the geometric primitive after being rebuild,
And the geometric primitive after reconstruction is spliced, obtain the threedimensional model of the target interior space.
Preferably, in above-mentioned indoor 3 d modeling system, the point cloud data acquisition module includes:
Registration unit is opened for using undistinguishable closest approach iteration ICP algorithm from collected first frame point cloud data
Begin, the two frame point cloud datas adjacent to position are registrated and are merged, and finally obtain the point cloud data model being registrated.
Preferably, in above-mentioned indoor 3 d modeling system, the point cloud data processing module includes:
Model cutting unit, for using the consistent RANSAC algorithm of random sampling, to the point in the point cloud data model
Cloud data carry out curved surface identification, identify the type of curved surface;It is mentioned according to the type of curved surface using corresponding the geometric parameter equation
Obtain the basic geometric primitive.
Preferably, in above-mentioned indoor 3 d modeling system, the point cloud data processing module further include:
Inverted image deletes unit and determines ground for calculating the basic geometric primitive in the point cloud data model
The corresponding ground point cloud of plane;The normal vector of the ground point cloud is calculated, and determines the central point of the ground point cloud;To each
The basic geometric primitive of a non-horizontal directions calculates each point of the basic geometric primitive in the normal direction of the central point
Whether subpoint is less than preset threshold at a distance from the central point, and according to the distance, determines whether the point is inverted image point;
Delete the inverted image point in the basic geometric primitive.
Preferably, in above-mentioned indoor 3 d modeling system, the point cloud data processing module further include:
Plane combining unit calculates any two plane for traversing to the plane in the point cloud data model
Normal vector between angle and preset coordinate origin to any two plane distance difference;In the angle less than
One pre-determined threshold and when the distance difference is less than the second pre-determined threshold, merges into a plane for any two plane.
Preferably, in above-mentioned indoor 3 d modeling system, the point cloud data processing module further include:
Door opening extraction unit, for extracting metope plane from the point cloud data model, and from the metope plane
In extract there are the first metope plane of partial dot cloud missing, the metope plane is vertical with the floor;According to institute
The space of floor and the first metope plane is stated, the direction of the first metope plane is calculated;It is flat according to first metope
The direction in face calculates four vertex of the first metope plane;Using four vertex of the first metope plane and direction,
Construction one completely refers to metope point cloud;By the reference metope point cloud and corresponding cloud phase of the first metope plane
Subtract, obtains the corresponding door opening point cloud of door opening plane.
Preferably, in above-mentioned indoor 3 d modeling system, the three-dimensional modeling module includes:
Plane fitting unit carries out surface fitting for the point cloud data to the basic geometric primitive, obtains the base
Geometric primitive after the corresponding reconstruction of this geometric primitive.
Preferably, in above-mentioned indoor 3 d modeling system, the three-dimensional modeling module further include:
3D Model Reconstruction unit, for the direction according to the normal vector of plane, from the point of the geometric primitive after each reconstruction
Floor is isolated in cloud and corresponding first point of ceiling plane is converged conjunction and the corresponding second point of metope plane converges
It closes;According to the height of point, the point in closing is converged by described first point and is divided to ceiling point set or ground point set, and respectively
To the point in the ceiling point set and ground point set, plane fitting is carried out, the smallpox of the target interior space is obtained
Plate plane and floor;Determine the target interior space the first principal direction in the vertical direction and in the horizontal direction
Second principal direction and third principal direction;According to corresponding cloud of first principal direction and each metope plane, described in generation
The metope of the target interior space;It is flat with the ceiling of the target interior space respectively by the metope of the target interior space
Face is mutually spliced with floor, and is inserted into the corresponding rectangle of door opening plane, obtains the geometric primitive of the target interior space simultaneously
Triangle gridding filtering is carried out, the threedimensional model of the target interior space is generated.
Preferably, in above-mentioned indoor 3 d modeling system, the 3D Model Reconstruction unit is also used to according to described first
Corresponding cloud of principal direction and each metope plane before the metope for generating the target interior space, further utilizes institute
The first principal direction, the second principal direction and third principal direction are stated, each plane is corrected.
Preferably, in above-mentioned indoor 3 d modeling system, the 3D Model Reconstruction unit is also used to basis and floor
Perpendicular direction obtains the first principal direction on vertical direction;The direction of all metope planes is clustered, is obtained multiple
Cluster centre selects wherein group's number maximum and orthogonal two cluster centres in direction, by two cluster centres
Direction, as the second principal direction and third principal direction in horizontal direction.
Preferably, above-mentioned indoor 3 d modeling system further include:
Measurement module, the inquiry operation for being directed to threedimensional model input for receiving user;Respond the inquiry operation
Corresponding query result is calculated and exports, the query result includes space coordinate, distance, area or volume.
The embodiment of the invention also provides another indoor 3 d modeling systems, comprising: memory, processor and is stored in
On memory and the computer program that can run on a processor, realized such as when the computer program is executed by the processor
Above the step of indoor three-dimensional modeling method.
The embodiment of the invention also provides a kind of computer readable storage medium, deposited on the computer readable storage medium
Computer program is contained, the step of indoor three-dimensional modeling method as described above is realized when the computer program is executed by processor
Suddenly.
Compared with prior art, indoor three-dimensional modeling method provided in an embodiment of the present invention and system, utilize point cloud data
Acquisition position be registrated, without acquire equipment installation closed loop path carry out data acquisition, simplify data acquisition, expand
The big application range of indoor three-dimensional modeling method.In addition, the embodiment of the present invention also passes through inverted image delete processing, laser is deleted
Laser reflection in point cloud data acquisition influences, and improves the precision of reconstruction model.In addition, the embodiment of the present invention is based on rebuilding
The indoor threedimensional model arrived may be implemented to accurately measure real indoor environment.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below by institute in the description to the embodiment of the present invention
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention
Example, for those of ordinary skill in the art, without creative efforts, can also obtain according to these attached drawings
Obtain other attached drawings.
Fig. 1 is a kind of flow diagram of indoor three-dimensional modeling method provided in an embodiment of the present invention;
Fig. 2 is a kind of exemplary diagram that inverted image is deleted in the embodiment of the present invention;
Fig. 3 is a kind of exemplary diagram of plane over-segmentation in the embodiment of the present invention;
A kind of Fig. 4 exemplary diagram of angle between two plane normal direction in the embodiment of the present invention;
Fig. 5 is a kind of exemplary diagram of coordinate origin and two plane distances in the embodiment of the present invention;
Fig. 6 is a kind of structural schematic diagram of indoor 3 d modeling system provided in an embodiment of the present invention;
Fig. 7 is another structural schematic diagram of indoor 3 d modeling system provided in an embodiment of the present invention;
Fig. 8 is another structural schematic diagram of indoor 3 d modeling system provided in an embodiment of the present invention;
Fig. 9 is another structural schematic diagram of indoor 3 d modeling system provided in an embodiment of the present invention.
Specific embodiment
To keep the technical problem to be solved in the present invention, technical solution and advantage clearer, below in conjunction with attached drawing and tool
Body embodiment is described in detail.In the following description, such as specific configuration is provided and the specific detail of component is only
In order to help comprehensive understanding the embodiment of the present invention.It therefore, it will be apparent to those skilled in the art that can be to reality described herein
Example is applied to make various changes and modifications without departing from scope and spirit of the present invention.In addition, for clarity and brevity, it is omitted pair
The description of known function and construction.
It should be understood that " one embodiment " or " embodiment " that specification is mentioned in the whole text mean it is related with embodiment
A particular feature, structure, or characteristic is included at least one embodiment of the present invention.Therefore, occur everywhere in the whole instruction
" in one embodiment " or " in one embodiment " not necessarily refer to identical embodiment.In addition, these specific features, knot
Structure or characteristic can combine in any suitable manner in one or more embodiments.
In various embodiments of the present invention, it should be appreciated that the size of the serial number of following each processes is not meant to execute suitable
Sequence it is successive, the execution of each process sequence should be determined by its function and internal logic, the implementation without coping with the embodiment of the present invention
Process constitutes any restriction.
The embodiment of the invention provides a kind of indoor three-dimensional modeling methods, are used to threedimensional model in rapid build room.Pass through
Indoor laser point cloud data is acquired and be registrated, point cloud data model is generated;Basic geometry base is isolated from point cloud data model
Member, such as plane, cylindrical surface etc. describe indoor scene using these basic geometric primitives;To these basic geometric primitives into
Row fitting, calculates relevant parameter, reconstructs and draws out the geometric primitive after the corresponding reconstruction of these basic geometric primitives.Then,
Reconstruction after splicing reconstruct refers to primitive, to realize from point cloud data to the reconstruction of indoor threedimensional model.
Fig. 1 is please referred to, the embodiment of the invention provides a kind of indoor three-dimensional modeling methods, comprising:
Step 11, the point cloud data of the target interior space is obtained, and the point cloud data is registrated, generates point cloud number
According to model.
In embodiments of the present invention, laser point cloud data can be acquired by laser point cloud equipment.Laser point cloud data is
Refer to the space coordinate for obtaining each sampled point of body surface under the same space referential using laser, what is obtained is a series of tables
Up to the set of object space distribution and the massive point of target surface characteristic.Laser point cloud data generally includes the position of sampled point
(such as D coordinates value) and characteristic information (such as color information and laser intensity angle value).Illustrate to simplify, in the embodiment of the present invention
Sometimes laser point cloud data is also referred to as point cloud data.
Laser point cloud data is acquired by laser point cloud equipment, which can integrate in staff
Knapsack or moveable acquisition platform on.Needing the laser point cloud number to the target interior space of indoor map to be built
When according to being acquired, which traverses the major trunk roads in entire room, and in above-mentioned mistake with the movement of staff
Cheng Zhong, laser point cloud equipment can be spaced setting time and carry out one acquisition one frame of acquisition to the laser point cloud data in entire space
Point cloud data, alternatively, laser point cloud equipment is often moved a certain distance in the target interior space with regard to carrying out according to default acquisition trajectories
One acquisition obtains a frame point cloud data.By above procedure, the multiframe point cloud data of the target interior space can be obtained, often
Frame point cloud data corresponds to a time point and an acquisition position, and (acquisition position refers to that laser point cloud equipment carries out one acquisition
When specific location).Two frame point cloud datas of adjacent acquisition position acquisition are the point cloud data of consecutive frame.
