CN117315184A - Method for automatically generating LOD1.3 model of building - Google Patents
Method for automatically generating LOD1.3 model of building Download PDFInfo
- Publication number
- CN117315184A CN117315184A CN202311094611.2A CN202311094611A CN117315184A CN 117315184 A CN117315184 A CN 117315184A CN 202311094611 A CN202311094611 A CN 202311094611A CN 117315184 A CN117315184 A CN 117315184A
- Authority
- CN
- China
- Prior art keywords
- building
- point
- point cloud
- component
- top surface
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 230000011218 segmentation Effects 0.000 claims abstract description 41
- 238000005457 optimization Methods 0.000 claims abstract description 15
- 238000012512 characterization method Methods 0.000 claims abstract description 7
- 238000003062 neural network model Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 5
- 238000000638 solvent extraction Methods 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 3
- 230000001629 suppression Effects 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 238000005530 etching Methods 0.000 claims description 2
- 230000010354 integration Effects 0.000 claims description 2
- 238000007621 cluster analysis Methods 0.000 abstract 1
- 238000010606 normalization Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/10—Constructive solid geometry [CSG] using solid primitives, e.g. cylinders, cubes
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
- G06V10/763—Non-hierarchical techniques, e.g. based on statistics of modelling distributions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Geometry (AREA)
- Computer Graphics (AREA)
- Probability & Statistics with Applications (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention provides a method for automatically generating a LOD1.3 model of a building, which adopts point cloud instance segmentation to directly separate the point cloud of a single building, and then fitting and post-treatment optimization can automatically generate DLG of the single building; in addition, the method adopts cluster analysis to divide the top surface component point set, uses Euclidean distance clustering algorithm to directly perform cluster fitting on the top surface component, adopts integrated optimization algorithm to optimize geometry, can generate top component characterization through post-treatment optimization, and can automatically generate LOD1.3 model by combining according to the threshold requirement of LOD 1.3.
Description
Technical Field
The invention relates to the technical field of modeling, in particular to a method for automatically generating a LOD1.3 model of a building.
Background
The current main modeling scheme is: after data are collected, TDOM (true projection image) is generated by adopting three-dimensional calculation, single building DLG (digital line drawing) is drawn manually, a building LOD1.0 model is generated by combining a point cloud model, and the LOD1.3 model is generated by manually adjusting the height of a local component in 3D software.
Problems of prior art solutions (problems solved by this patent):
1. drawing the DLG of the single building needs to be conducted in an auxiliary mode of deriving various data, and has the advantages of large manual workload and quite low efficiency.
2. After LOD1.0 is modeled, the height of the local component is adjusted in 3D editing software, and 3D geometry needs to be edited again, because the influence of factors such as position, angle and the like is very time-consuming and labor-consuming.
Disclosure of Invention
In view of the above, the present invention proposes a method of automatically generating a building LOD1.3 model that overcomes or at least partially solves the above-mentioned problems.
The invention provides a method for automatically generating a LOD1.3 model of a building,
s1, acquiring point cloud original data by adopting an unmanned plane or an unmanned vehicle carrying a multi-view camera or a laser radar;
s2, generating a color dense point cloud based on the original point cloud data;
s3, preprocessing the colored dense point cloud, and carrying out building point cloud instance segmentation to obtain an instance segmentation result of the building point cloud;
s4, extracting independent building point clouds according to example segmentation results of the building point clouds, and performing digital line drawing map DLG fitting to obtain digital line drawing map DLG characterization of coarse granularity of the building;
s5, screening according to independent building point clouds to obtain a point cloud integration set of a building top surface component, and fitting the top surface component to obtain a projection geometric representation of the building top surface component;
s6, combining the DLG representation of the coarse-grained digital line drawing map of the building and the projection geometric representation of the top surface component of the building to generate a multi-detail-level LOD1.3 model of the building.
Optionally, preprocessing the color dense point cloud in step S3 includes:
separating point clouds of different planes in the color dense point clouds by adopting a noise density clustering algorithm, and then sampling each category in an equal ratio mode to extract the dense point clouds into sparse point clouds;
partitioning the sparse point cloud on an [ x, y ] plane according to a coordinate partitioning grid, independently storing each block of point cloud as a file, and interpolating or extracting the number of points of each block of point cloud; wherein the intensity of the interpolation point is set to 0, and the point intensity i of the other normal point clouds is 1.
Optionally, in step S3, performing building point cloud instance segmentation, and obtaining an instance segmentation result of the building point cloud includes:
for the preprocessed sparse point cloud, marking part of the point cloud as a data set;
performing learning training by using an example segmentation network based on PointNet++, and performing recognition reasoning on unlabeled point clouds by using a trained neural network model; in the identification process, a nine-grid overlapping sampling mode is adopted, grid blocks of 3x3 around each grid block are combined in sequence to form point cloud blocks, then a neural network model is loaded to conduct instance segmentation, the non-maximum suppression of the [ x, y ] plane projection area is conducted on the segmentation result, the repeated identification of the building is combined, the damaged point cloud identification result is removed, and the instance segmentation result of the building point cloud is generated.
Optionally, step S4 extracts an independent building point cloud according to an example segmentation result of the building point cloud, and performs DLG fitting to obtain DLG characterization of the coarse-grained digital line map of the building, including:
projecting independent building point clouds to the ground according to example segmentation results of the building point clouds to form a ground plane point set of the building;
and expanding the points in the ground point set into a plane according to resolution, and combining the planes of all the point sets to fit into a polygon, wherein