After collecting point cloud data, need to be registrated point cloud data.Registration refers to different location acquisition
Laser point cloud data converts the process to the same coordinate system.Specifically, being adopted in the collection process of laser point cloud equipment
Each frame laser point cloud data of collection is relative to the space coordinate of laser point cloud data acquisition moment laser point cloud equipment
For system, and different acquisition position, the space coordinates of laser point cloud equipment are different.In order to construct in target chamber to be built
The three-dimensional figure in space needs to relocate the laser point cloud data under different spaces coordinate system by registration, generates one
Three-dimensional figure under a unified coordinate system.
The embodiment of the present invention is when carrying out cloud data registration processing, according to the acquisition position of the point cloud data, to institute
It states point cloud data to be registrated, specifically, undistinguishable closest approach iteration (ICP, Iterative Closest can be used
Point) algorithm, in the point cloud data first frame point cloud data and the second frame point cloud data be registrated and merged, obtain
To being registrated point cloud data, wherein the acquisition position of the first frame point cloud data and the second frame point cloud data is adjacent.Here
First frame point cloud data can be start frame in all frames, a certain frame being also possible to except start frame.Then, to described cloud
Remaining point cloud data (not yet carrying out the point cloud data of registration process) in data, executes following steps frame by frame: remaining from described
The next frame point cloud data adjacent with the acquisition position for being registrated point cloud data is selected in remaining point cloud data, as point subject to registration
Cloud data have been registrated point cloud data and the point cloud data subject to registration is matched to described using the undistinguishable ICP algorithm
It is quasi- and merge, it obtains updated described being registrated point cloud data.It, can be according to most after all point cloud datas are disposed
Whole registration point cloud data, obtains the point cloud data model of the target interior space.
Here, the embodiment of the present invention uses undistinguishable ICP algorithm, is directly matched using the adjacent point cloud of acquisition position
Standard, therefore do not have to feature point for calibration in prior scene indoors, data acquisition is carried out according to closed loop path without acquisition equipment,
Data acquisition is simplified, the application range of indoor three-dimensional modeling method is expanded.Furthermore it is also possible to certainly using algorithm routine
It is dynamic to judge whether registration succeeds, it is ensured that the precision of registration.In addition, the embodiment of the present invention can also use K-
Dimensional tree (i.e. KD tree) algorithm accelerates to search closest approach, and solving ICP algorithm causes because computationally intensive with Quasi velosity
Slow problem can be quickly generated a cloud 3D model.
Step 12, basic geometric primitive is extracted from the point cloud data model.
Here, the embodiment of the present invention can be by using consistent (RANSAC, the Random Sample of random sampling
Consensus) algorithm carries out curved surface identification to the point cloud data in the point cloud data model, identifies the type of curved surface.So
Afterwards, according to the type of curved surface, using corresponding the geometric parameter equation, extraction obtains the basic geometric primitive.Specifically, described
Basic geometric primitive includes plane and curved surface (such as cylinder).
According to common surface classification method, characteristic surface includes quadratic surface (plane is the quadratic surface degenerated), mistake
Cross curved surface and free form surface (non-uniform rational B-spline (NURBS, Non-Uniform Rational B-Splines) curved surface).
Quadratic surface includes common spherical surface, cylindrical surface, circular conical surface, paraboloid etc..In actual scene, surface in kind usually can be by
Or approximation is made of plane and quadratic surface.In embodiments of the present invention, it is contemplated that the most common one is planes for the interior space (such as
Ground, ceiling and metope are usually all plane), it can simplify to be identified and being extracted for plane.
In the point cloud data collection process of certain indoor scenes, due to there is metal skin, smooth surface etc.
Reflection can form the noise spot of similar inverted image in point cloud data.These inverted images will affect the reconstruction of threedimensional model, therefore this hair
Bright embodiment can also provide a kind of processing side that inverted image is deleted after extracting basic geometric primitive in above-mentioned steps 12
Formula, to improve the precision of following model reconstruction, which is specifically included: to the basic geometry in the point cloud data model
Primitive is calculated, and determines the corresponding ground point cloud of floor;The normal vector of the ground point cloud is calculated, and is determined describedly
The central point of millet cake cloud;To the basic geometric primitive of each non-horizontal directions, each point of the basic geometric primitive is calculated
Whether it is less than default threshold at a distance from the central point, and according to the distance in the subpoint in the normal direction of the central point
Value, determines whether the point is inverted image point;Delete the inverted image point in the basic geometric primitive.Here preset threshold, can basis
Specific scene optimizes and is arranged.
Further, it is contemplated that the curved surface in the measurement error of point cloud data, registration error and curved surface identification, which is divided, to be missed
Difference, approximately the same plane may be divided into multiple and different sub- planes, therefore, in above-mentioned steps 12, can also extract base
After this geometric primitive, alternatively, processing is merged to above-mentioned sub- plane, specifically, the conjunction after above-mentioned inverted image delete processing
And handling includes: to traverse to the plane in the point cloud data model, between the normal vector for calculating any two plane
The distance difference of angle and preset coordinate origin to any two plane;In the angle less than the first pre-determined threshold and institute
When stating distance difference less than the second pre-determined threshold, which is merged into a plane.Here first, second is pre-
Gating limit, can empirically be worth and be configured, and be optimized according to the reconstructing three-dimensional model result of the target interior space.
In actual scene, there may be one or more doors on metope, each door corresponds to a door opening on metope.
In point cloud data collection process, door can be opened, make that door opening is presented on metope, thus the point cloud data of collected metope
In, by point cloud data corresponding to missing door opening.Due to sampled point of the point cloud data in some plane usually have it is certain close
Degree, therefore can identify that there are the planes of fractional-sample point missing, and then calculate door opening.Above-mentioned in the embodiment of the present invention
In step 12, door opening plane and its corresponding point cloud data (door opening point cloud) can also be obtained: from the point in the following way
Metope plane is extracted in cloud data model, and extracts from the metope plane that there are the first metopes of partial dot cloud missing
Plane, here, the metope plane are vertical with the floor, and the normal vector of floor is usually closer to vertical side
To, can with include vertical direction some Direction interval, identify floor;Then, according to the floor and
The space of one metope plane calculates the direction of the first metope plane;According to the direction of the first metope plane, institute is calculated
State four vertex of the first metope plane;And then using four vertex and direction of the first metope plane, construction one is complete
Whole reference metope point cloud;Subtract each other described with reference to metope point cloud with corresponding cloud of the first metope plane, so as to
Obtain the corresponding door opening point cloud of door opening plane.
Step 13, curve reestablishing is carried out to the basic geometric primitive, the geometric primitive after being rebuild, and to reconstruction after
Geometric primitive spliced, obtain the target interior space threedimensional model.
It here, can be to the basic geometric primitive extracted in order to preferably rebuild the threedimensional model of the target interior space
Curve reestablishing is carried out, to improve the precision of reconstruction model.Specifically, can be carried out to the point cloud data of the basic geometric primitive
Surface fitting, the geometric primitive after obtaining the corresponding reconstruction of the basic geometric primitive.Likewise, to simplify the process, after reconstruction
Geometric primitive be specifically as follows plane.
The embodiment of the present invention may comprise steps of when splicing to the geometric primitive after the reconstruction:
1) according to the direction of the normal vector of plane, floor is isolated from the point cloud of the geometric primitive after each reconstruction
First point corresponding with ceiling plane is converged conjunction and the corresponding second point of metope plane converges conjunction.
2) according to the height of point, the point in closing is converged by described first point and is divided to ceiling point set or ground point set
It closes, and is carried out by plane fitting, obtains the target Interior Space for the point in the ceiling point set and ground point set respectively
Between ceiling plane and floor;
Here it is possible to calculate the described first point height for converging the central point of conjunction, and whether high according to the height of each point
In the height of the central point, the point in closing is converged by described first point and is added separately to ceiling point set and ground point set.
3) the target interior space the first principal direction in the vertical direction and in the horizontal direction second main is determined
Direction and third principal direction;
Here it is possible to obtain the first principal direction on vertical direction according to the direction perpendicular with floor.Then,
The direction of all metope planes in geometric primitive after reconstruction is clustered, multiple cluster centres is obtained, selects wherein
Group's number maximum and orthogonal two cluster centres in direction, by the direction of two cluster centres, as in horizontal direction
The second principal direction and third principal direction.
After obtaining above three principal direction, the embodiment of the present invention can also carry out above three principal direction orthogonal
Correction, so that above three principal direction is mutually perpendicular to.Further, the embodiment of the present invention can also utilize the described first main side
To, the second principal direction and third principal direction, to each plane (such as ceiling plane and floor, door opening plane, metope plane
Deng) corrected, for example, if some plane (can specifically pass through plane normal vector and principal direction close to some principal direction
Between angle whether be less than preset threshold value to determine), then can guarantee mesh in this way on the plane rectification to the principal direction
Marking practical vertical structure in the interior space is also vertical in final threedimensional model.
4) according to corresponding cloud of first principal direction and each metope plane, the target interior space is generated
Metope.
It take the first principal direction as the y-axis of the metope plane, according to y-axis and metope here it is possible to be directed to each metope plane
The multiplication cross of plane normal vector determines the x-axis of the metope plane, and then, the interior point in corresponding cloud of the metope plane is projected to
In the metope plane and transform to plane coordinate system.And then according to the point in the plane coordinate system, corresponding bounding box is constructed
Son, and according to four vertex for surrounding box, metope shape is determined, to obtain the corresponding metope of metope plane.It is logical
Above step is crossed, the metope of the target interior space can be obtained.
5) flat with the ceiling plane of the target interior space and ground respectively by the metope of the target interior space
Face is mutually spliced, and is inserted into the corresponding rectangle of door opening plane, is obtained the geometric primitive of the target interior space and is carried out the triangulation network
Lattice filtering, generates the threedimensional model of the target interior space.
Here, it is contemplated that data acquisition or processing error, after the completion of plain splice in, can also further judge appoint
Whether two planes of anticipating need to splice, for example, when the first plane vertex to the second plane first distance be less than it is preset away from
From thresholding, then can active the boundary of the first plane and/or the second plane is adjusted, so that two planes is connected to one
It rises, to guarantee to generate closed three-dimensional space model as far as possible.
By above-mentioned splicing, the geometric primitive of the target interior space is obtained, these geometric primitives, such as metope,
The vertex representation of polygon usually by quadrangle or including more polygon (being more than 4 sides).It shows and draws for the ease of machine
System processing and the problems such as avoid ghost image, the embodiment of the present invention can also geometric primitive to the target interior space described above simultaneously
Triangle gridding filtering is carried out, the threedimensional model of the target interior space is generated.
From above step as can be seen that interior three-dimensional modeling method provided in an embodiment of the present invention, utilizes acquisition position phase
Adjacent point cloud data is registrated, without feature point for calibration in prior scene indoors, without acquisition equipment according to closed-loop
Diameter carries out data acquisition, simplifies data acquisition, expands the application range of indoor three-dimensional modeling method.In addition, this hair
Bright embodiment also passes through inverted image delete processing, and the laser reflection deleted in laser point cloud data acquisition influences, and improves reconstruction
The precision of model.
After obtaining the indoor threedimensional model in above-mentioned steps 13, the above-mentioned indoor three-dimensional modeling method of the embodiment of the present invention,
The inquiry operation that user is directed to threedimensional model input, such as detection user's selection operation can also be received, determines the selection
Operate selected object (point, line or face in such as threedimensional model) and the inquiry instruction for the object;Then, it responds
The inquiry operation calculates and exports corresponding query result, the query result may include space coordinate, distance, area or
The information such as volume.By above step, the embodiment of the present invention realizes the indoor threedimensional model obtained based on reconstruction, measurement model
The parameters such as the area of the distances of interior any two points and plane, realization accurately measure real indoor environment.
The process of the indoor three-dimensional modeling method of the embodiment of the present invention is illustrated above.To help to better understand
Process is stated, more careful description will be carried out to the specific implementation of each step of the above process further combined with attached drawing below.
The specific implementation of above-mentioned steps 11:
Acquisition point cloud data is the premise of analysis processing point cloud data, and the quality of the quality of data of acquisition is to subsequent processing
There is strong influence with analysis, it is thus determined that a good data acquisition modes, obtain the point cloud data of high quality as far as possible
It is very important.Simultaneously for the requirement of subsequent registration, acquire the setting of density, acquisition mode all and be need to pay close attention to it is interior
Hold.
Point cloud data is acquired there are two types of mode, one is the introducing extrinsic calibration positions of the prior art as reference marker,
Directly by the cloud data registration in every frame into global coordinate system.But in large scale scene acquisition, this mode needs
Lay a large amount of mark position, very time and effort consuming.It is used in the embodiment of the present invention, it is another scheme, i.e., directly using adjacent
The point cloud data of acquisition position is registrated, without measuring control, without a large amount of mark points of offer.The embodiment of the present invention
By the way of being directly registrable, based on the function that point Yun Ku (PCL, Point Cloud Library) is provided, realize continuous
Data acquisition.In order to obtain better point cloud data, the pretreatment before being registrated simultaneously during acquisition is (as point cloud is gone
Make an uproar, isolated point is rejected) and registration operation, acquisition, registration and the synchronous progress of registration result visualization are realized, can be seen at any time
The point cloud and the point cloud data model after registration for observing acquisition, realize the effect of What You See Is What You Get, provide for next step processing
Reference.It is the detailed process of point cloud acquisition below.
1) laser sensor is adjusted to horizontal direction.
2) since some starting acquisition position point of the target interior space, sensor is placed, is acquired every time towards needs
The mobile preset distance in direction reaches next acquisition position point, such as 1 meter to 2 meters.Guarantee that sensor does not have as far as possible in this process
There is rotation.
3) point cloud data of present frame is acquired in current acquisition position point and be registrated, and can observe program output
Registration result can export the information of registration failure, can resurvey at this time when registration result is undesirable.It is registrated successfully
Afterwards, it can continue to translate toward next acquisition position point to carry out the acquisition of the data of next frame, be completed until data acquire.Separately
Outside, as a result, data picker can be intuitive to see the result of registration after system synchronously visualization registration.
In above-mentioned collection process, avoid as far as possible moving object it is very more or have to sensor time for seriously blocking into
Row acquisition.
Point cloud registering mainly needs the problem of handling two parts, and first is how to find the part of overlapping and find phase
The corresponding relationship at point cloud data midpoint and point that adjacent twice sweep obtains, second is that transformation is solved by the corresponding relationship of point
Matrix.
The treatment process of point cloud registering mainly includes two steps, and the first step finds the three-dimensional data frame for currently needing to be registrated and it
Before corresponding relationship between the three-dimensional data frame that collects, determine corresponding points pair.The positional relationship that second step passes through corresponding points
Solve the transformation relation under present frame to world coordinate system.The output of point cloud registering is obtained complete after multiframe Registration of Measuring Data
Point cloud data model (three-dimensional point cloud model of place), which has recorded scene using cloud as representation
Information is supplied to subsequent step and is handled.
The present invention uses undistinguishable closest approach iteration (ICP) point cloud registration algorithm, to by pretreated cloud number
According to being registrated, complete cloud three-dimensional scene models relatively can be obtained, tentatively obtain the entity indicated by spatial point
Model.Closest approach iteration (ICP) algorithm is a kind of undistinguishable registration Algorithm, based on the vacation for only having minor shifts between cloud
If choosing the closest approach of each point in the step of looking for corresponding points as possible after the distance between defining point and point
Match, recycles current corresponding solution transformation matrix to be converted, then the two processes of continuous iteration are until convergence.Point pair
Search and the solution of transformation parameter are all iterative process, which belongs to accuracy registration algorithm.It is assumed that with Q representation space
One point set, the alignment matched transform of second point set P are to keep the objective function of following formula minimum:
R in above-mentioned formula is three-dimensional rotation matrix, and T is translation matrix.
The purpose of ICP algorithm is to find the transformation of spin matrix R and translation matrix T between target point set and reference point.
The search process of arest neighbors is the very big part of calculation amount in iterative closest point algorithm, when the quantity at cloud midpoint is very big
When, can generate very big operand by the way of force search causes registration process very slow.The embodiment of the present invention
Accelerate the search process of closest approach using the form of KD tree during corresponding point matching.KD tree be it is a kind of by data space according to
Particular form is split, and accelerates the search of arest neighbors by the segmentation building tree construction in this space.Pass through the side of KD number
Formula can be by the O (n of force search2) complexity be reduced to the complexity of O (nlogn), the speed of registration can be improved.
ICP algorithm processing in, calculate current point cloud between distance and when, due to the degree of reliability of each point, point
The degree of reliability between point correspondence is all different, therefore can use different weights to different points.On the one hand,
For the confidence level of each sampled point, the characteristic of data can be acquired according to laser sensor, utilizes laser beam and sampled point
Angle, the laser sensor of surface normal estimate the possible credibility of each sampled point at a distance from sampled point, generally may be used
To think that smaller with surface normal angle and in the use of sensor distance point is more accurate.On the other hand, it puts and puts it
Between corresponding relationship credibility, it is general using between two o'clock between normal vector angle or the similarity degree of local curvature come degree
Amount.Therefore, calculate the distance and when, adjusted the distance according to the weight of sampled point and be weighted summation, for example, the sampled point
Weight, with the angle of laser beam and sampled point surface normal negative correlation.By introducing the different weights of each point, can make
The result that must be registrated is more robust, is capable of handling some noises and accelerates convergence rate.
The Optimization Solution process of ICP algorithm can be by the way of least square or the mode of nonlinear optimization is completed, and leads to
The weighted sum of the distance between minimum all-pair is crossed to solve to obtain optimal transformation matrix (R and T), such as directlys adopt surprise
The mode that different value decomposes (SVD, Singular Value Decomposition) solve or is solved using quaternary number substep
Translation and rotation amount etc..In the point cloud data of different characteristics, Different treatments may have the effect of different.
After point cloud registering processing, a cloud noise reduction process can also be carried out.For example, Gaussian function is utilized in gaussian filtering
Still with the characteristic of Gaussian function after being fourier transformed, enabling the weight in specified domain is Gaussian Profile, thus making an uproar high frequency
Sound filters out.Specifically a certain data point and its each n data point in front and back are weighted and averaged, those are much larger than the point quilt of operating distance
It is processed into the endpoint of Zhou Ding, this helps to identify gap and endpoint.Since gaussian filtering is while filtering, can preferably keep
Data original appearance, thus be often used.In embodiments of the present invention, the Gaussian filter of PCL offer, basic principle are provided
To assume to obtain the result is that a Gaussian Profile, shape are determined by mean value and standard deviation, average distance critical field (by
Global distance average and variance definition) except point, noise can be defined as and can be got rid of from data set.
The specific implementation of above-mentioned steps 12:
In above-mentioned steps 11, using laser sensor scanning collection point cloud data, and preprocessed and registration and form
One preliminary more complete three-dimensional point cloud model of place.But in general, each of existing point cloud data model
Point generally all only stores position (D coordinates value) and the characteristic information (color information and laser intensity angle value) of sampled point, and
The topological relation between a little is not stored.I.e. put cloud in point be it is at random, spatial relationship between points it is unknown and exist
Noise can make the modeling in later period become difficult.Therefore, how collected point cloud data to be optimized and carry out modeling be
Important problem during entire three-dimensional modeling, essence are how to become " point " into the process of " body ".In this process
In, how be to solve by " point " becomes " face " first, this is related to the extraction and reconstruction of geometric primitive in a cloud.
In order to which the reconstruction of threedimensional model is better achieved, need to extract common basic geometric primitive from point cloud data
Indoor scene is described.The main geometric primitive of indoor scene includes the various combinations of plane, cylinder and plane, this hair
Bright embodiment extracts above-mentioned geometric primitive from point cloud data, so as to rebuild indoor scene.
The three-dimensional point cloud model of place obtained first to step 11 is split.
Specific form in kind in reality is diversified, but is modeled for indoor scene, always can be by its point
Solution is a series of geometric primitive, is indicated using characteristic surface, such as: plane, cylindrical surface.
According to common surface classification method, characteristic surface can be divided into three categories, respectively (plane is quadratic surface
The quadratic surface of degeneration), fillet surface and free form surface (nurbs surface).Quadratic surface include common spherical surface, cylindrical surface,
Circular conical surface, paraboloid etc..In actual scene, surface in kind usually can by or approximation be made of plane and quadratic surface.
In three dimensions, curved surface meets equation:
F (x, y, z)=c1x2+c2y2+c3z2+c4xy+c5yz+c6zx+c7x+c8y+c9z+c0=0 (2)
C thereini, i=0, Λ 9 is quadric parameter.It can be indicated by substituting into or solving different parameters
Different quadratic surfaces.In order to which the reconstruction of scene is better achieved, needs to extract common quadratic surface from cloud and (calculate two
The parameter value of secondary curved surface) scene is described as geometric primitive, by the curved surface to composition object each surface extract with
And the extraction of the constraint relationship, the precision of reconstruction model can be greatly improved.
Quadratic surface extraction is directly carried out for discrete point cloud data can be divided into following two step:
1) quadric identification is carried out to the point cloud data block of single type;
2) curved surface extraction is carried out using its geometric parameter equation for the different types of quadratic surface identified.
In embodiments of the present invention, curved surface identification can be carried out using RANSAC algorithm.RANSAC algorithm is in LiDAR point cloud
Data processing has a wide range of applications, this is determined by its algorithm characteristic: RANSAC has high breakdown point, works as sample
It still is able to obtain correct calculated result when this error is more than 50%, is a kind of efficient, steady random sampling algorithm.For side
Assistant's solution, has carried out simple introduction to the background of related of RANSAC algorithm below.
RANSAC algorithm principle are as follows: for including correct data (inliers or interior point) and abnormal data
The sample set data of (outliers or exterior point), the algorithm can solve parameter model to be asked by iterative calculation
Meet the correct model parameter and correct sample data under certain fiducial probability out, specific implementation is divided into following steps:
1) for sample set Data, there are the model M that smallest sample number is n, n is required for initialization M model parameter
Smallest sample number, and Data > n;
2) the subset S comprising n sample is randomly selected from Data, constitutes initial model M;
3) sample set and S with model M error less than threshold value t are extracted from complementary set SC=Data-S constitutes sample together
Consistent collection S*;
If 4) S*Number of samples >=given threshold value N, then it is assumed that this obtains correct model parameter, records this model
M is Maybe_Best_Model;
5) it repeats the above steps 1~4, in an iterative process, records best statistical model;
6) after reaching given the number of iterations K, if Maybe_Best_Model meets threshold value N, Maybe_Best_Model
For optimum Best_Model;Otherwise failure is calculated.
RANSAC algorithm be based on the random sampling probability of success solve optimum model parameter, can be used to fit Plane, sphere,
The geometrical models such as cylinder.In the parametric procedure for solving geometrical model, it is thus necessary to determine that 6 parameters of RANSAC algorithm, including need
Want preset 5 parameters: the pollution rate w of sample mother collection, the minimum sampling number n for meeting model M, successfully confidence of sampling are general
The refusal threshold value t and correct model smallest sample collection N of rate P, model, the maximum number of iterations of the 6th parameter can be 5 by front
Parameter is calculated.
In data acquisition, due to the characteristic of laser, certain wall bottoms in the presence of can cause strong reflection or
When the medium of refraction, a large amount of point clouds for being lower than level ground can be collected.The inverted image generated after laser reflection and refraction,
When the subsequent parameter for obtaining a single point cloud plane, need to calculate its corresponding several vertex, if retaining these inverted images will lead to
Mistake is calculated, therefore the embodiment of the present invention identifies inverted image and eliminates it from scene, to improve the accuracy of subsequent calculating.
Referring to figure 2., relative position of the embodiment of the present invention according to these inverted image point clouds and ground point cloud, a kind of deletion scheme provided
Are as follows:
1) simple computation that parameter is carried out to the basic geometric primitive in scene, acquires the point cloud where ground;
2) is to its normal direction of ground point cloud computing, and acquires the central point of ground point cloud, obtains StoreFront central point;
3) the basic geometric primitive of the point cloud representation of each non-horizontal directions is handled as follows:
3.1) by the normal direction of each point projection of the basic geometric primitive to ground central point;
3.2) the distance between point and ground central point after calculating projection;
3.3) when distance is less than given threshold value, then it is judged as inverted image point, and reject from original point cloud.
By taking Fig. 2 as an example, two points of some corresponding metope of metope point cloud, such as the P in Fig. 21And P2, in ground central point
The subpoint in direction be respectively P'1 and P'2, wherein P'2 is less than given threshold value at a distance from ground central point, therefore as falling
Shadow point is deleted.
In addition, approximately the same plane may excessively be divided due to an influence for cloud measurement error, registration error and segmentation threshold
Two or more different sub- planes are cut into, plane as shown in Figure 3 is two sub- planes of A and B by over-segmentation, and the present invention is real
Applying example can be by merging processing to similar plane, to obtain the set of final correct planar point cloud.Specifically, of the invention
Embodiment considers whether two planes meet following two condition in merging treatment:
1) the normal vector angle theta of two planes is less than preset threshold t_ θ, and Fig. 4 gives the one of two planar process vector angles
A example;
2) two planes to coordinate origin distance difference be less than preset threshold t_d, Fig. 5 give two planes to seat
Mark an example of the difference of the distance of origin.
The embodiment of the present invention is successively traversed as the planar point cloud to obtained by, calculate the mutual angle of two cloud planes with
The difference of above-mentioned distance can then find the planar point cloud of over-segmentation and be closed, to obtain larger and complete point Yun Ping
Face.
In addition, especially in building, door is a very common component part in reality scene.Especially in people
In the application scenarios such as flow point analysis, door needs to be identified as basic channel.The embodiment of the present invention was acquired in point cloud data
Cheng Zhong, if door is concordant with wall and is when closing, what is obtained will be point cloud data of the entirety together with wall, and
Such point cloud data is difficult to identify by cloud itself.Therefore, the embodiment of the present invention is in acquisition point cloud data
When, door is opened, in the point cloud data collected in this way, there are the metopes of door, then can show as in biggish wall point cloud
Lack the point cloud of a part of (door).Based on this, the extraction scheme that the embodiment of the present invention can design door opening is as follows:
1) according to the space of some metope plane and ground, the both direction of the metope plane is calculated;
2) according to four vertex of the two direction calculatings metope plane;
3) four vertex of the metope plane and directional structure vectorical structure one complete metope point cloud are utilized;
4) by subtracting the initial point cloud of the metope plane from complete metope point cloud, initial hollow sectors be can be obtained
Point cloud, i.e., mainly include door opening point cloud.
The specific implementation of above-mentioned steps 13:
Above, the embodiment of the present invention divides scene, has obtained many basic geometric primitives, each
A basic geometric primitive can be considered as a part of entire scene.It is fitted drafting here by basic geometric primitive,
Geometric primitive after obtained reconstruction is still a part in scene, then, in order to obtain whole model, needs to gained
Reconstruction after geometric primitive spliced, to complete the building of three-dimensional indoor scene.
The input of the reconstruction process of indoor threedimensional model is the independent cloud that segmentation obtains, and output is then using weight
The indoor threedimensional model that geometric primitive (predominantly plane) after building is composed, specifically need by plane fitting, ground and
The extraction and fitting of ceiling find principal direction, find metope shape information, plain splice, door opening rejecting and triangle gridding
The processing such as filtering, next introduces the specific implementation of above-mentioned processing respectively.
1) plane fitting is handled
In order to preferably rebuild model of place, need to be fitted the scene surface extracted to rebuild surface, and with
This precision to improve reconstruction model.
The process of resurfacing can be described as: known sample point set X={ x1,x2,x3,Λ,xnRelated to some samplings
Information (for example, noise level Δ, sample rate Θ etc.), determine face M', it made to approach a unknown face M.Curve reestablishing algorithm
Surface fitting, piecewise-linear reconstruction, the reconstruction based on physics etc. can be divided into.Surface fitting is used in embodiments of the present invention
Method, specifically approached using implicit surface or parametric surface (such as quadratic surface) or fitting data point set.It is a kind of
Basic fitting algorithm is least square fitting method.
Quadratic surface extraction belongs to data fitting problems, therefore relatively conventional least square method can be used also to ask
Solution.Least square method is usually to pass through the least residual quadratic sum of the objective function f (x, y, z) asked, it may be assumed that
To solve the problems, such as over-determined systems.
For different types of Quadratic Surface Equation, such as plane, cylindrical surface, linear least square and non-can be used respectively
Linear least square is handled.
2) extraction and fitting of ground and ceiling
Each of Model Reconstruction input cloud is the relatively independent and complete plane that segmentation obtains.It is final in order to obtain
The indoor model that is composed of plane primitive, need to carry out plane primitive fit operation to the point cloud of input, extract point cloud
Areal model and the interior point of cloud sheet section of ordering, that is, meet the interior point of plane equation, and the boundary of the plane primitive is found using interior point,
Construct quadrangle model as accurate as possible.
In result after model segmentation, the point cloud of metope was both contained, the point cloud of ground and ceiling is also contained.Due to
The shape of ground and ceiling depends primarily on the wall of surrounding, so ground and ceiling are abstracted into plane.It is indoor
In scene, can often there be wall and ground, ceiling are the prior informations being connected to, in order to use this information, the present invention is real
Ground, ceiling and common metope can be separated according to the direction of plane normal vector by applying example, individually retain ground and day
The information of card.
In the collection process of point cloud data, laser sensor and ground keep connecing subparallel state, therefore can be with method
The plane of vector straight up is possible ground and ceiling.Separation process is mainly one by one to the cut-point cloud sheet section of input
Be fitted, it is close with direction straight up to normal vector (as both angle be less than some preset threshold value) order cloud sheet section
It is added to possible ground ceiling plate point set.Handled it is all order after cloud sheet section, to obtained possible ground ceiling
The point of plate converges conjunction and seeks center, then thinks that the point higher than center is the point on ceiling, is ground lower than central point
Point.Then, ceiling plane and floor are fitted respectively using this two groups of points.
3) principal direction is found
The usual interior space has obvious square structure, and this square structure often has orthogonal three
Principal direction, one is perpendicular to ground, other two is orthogonal and in the horizontal direction.This can be used in the embodiment of the present invention
The rectangular prior information of kind, the model obtained come the fitting that standardizes, it would be possible to orthogonal plane rectification to most probable phase
Mutually on vertical direction.
The principal direction of vertical direction is defaulted as the normal orientation on the ground that fitting obtains.The mistake of horizontal direction searching principal direction
Journey it is a kind of realize can be the direction of wall plane is clustered, group number is maximum and mean level of the sea direction orthogonal two
A cluster centre, it is believed that be the principal direction of model.Orthogonal correction finally is carried out to three principal directions, guarantees three principal direction phases
It is mutually vertical.After obtaining three principal directions, traverse all planes, if close to if some principal direction actively by the plane rectification arrive this
In principal direction.It can guarantee that the vertical structure in model is always vertical in this way.
4) metope shape information is found
After finding the principal direction of model, for each plane, y-axis can regard as model in plane coordinate system
Vertical direction, x-axis direction can be obtained by y-axis and the multiplication cross of plane normal vector.It is all in this way to use when being fitted the plane
The interior point selected in cloud sheet section can project in the plane, and transform in the unilateral coordinate system.It is sought in the coordinate system of plane
It looks for and orders the length and width of cloud sheet section, respectively point cloud distribution upper farthest value in the x and y direction.It, can be with using the information of farthest point
The encirclement box (bounding box) for constructing this edge point cloud sheet section, since all interior points are all on a three-dimensional planar, so
Bounding box is a rectangle in fact, and the embodiment of the present invention retains its four vertex, to indicate the shape information of this wall.
5) plain splice
Due to the error being difficult to avoid that in data acquisition, the quadrangle of obtained metope is approximate always to have certain error, respectively
It is difficult to be directly connected into a complete model between block plane primitive, has crack etc. between different metopes.The present invention is implemented
Example is by judging whether two planes should connect, to connect the plane of the connection.
All metopes need first and ground, day flower stencil are spliced, if the boundary of the metope is from the ground or the variation of day flower stencil
Less, which is just actively connected on ground and day flower stencil by that, eliminates crack.Later plane need between any two attempt be
It is no to splice, when the vertex of two planes is very close to another plane, actively the boundary in two faces is adjusted
It is whole, so that two planes connect together.It can also guarantee to generate closed model as far as possible in this way.
6) door opening is rejected
Since some information such as door are important information in some applications, before needing to utilize in final model
The door opening information extracted carries out model perfect.The embodiment of the present invention assumes that there is connection on the bottom edge of door and ground, in this way
After detecting door opening in certain face wall, the embodiment of the present invention only needs the rectangle by the polygon of wall and door to be attached.Most
Simple wall model is rectangle, and there may be multiple door openings on a metope.Assuming that adding in the vertex sequence of wall before
Several doors are entered, if the vertex sequence of wall is { V at this time1,V2,…,VnAnd vertex be from the upper left corner arranged clockwise to lower-left
Angle, the vertex sequence of door are { Vlu,Vru,Vrb,VlbThe lower left corner is traversed clockwise from the upper left corner, the sequence of door is inserted in this way
Enter to front wall vertex sequence it is suitable place.
7) triangle gridding filters
Current computer is operated in progress graphic plotting generally using tri patch as basic element, the embodiment of the present invention
The plane primitive that above procedure fits be generally include 4 while or more while polygon, it is therefore desirable to carry out the triangulation network
Lattice are filtered to complete final drafting.
In practical applications, the embodiment of the present invention assumes that the interior space mainly using the metope of rule as basic element, is led to
It crosses and splices and combines into indoor model.But simultaneously because may have the cavities such as door and window on some metopes, in order to final
Model in performance go out equal elements, need to be indicated metope using polygon.In the expression of polygon, each geometry base
Member with the vertex representation of one group of sequence such as { V_1, V_2 ..., V_n }.It is drawn in drawing process if direct sequence inputs to vertex
Library processed, often will cause the effect of ghost image, therefore need to carry out a triangle gridding filtering, and the polygon of input is transformed into triangle
The form of grid.
By handling above, complete indoor threedimensional model is can be generated in the embodiment of the present invention.The embodiment of the present invention may be used also
To provide basic function based on visual post process (VTK, Visualization ToolKit), after designing and Implementing reconstruction
The visualization of indoor threedimensional model.
By the above detailed process, the embodiment of the present invention during reconstructing three-dimensional model, solves point cloud data indoors
The problem of registration calibration and indoor metal skin simplify indoor three-dimensional because of inverted image problem caused by reflection laser/radar
The modeling process of model, and improve the precision of modeling.
Further, after rebuilding indoor threedimensional model, the embodiment of the present invention can also design and Implement the algorithm of measurement,
To realize accurately measuring based on indoor threedimensional model.
For the indoor threedimensional model rebuild, a certain amount is carried out, is had for the understanding and cognition of model of place
Very great meaning.So the module of geometric dimension calculating is added in the embodiment of the present invention in a model, such as most basic
The calculating of the distance of point, the calculating of angle based on coordinate system, distance, height of column etc. from sluice gate to staircase.For
The calculating of these geometric dimensions is mainly realized using calculating of the point at a distance from point.The embodiment of the invention provides correlation-measuring instruments
The content of calculation includes:
A) three dimensional space coordinate is inquired
In the Visual Scene based on threedimensional model, most basic space querying includes that the three-dimensional coordinate of spatial point is looked into
It askes, it is the basis of other interactive operations and spatial analysis.The specific implementation of coordinate inquiry are as follows: put down the computer that user selects
The two-dimensional screen coordinates of some point on face instead solve as three dimensional space coordinate.The above process is an inverse process of perspective projection.
Mainly there are normal solution transformation and anti-solution to convert two ways, in which:
Normal solution transformation: selected with mouse by the two-dimensional screen coordinates of the coordinate transform of three-dimensional space point to computer
Target point is matched, to obtain the three-dimensional coordinate of target point.By the project objects in three-dimensional space to dimension computer screen
On, it can be realized by projective transformation matrix.
Anti- solution transformation: window coordinates are mapped to by object coordinates according to the inverse matrix of projection matrix, to obtain window mesh
The three-dimensional coordinate of punctuate.
B) space length inquiry and calculating
Mouse or touch operation can be used in user, arbitrarily chooses two different points on threedimensional model indoors, therebetween
Sum of the distance of the distance between the line and a series of intersection points of model, the available line of algorithm and model before use
A series of intersection points three-dimensional coordinate, the distance of space two o'clock can be calculated using following formula:
S=∑ DI, i+1 (4)
Wherein:
D in above-mentioned formulai,i+1Indicate point the distance between i and i+1, Xi, Yi, ZiRespectively indicate the three-dimensional coordinate of point i.
C) three-dimensional areal calculation:
For indoor threedimensional model, the three-dimensional dimension information of selected plane can also be directly inquired.Three-dimensional dimension information master
Refer to the length and width of the rectangle that metope is surrounded.Need to calculate projected area after section cutting, i.e., arbitrary polygon is in horizontal plane
On area, can specifically be calculated using Heron's formula, or projected area can be calculated using trapezoidal rule.
The indoor three-dimensional modeling method of the embodiment of the present invention is described in detail above.The embodiment of the present invention also provides
A kind of computer readable storage medium, is stored thereon with computer program, real when the computer program is executed by processor
The step in indoor three-dimensional modeling method in existing above method embodiment.
Based on above method, the embodiment of the invention also provides the systems for implementing the above method, please refer to Fig. 6, the present invention
Embodiment provides a kind of indoor 3 d modeling system 60, comprising:
Point cloud data acquisition module 61, for obtaining the point cloud data of the target interior space, and to the point cloud data into
Row registration, generates point cloud data model;
Point cloud data processing module 62, for extracting basic geometric primitive from the point cloud data model.Here, described
Basic geometric primitive includes plane and curved surface etc.;
Three-dimensional modeling module 63, for carrying out curve reestablishing to the basic geometric primitive, the geometry base after being rebuild
Member, and the geometric primitive after the reconstruction is spliced, obtain the threedimensional model of the target interior space.
Please refer to Fig. 7, another indoor 3 d modeling system 70 provided in an embodiment of the present invention, in addition to including in Fig. 6
Outside equal modules, further includes:
Measurement module 64, the inquiry operation for being directed to threedimensional model input for receiving user;Respond the inquiry behaviour
Make to calculate and export corresponding query result, the query result includes space coordinate, distance, area or volume.
Please refer to Fig. 8, another indoor 3 d modeling system 80 provided in an embodiment of the present invention, as shown in Figure 8, in which:
The point cloud data acquisition module 61 includes:
Registration unit 611, for using undistinguishable closest approach iteration ICP algorithm, from collected first frame point cloud number
According to beginning, the two frame point cloud datas adjacent to position are registrated and are merged, and finally obtain the point cloud data model being registrated.
Specifically, can in the point cloud data first frame point cloud data and the second frame point cloud data carry out registration and
Merge, has been registrated point cloud data, wherein the acquisition position phase of the first frame point cloud data and the second frame point cloud data
It is adjacent;To the remaining point cloud data in the point cloud data, following steps are executed frame by frame: being selected from the remaining point cloud data
The next frame point cloud data adjacent with the acquisition position for being registrated point cloud data is calculated as point cloud data subject to registration using ICP
Method has been registrated point cloud data and the point cloud data subject to registration is registrated and is merged to described, obtain it is updated it is described
It is registrated point cloud data;It has been registrated point cloud data according to described, has obtained the point cloud data model of the target interior space.
Wherein, the point cloud data processing module 62 includes:
Model cutting unit 621, for using the consistent RANSAC algorithm of random sampling, in the point cloud data model
Point cloud data carries out curved surface identification, identifies the type of curved surface;According to the type of curved surface, using corresponding the geometric parameter equation,
Extraction obtains the basic geometric primitive;
Inverted image deletes unit 622, for calculating the basic geometric primitive in the point cloud data model, definitely
The corresponding ground point cloud of facial plane;The normal vector of the ground point cloud is calculated, and determines the central point of the ground point cloud;To every
The basic geometric primitive of one non-horizontal directions calculates each point of the basic geometric primitive in the normal direction of the central point
Subpoint whether be less than preset threshold at a distance from the central point, and according to the distance, determine whether the point is inverted image
Point;Delete the inverted image point in the basic geometric primitive;
It is flat to calculate any two for traversing to the plane in the point cloud data model for plane combining unit 623
The distance difference of angle and preset coordinate origin between the normal vector in face to any two plane;It is less than in the angle
First pre-determined threshold and when the distance difference is less than the second pre-determined threshold, merges into a plane for any two plane;
Door opening extraction unit 624, for extracting metope plane from the point cloud data model, and it is flat from the metope
It is extracted in face there are the first metope plane of partial dot cloud missing, the metope plane is vertical with the floor;According to
The space of the floor and the first metope plane calculates the direction of the first metope plane;According to first metope
The direction of plane calculates four vertex of the first metope plane;Utilize four vertex of the first metope plane and side
To construction one completely refers to metope point cloud;By the reference metope point cloud and corresponding cloud of the first metope plane
Subtract each other, obtains the corresponding door opening point cloud of door opening plane.
Wherein, the three-dimensional modeling module 63 includes:
Plane fitting unit 631 carries out surface fitting for the point cloud data to the basic geometric primitive, obtains described
Geometric primitive after the corresponding reconstruction of basic geometric primitive;
3D Model Reconstruction unit 632, for the direction according to the normal vector of plane, from the geometric primitive after each reconstruction
Floor is isolated in point cloud and corresponding first point of ceiling plane is converged conjunction and the corresponding second point cloud of metope plane
Set;According to the height of point, the point in closing is converged by described first point and is divided to ceiling point set or ground point set, and divide
The other point in the ceiling point set and ground point set carries out plane fitting, obtains the day of the target interior space
Card plane and floor;Determine the target interior space the first principal direction in the vertical direction and in the horizontal direction
The second principal direction and third principal direction;According to corresponding cloud of first principal direction and each metope plane, institute is generated
State the metope of the target interior space;By the metope of the target interior space, respectively with the ceiling of the target interior space
Plane is mutually spliced with floor, and is inserted into the corresponding rectangle of door opening plane, obtains the geometric primitive of the target interior space
And triangle gridding filtering is carried out, generate the threedimensional model of the target interior space.
Further, in the embodiment of the present invention, the 3D Model Reconstruction unit 632 is also used to main according to described first
Corresponding cloud in direction and each metope plane, before the metope for generating the target interior space, further using described
First principal direction, the second principal direction and third principal direction, correct each plane.
Further, in the embodiment of the present invention, the 3D Model Reconstruction unit 632 is also used to basis and floor phase
Vertical direction obtains the first principal direction on vertical direction;The direction of all metope planes is clustered, is obtained multiple poly-
Wherein group's number maximum and orthogonal two cluster centres in direction are selected, by the side of two cluster centres in class center
To as the second principal direction and third principal direction in horizontal direction.
Referring to FIG. 9, the embodiment of the invention provides another hardware configurations of another indoor 3 d modeling system 900 to show
It is intended to, comprising: processor 901, network interface 902, memory 903, user interface 904 and bus interface, in which:
In embodiments of the present invention, the indoor 3 d modeling system 900 further include: storage on a memory 903 and can
The computer program run on processor 901 realizes following steps when computer program is by processor 901, execution:
The point cloud data of the target interior space is obtained, and the point cloud data is registrated, generates point cloud data model;
Basic geometric primitive is extracted from the point cloud data model;
Curve reestablishing is carried out to the basic geometric primitive, the geometric primitive after being rebuild, and to the reconstruction after
Geometric primitive is spliced, and the threedimensional model of the target interior space is obtained.
In Fig. 9, bus architecture may include the bus and bridge of any number of interconnection, specifically be represented by processor 901
One or more processors and the various circuits of memory that represent of memory 903 link together.Bus architecture can be with
Various other circuits of such as peripheral equipment, voltage-stablizer and management circuit or the like are linked together, these are all these
Well known to field, therefore, it will not be further described herein.Bus interface provides interface.Network interface 902 can be with
It is wired or wireless network card equipment, realizes transmission-receiving function of the data on network.For different user equipmenies, user interface
904 can also be and external the interface for needing equipment can be inscribed, and the equipment of connection includes but is not limited to keypad, display, raises
Sound device, microphone, control stick etc..
Processor 901, which is responsible for management bus architecture and common processing, memory 903, can store processor 901 and is holding
Used data when row operation.
Optionally, following steps be can also be achieved when computer program is executed by processor 901:
Using undistinguishable closest approach iteration ICP algorithm, to the first frame point cloud data and second in the point cloud data
Frame point cloud data is registrated and is merged, and point cloud data has been registrated, wherein the first frame point cloud data and the second frame point
The acquisition position of cloud data is adjacent;
To the remaining point cloud data in the point cloud data, following steps are executed frame by frame: from the remaining point cloud data
The next frame point cloud data adjacent with the acquisition position for being registrated point cloud data is selected, as point cloud data subject to registration, is used
The ICP algorithm has been registrated point cloud data and the point cloud data subject to registration is registrated and is merged to described, after obtaining update
Described be registrated point cloud data;
It has been registrated point cloud data according to described, has obtained the point cloud data model of the target interior space.
Using the consistent RANSAC algorithm of random sampling, curved surface knowledge is carried out to the point cloud data in the point cloud data model
Not, the type of curved surface is identified;
According to the type of curved surface, using corresponding the geometric parameter equation, extraction obtains the basic geometric primitive.
Optionally, following steps be can also be achieved when computer program is executed by processor 901:
Basic geometric primitive in the point cloud data model is calculated, determines the corresponding ground point of floor
Cloud;
The normal vector of the ground point cloud is calculated, and determines the central point of the ground point cloud;
To the basic geometric primitive of each non-horizontal directions, each point of the basic geometric primitive is calculated in described
Whether the subpoint in the normal direction of heart point is less than preset threshold at a distance from the central point, and according to the distance, and determining should
Whether point is inverted image point;
Delete the inverted image point in the basic geometric primitive.
Optionally, following steps be can also be achieved when computer program is executed by processor 901:
Plane in the point cloud data model is traversed, the angle between the normal vector of any two plane is calculated
And preset coordinate origin is to the distance difference of any two plane;
When the angle is less than the first pre-determined threshold and the distance difference is less than the second pre-determined threshold, by this any two
A plane merges into a plane.
Optionally, following steps be can also be achieved when computer program is executed by processor 901:
Metope plane is extracted from the point cloud data model, and extracts from the metope plane that there are partial dots
First metope plane of cloud missing, the metope plane are vertical with the floor;
According to the space of the floor and the first metope plane, the direction of the first metope plane is calculated;
According to the direction of the first metope plane, four vertex of the first metope plane are calculated;
Using four vertex of the first metope plane and direction, constructs one and completely refer to metope point cloud;
Subtract each other described with reference to metope point cloud with corresponding cloud of the first metope plane, it is corresponding to obtain door opening plane
Door opening point cloud.
Optionally, following steps be can also be achieved when computer program is executed by processor 901:
Surface fitting is carried out to the point cloud data of the basic geometric primitive, it is corresponding heavy to obtain the basic geometric primitive
Geometric primitive after building.
Optionally, following steps be can also be achieved when computer program is executed by processor 901:
According to the direction of the normal vector of plane, isolated from the point cloud of the geometric primitive after each reconstruction floor and
Corresponding first point of ceiling plane is converged conjunction and the corresponding second point of metope plane converges conjunction;
According to the height of point, the point in closing is converged by described first point and is divided to ceiling point set or ground point set,
And is carried out by plane fitting, obtains the target interior space for the point in the ceiling point set and ground point set respectively
Ceiling plane and floor;
Determine the target interior space the first principal direction in the vertical direction and the second main side in the horizontal direction
To with third principal direction;
According to corresponding cloud of first principal direction and each metope plane, the wall of the target interior space is generated
Face;
By the metope of the target interior space, respectively with the ceiling plane and floor of the target interior space
Mutually splice, and is inserted into the corresponding rectangle of door opening plane, obtains the geometric primitive of the target interior space and carries out triangle gridding
Filtering, generates the threedimensional model of the target interior space.
Optionally, following steps be can also be achieved when computer program is executed by processor 901:
According to the direction perpendicular with floor, the first principal direction on vertical direction is obtained;
The direction of all metope planes is clustered, multiple cluster centres are obtained, select wherein group's number it is maximum and
Orthogonal two cluster centres in direction, by the direction of two cluster centres, as the second principal direction in horizontal direction
With third principal direction.
Optionally, following steps be can also be achieved when computer program is executed by processor 901:
According to corresponding cloud of first principal direction and each metope plane, the target interior space is generated
Before metope, first principal direction, the second principal direction and third principal direction are further utilized, each plane is corrected.
Optionally, following steps be can also be achieved when computer program is executed by processor 901:
Receive the inquiry operation that user is directed to threedimensional model input;
Respond the inquiry operation and calculate and export corresponding query result, the query result include space coordinate, away from
From, area or volume.
Those of ordinary skill in the art may be aware that list described in conjunction with the examples disclosed in the embodiments of the present disclosure
Member and algorithm steps can be realized with the combination of electronic hardware or computer software and electronic hardware.These functions are actually
It is implemented in hardware or software, the specific application and design constraint depending on technical solution.Professional technician
Each specific application can be used different methods to achieve the described function, but this realization is it is not considered that exceed
The scope of the present invention.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description,
The specific work process of device and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
In embodiment provided herein, it should be understood that disclosed device and method can pass through others
Mode is realized.For example, the apparatus embodiments described above are merely exemplary, for example, the division of the unit, only
A kind of logical function partition, there may be another division manner in actual implementation, for example, multiple units or components can combine or
Person is desirably integrated into another system, or some features can be ignored or not executed.Another point, shown or discussed is mutual
Between coupling, direct-coupling or communication connection can be through some interfaces, the INDIRECT COUPLING or communication link of device or unit
It connects, can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.Some or all of unit therein can be selected to realize the embodiment of the present invention according to the actual needs
Purpose.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.
It, can be with if the function is realized in the form of SFU software functional unit and when sold or used as an independent product
It is stored in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially in other words
The part of the part that contributes to existing technology or the technical solution can be embodied in the form of software products, the meter
Calculation machine software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be a
People's computer, server or network equipment etc.) it performs all or part of the steps of the method described in the various embodiments of the present invention.
And storage medium above-mentioned includes: that USB flash disk, mobile hard disk, ROM, RAM, magnetic or disk etc. are various can store program code
Medium.
The indoor three-dimensional modeling method and system of the embodiment of the present invention are described in detail above.As can be seen that this
The indoor three-dimensional modeling method and system that inventive embodiments provide, are registrated, nothing using the adjacent point cloud data of acquisition position
Feature point for calibration in scene indoors in advance is needed, data acquisition is carried out according to closed loop path without acquisition equipment, simplifies number
According to collection process, the application range of indoor three-dimensional modeling method is expanded.In addition, the embodiment of the present invention also passes through at inverted image deletion
Reason, the laser reflection deleted in laser point cloud data acquisition influence, and improve the precision of reconstruction model.In addition, the present invention is real
The indoor threedimensional model that example is obtained based on reconstruction is applied, may be implemented to accurately measure real indoor environment.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can easily think of the change or the replacement, and should all contain
Lid is within protection scope of the present invention.Therefore, protection scope of the present invention should be subject to the protection scope in claims.
Claims (24)
1. a kind of interior three-dimensional modeling method characterized by comprising
The point cloud data of the target interior space is obtained, and the point cloud data is registrated, generates point cloud data model;
Basic geometric primitive is extracted from point cloud data model;
Curve reestablishing is carried out to basic geometric primitive, the geometric primitive after being rebuild, and the geometric primitive after reconstruction is carried out
Splicing obtains the threedimensional model of the target interior space.
2. the method as described in claim 1, which is characterized in that be registrated to the point cloud data, comprising:
Using undistinguishable closest approach iteration ICP algorithm, since collected first frame point cloud data, adjacent to position two
Frame point cloud data is registrated and is merged, and finally obtains the point cloud data model being registrated.
3. the method as described in claim 1, which is characterized in that basic geometric primitive is extracted from the point cloud data model,
Include:
Using the consistent RANSAC algorithm of random sampling, curved surface identification is carried out to the point cloud data in the point cloud data model, is known
Not Chu curved surface type;
According to the type of curved surface, using corresponding the geometric parameter equation, extraction obtains the basic geometric primitive.
4. method as claimed in claim 3, which is characterized in that basic geometric primitive is extracted from the point cloud data model,
Further include:
Basic geometric primitive in the point cloud data model is calculated, determines the corresponding ground point cloud of floor;
The normal vector of the ground point cloud is calculated, and determines the central point of the ground point cloud;
To the basic geometric primitive of each non-horizontal directions, each point of the basic geometric primitive is calculated in the central point
Normal direction on subpoint whether be less than preset threshold at a distance from the central point, and according to the distance, determine that the point is
No is inverted image point;
Delete the inverted image point in the basic geometric primitive.
5. method as claimed in claim 4, which is characterized in that basic geometric primitive is extracted from the point cloud data model,
Further include:
Plane in the point cloud data model is traversed, calculate any two plane normal vector between angle and
Distance difference of the preset coordinate origin to any two plane;
When the angle is less than the first pre-determined threshold and the distance difference is less than the second pre-determined threshold, which is put down
A plane is merged into face.
6. method as claimed in claim 5, which is characterized in that basic geometric primitive is extracted from the point cloud data model,
Further include:
Metope plane is extracted from the point cloud data model, and extracts from the metope plane that there are partial dot clouds to lack
The the first metope plane lost, the metope plane are vertical with the floor;
According to the space of the floor and the first metope plane, the direction of the first metope plane is calculated;
According to the direction of the first metope plane, four vertex of the first metope plane are calculated;
Using four vertex of the first metope plane and direction, constructs one and completely refer to metope point cloud;
Subtract each other described with reference to metope point cloud with corresponding cloud of the first metope plane, obtains the corresponding door opening of door opening plane
Point cloud.
7. the method as described in claim 1, which is characterized in that carry out curve reestablishing to the basic geometric primitive, comprising:
Surface fitting is carried out to the point cloud data of the basic geometric primitive, after obtaining the corresponding reconstruction of the basic geometric primitive
Geometric primitive.
8. method as described in any one of claim 1 to 7, which is characterized in that spell the geometric primitive after the reconstruction
The step of connecing, obtaining the threedimensional model of the target interior space, comprising:
According to the direction of the normal vector of plane, floor and smallpox are isolated from the point cloud of the geometric primitive after each reconstruction
Corresponding first point of plate plane is converged conjunction and the corresponding second point of metope plane converges conjunction;
According to the height of point, the point in closing is converged by described first point and is divided to ceiling point set or ground point set, and divide
The other point in the ceiling point set and ground point set carries out plane fitting, obtains the day of the target interior space
Card plane and floor;
Determine the target interior space the first principal direction in the vertical direction and the second principal direction in the horizontal direction and
Third principal direction;
According to corresponding cloud of first principal direction and each metope plane, the metope of the target interior space is generated;
By the metope of the target interior space, mutually spelled with the ceiling plane of the target interior space and floor respectively
It connects, and is inserted into the corresponding rectangle of door opening plane, obtain the geometric primitive of the target interior space and carry out triangle gridding filtering,
Generate the threedimensional model of the target interior space.
9. method according to claim 8, which is characterized in that determine first of the target interior space in the vertical direction
Principal direction and the second principal direction in the horizontal direction and third principal direction, comprising:
According to the direction perpendicular with floor, the first principal direction on vertical direction is obtained;
The direction of all metope planes is clustered, multiple cluster centres are obtained, selects wherein group's number maximum and direction
Orthogonal two cluster centres, by the direction of two cluster centres, as the second principal direction and in horizontal direction
Three principal directions.
10. method according to claim 8, which is characterized in that according to first principal direction and each metope plane
Corresponding cloud before the metope for generating the target interior space, further utilizes first principal direction, the second principal direction
With third principal direction, each plane is corrected.
11. the method as described in claim 1, which is characterized in that further include:
Receive the inquiry operation that user is directed to threedimensional model input;
It responds the inquiry operation and calculates and export corresponding query result, the query result includes space coordinate, distance, face
Long-pending or volume.
12. a kind of interior 3 d modeling system characterized by comprising
Point cloud data acquisition module is registrated for obtaining the point cloud data of the target interior space, and to the point cloud data,
Generate point cloud data model;
Point cloud data processing module, for extracting basic geometric primitive from the point cloud data model;
Three-dimensional modeling module, for carrying out curve reestablishing to the basic geometric primitive, the geometric primitive after being rebuild, and it is right
Geometric primitive is spliced after reconstruction, obtains the threedimensional model of the target interior space.
13. interior 3 d modeling system as claimed in claim 12, which is characterized in that the point cloud data acquisition module packet
It includes:
Registration unit, for using undistinguishable closest approach iteration ICP algorithm, since collected first frame point cloud data,
The two frame point cloud datas adjacent to position are registrated and are merged, and finally obtain the point cloud data model being registrated.
14. interior 3 d modeling system as claimed in claim 12, which is characterized in that the point cloud data processing module packet
It includes:
Model cutting unit, for using the consistent RANSAC algorithm of random sampling, to the point cloud number in the point cloud data model
According to curved surface identification is carried out, the type of curved surface is identified;It is extracted according to the type of curved surface using corresponding the geometric parameter equation
To the basic geometric primitive.
15. interior 3 d modeling system as claimed in claim 14, which is characterized in that the point cloud data processing module also wraps
It includes:
Inverted image deletes unit and determines floor for calculating the basic geometric primitive in the point cloud data model
Corresponding ground point cloud;The normal vector of the ground point cloud is calculated, and determines the central point of the ground point cloud;It is non-to each
The basic geometric primitive of horizontal direction calculates each projection of point in the normal direction of the central point of the basic geometric primitive
Whether point is less than preset threshold at a distance from the central point, and according to the distance, determines whether the point is inverted image point;It deletes
Inverted image point in the basic geometric primitive.
16. interior 3 d modeling system as claimed in claim 15, which is characterized in that the point cloud data processing module also wraps
It includes:
Plane combining unit calculates the method for any two plane for traversing to the plane in the point cloud data model
The distance difference of angle and preset coordinate origin between vector to any two plane;It is pre- less than first in the angle
When gating limits and the distance difference is less than the second pre-determined threshold, which is merged into a plane.
17. interior 3 d modeling system as claimed in claim 16, which is characterized in that the point cloud data processing module also wraps
It includes:
Door opening extraction unit is mentioned for extracting metope plane from the point cloud data model, and from the metope plane
It takes out there are the first metope plane that partial dot cloud lacks, the metope plane is vertical with the floor;According to describedly
The space of facial plane and the first metope plane calculates the direction of the first metope plane;According to the first metope plane
Direction calculates four vertex of the first metope plane;Utilize four vertex of the first metope plane and direction, construction
One completely refers to metope point cloud;Subtract each other described with reference to metope point cloud with corresponding cloud of the first metope plane, obtains
To the corresponding door opening point cloud of door opening plane.
18. interior 3 d modeling system as claimed in claim 12, which is characterized in that the three-dimensional modeling module includes:
Plane fitting unit carries out surface fitting for the point cloud data to the basic geometric primitive, obtains described substantially several
Geometric primitive after what corresponding reconstruction of primitive.
19. such as the described in any item indoor 3 d modeling systems of claim 12 to 18, which is characterized in that the three-dimensional modeling mould
Block further include:
3D Model Reconstruction unit, for the direction according to the normal vector of plane, from the point cloud of the geometric primitive after each reconstruction
It isolates floor and corresponding first point of ceiling plane is converged conjunction and the corresponding second point of metope plane converges conjunction;Root
The height at strong point converges the point in closing for described first point and is divided to ceiling point set or ground point set, and respectively to institute
The point in ceiling point set and ground point set is stated, plane fitting is carried out, the ceiling for obtaining the target interior space is flat
Face and floor;Determine the first principal direction and in the horizontal direction second of the target interior space in the vertical direction
Principal direction and third principal direction;According to corresponding cloud of first principal direction and each metope plane, the target is generated
The metope of the interior space;By the metope of the target interior space, respectively with the ceiling plane of the target interior space and
Floor mutually splices, and is inserted into the corresponding rectangle of door opening plane, obtains the geometric primitive of the target interior space and progress
Triangle gridding filtering, generates the threedimensional model of the target interior space.
20. interior 3 d modeling system as claimed in claim 19, which is characterized in that the 3D Model Reconstruction unit is also used
In the metope for according to corresponding cloud of first principal direction and each metope plane, generating the target interior space it
Before, first principal direction, the second principal direction and third principal direction are further utilized, each plane is corrected.
21. interior 3 d modeling system as claimed in claim 19, which is characterized in that the 3D Model Reconstruction unit is also used
According to the direction perpendicular with floor, the first principal direction on vertical direction is obtained;To the direction of all metope planes
It is clustered, obtains multiple cluster centres, select wherein group's number maximum and orthogonal two cluster centres in direction, it will
The direction of two cluster centres, as the second principal direction and third principal direction in horizontal direction.
22. interior 3 d modeling system as claimed in claim 12, which is characterized in that further include:
Measurement module, the inquiry operation for being directed to threedimensional model input for receiving user;The inquiry operation is responded to calculate
And corresponding query result is exported, the query result includes space coordinate, distance, area or volume.
23. a kind of interior 3 d modeling system characterized by comprising memory, processor and storage are on a memory and can
The computer program run on a processor, when the computer program is executed by the processor realize as claim 1 to
The step of indoor three-dimensional modeling method described in any one of 11.
24. a kind of computer readable storage medium, which is characterized in that be stored with computer on the computer readable storage medium
Program realizes that the indoor three-dimensional as described in any one of claims 1 to 11 is built when the computer program is executed by processor
The step of modular system.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810321028.3A CN110363849A (en) | 2018-04-11 | 2018-04-11 | A kind of interior three-dimensional modeling method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810321028.3A CN110363849A (en) | 2018-04-11 | 2018-04-11 | A kind of interior three-dimensional modeling method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110363849A true CN110363849A (en) | 2019-10-22 |
Family
ID=68214161
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810321028.3A Pending CN110363849A (en) | 2018-04-11 | 2018-04-11 | A kind of interior three-dimensional modeling method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110363849A (en) |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110827339A (en) * | 2019-11-05 | 2020-02-21 | 北京深测科技有限公司 | Method for extracting target point cloud |
CN110969701A (en) * | 2019-11-29 | 2020-04-07 | 司空定制家居科技有限公司 | Method and device for modeling house and storage medium |
CN111125585A (en) * | 2019-12-24 | 2020-05-08 | 武汉珞珈灵智科技有限公司 | Visualization method and system for Web-end three-dimensional model |
CN111127655A (en) * | 2019-12-18 | 2020-05-08 | 北京城市网邻信息技术有限公司 | House layout drawing construction method and device, and storage medium |
CN111583391A (en) * | 2020-04-29 | 2020-08-25 | 北京深测科技有限公司 | Object three-dimensional reconstruction method and system |
CN111710040A (en) * | 2020-06-03 | 2020-09-25 | 纵目科技(上海)股份有限公司 | Method, system, terminal and storage medium for constructing high-precision map |
CN112045654A (en) * | 2020-08-11 | 2020-12-08 | 上海有个机器人有限公司 | Detection method and device for unmanned closed space and robot |
CN112417579A (en) * | 2021-01-25 | 2021-02-26 | 深圳大学 | Semantic-constrained planar primitive topological relation rule detection and recovery method |
CN112528728A (en) * | 2020-10-16 | 2021-03-19 | 深圳市银星智能科技股份有限公司 | Image processing method and device for visual navigation and mobile robot |
CN112540593A (en) * | 2019-11-22 | 2021-03-23 | 百度(美国)有限责任公司 | Method and apparatus for registering point clouds for autonomous vehicles |
CN112686934A (en) * | 2020-12-29 | 2021-04-20 | 广州广电研究院有限公司 | Point cloud data registration method, device, equipment and medium |
CN112966327A (en) * | 2021-03-17 | 2021-06-15 | 清华大学 | Three-dimensional indoor scene generation method and system based on spatial incidence relation |
CN113220018A (en) * | 2021-04-23 | 2021-08-06 | 上海发电设备成套设计研究院有限责任公司 | Unmanned aerial vehicle path planning method and device, storage medium and electronic equipment |
CN114140586A (en) * | 2022-01-29 | 2022-03-04 | 苏州工业园区测绘地理信息有限公司 | Indoor space-oriented three-dimensional modeling method and device and storage medium |
CN114419278A (en) * | 2022-01-19 | 2022-04-29 | 厦门大学 | Indoor three-dimensional color grid model generation method and system |
CN116453456A (en) * | 2023-06-14 | 2023-07-18 | 北京七维视觉传媒科技有限公司 | LED screen calibration method and device, electronic equipment and storage medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110304628A1 (en) * | 2010-06-10 | 2011-12-15 | Autodesk, Inc. | Pipe reconstruction from unorganized point cloud data |
US20110304619A1 (en) * | 2010-06-10 | 2011-12-15 | Autodesk, Inc. | Primitive quadric surface extraction from unorganized point cloud data |
CN102608620A (en) * | 2012-03-12 | 2012-07-25 | 北京北科安地科技发展有限公司 | Laser scanning point cloud vegetation filtering method on basis of reflection strength and terrain |
US20140192050A1 (en) * | 2012-10-05 | 2014-07-10 | University Of Southern California | Three-dimensional point processing and model generation |
CN105702151A (en) * | 2016-03-31 | 2016-06-22 | 百度在线网络技术(北京)有限公司 | Indoor map constructing method and device |
CN105844064A (en) * | 2016-05-23 | 2016-08-10 | 厦门亿力吉奥信息科技有限公司 | Three-dimensional transformer station semi-automatic reconstruction method based on laser point cloud data |
-
2018
- 2018-04-11 CN CN201810321028.3A patent/CN110363849A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110304628A1 (en) * | 2010-06-10 | 2011-12-15 | Autodesk, Inc. | Pipe reconstruction from unorganized point cloud data |
US20110304619A1 (en) * | 2010-06-10 | 2011-12-15 | Autodesk, Inc. | Primitive quadric surface extraction from unorganized point cloud data |
CN102608620A (en) * | 2012-03-12 | 2012-07-25 | 北京北科安地科技发展有限公司 | Laser scanning point cloud vegetation filtering method on basis of reflection strength and terrain |
US20140192050A1 (en) * | 2012-10-05 | 2014-07-10 | University Of Southern California | Three-dimensional point processing and model generation |
CN105702151A (en) * | 2016-03-31 | 2016-06-22 | 百度在线网络技术(北京)有限公司 | Indoor map constructing method and device |
CN105844064A (en) * | 2016-05-23 | 2016-08-10 | 厦门亿力吉奥信息科技有限公司 | Three-dimensional transformer station semi-automatic reconstruction method based on laser point cloud data |
Cited By (27)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110827339B (en) * | 2019-11-05 | 2022-08-26 | 北京深测科技有限公司 | Method for extracting target point cloud |
CN110827339A (en) * | 2019-11-05 | 2020-02-21 | 北京深测科技有限公司 | Method for extracting target point cloud |
CN112540593A (en) * | 2019-11-22 | 2021-03-23 | 百度(美国)有限责任公司 | Method and apparatus for registering point clouds for autonomous vehicles |
CN112540593B (en) * | 2019-11-22 | 2023-11-17 | 百度(美国)有限责任公司 | Method and device for registering a point cloud for an autonomous vehicle |
CN110969701A (en) * | 2019-11-29 | 2020-04-07 | 司空定制家居科技有限公司 | Method and device for modeling house and storage medium |
CN110969701B (en) * | 2019-11-29 | 2023-12-29 | 司空定制家居科技有限公司 | Method, device and storage medium for modeling house |
CN111127655A (en) * | 2019-12-18 | 2020-05-08 | 北京城市网邻信息技术有限公司 | House layout drawing construction method and device, and storage medium |
CN111127655B (en) * | 2019-12-18 | 2021-10-12 | 北京城市网邻信息技术有限公司 | House layout drawing construction method and device, and storage medium |
CN111125585A (en) * | 2019-12-24 | 2020-05-08 | 武汉珞珈灵智科技有限公司 | Visualization method and system for Web-end three-dimensional model |
CN111583391A (en) * | 2020-04-29 | 2020-08-25 | 北京深测科技有限公司 | Object three-dimensional reconstruction method and system |
CN111710040B (en) * | 2020-06-03 | 2024-04-09 | 纵目科技(上海)股份有限公司 | High-precision map construction method, system, terminal and storage medium |
CN111710040A (en) * | 2020-06-03 | 2020-09-25 | 纵目科技(上海)股份有限公司 | Method, system, terminal and storage medium for constructing high-precision map |
CN112045654B (en) * | 2020-08-11 | 2022-01-28 | 上海有个机器人有限公司 | Detection method and device for unmanned closed space and robot |
CN112045654A (en) * | 2020-08-11 | 2020-12-08 | 上海有个机器人有限公司 | Detection method and device for unmanned closed space and robot |
CN112528728A (en) * | 2020-10-16 | 2021-03-19 | 深圳市银星智能科技股份有限公司 | Image processing method and device for visual navigation and mobile robot |
CN112528728B (en) * | 2020-10-16 | 2024-03-29 | 深圳银星智能集团股份有限公司 | Image processing method and device for visual navigation and mobile robot |
CN112686934A (en) * | 2020-12-29 | 2021-04-20 | 广州广电研究院有限公司 | Point cloud data registration method, device, equipment and medium |
CN112417579B (en) * | 2021-01-25 | 2021-05-18 | 深圳大学 | Semantic-constrained planar primitive topological relation rule detection and recovery method |
CN112417579A (en) * | 2021-01-25 | 2021-02-26 | 深圳大学 | Semantic-constrained planar primitive topological relation rule detection and recovery method |
CN112966327A (en) * | 2021-03-17 | 2021-06-15 | 清华大学 | Three-dimensional indoor scene generation method and system based on spatial incidence relation |
CN112966327B (en) * | 2021-03-17 | 2022-09-16 | 清华大学 | Three-dimensional indoor scene generation method and system based on spatial incidence relation |
CN113220018A (en) * | 2021-04-23 | 2021-08-06 | 上海发电设备成套设计研究院有限责任公司 | Unmanned aerial vehicle path planning method and device, storage medium and electronic equipment |
CN114419278A (en) * | 2022-01-19 | 2022-04-29 | 厦门大学 | Indoor three-dimensional color grid model generation method and system |
CN114140586B (en) * | 2022-01-29 | 2022-05-17 | 苏州工业园区测绘地理信息有限公司 | Three-dimensional modeling method and device for indoor space and storage medium |
CN114140586A (en) * | 2022-01-29 | 2022-03-04 | 苏州工业园区测绘地理信息有限公司 | Indoor space-oriented three-dimensional modeling method and device and storage medium |
CN116453456A (en) * | 2023-06-14 | 2023-07-18 | 北京七维视觉传媒科技有限公司 | LED screen calibration method and device, electronic equipment and storage medium |
CN116453456B (en) * | 2023-06-14 | 2023-08-18 | 北京七维视觉传媒科技有限公司 | LED screen calibration method and device, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110363849A (en) | A kind of interior three-dimensional modeling method and system | |
Wang et al. | Lidar point clouds to 3-D urban models $: $ A review | |
Berger et al. | State of the art in surface reconstruction from point clouds | |
Zhang et al. | Online structure analysis for real-time indoor scene reconstruction | |
Murali et al. | Indoor Scan2BIM: Building information models of house interiors | |
CN110009727A (en) | A kind of indoor threedimensional model automatic reconfiguration method and system with structure semantics | |
EP1694821B1 (en) | Probable reconstruction of surfaces in occluded regions by computed symmetry | |
Ji et al. | A novel simplification method for 3D geometric point cloud based on the importance of point | |
Xu et al. | Reconstruction of scaffolds from a photogrammetric point cloud of construction sites using a novel 3D local feature descriptor | |
CN109544677A (en) | Indoor scene main structure method for reconstructing and system based on depth image key frame | |
Karantzalos et al. | Large-scale building reconstruction through information fusion and 3-d priors | |
CN109544612A (en) | Point cloud registration method based on the description of characteristic point geometric jacquard patterning unit surface | |
Wei et al. | Automatic coarse registration of point clouds using plane contour shape descriptor and topological graph voting | |
Shufelt | Geometric constraints for object detection and delineation | |
Liu et al. | Coarse registration of point clouds with low overlap rate on feature regions | |
Guo et al. | Line-based 3d building abstraction and polygonal surface reconstruction from images | |
Zhao et al. | Completing point clouds using structural constraints for large-scale points absence in 3D building reconstruction | |
Palma et al. | Detection of geometric temporal changes in point clouds | |
Álvarez et al. | Junction assisted 3d pose retrieval of untextured 3d models in monocular images | |
Seo | Model-based automatic building extraction from lidar and aerial imagery | |
Palma et al. | Scalable non-rigid registration for multi-view stereo data | |
Huang et al. | From bim to pointcloud: automatic generation of labeled indoor pointcloud | |
Xiong | Reconstructing and correcting 3d building models using roof topology graphs | |
Favorskaya et al. | Realistic tree modelling | |
Gkeli et al. | Automatic 3D reconstruction of buildings roof tops in densely urbanized areas |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20191022 |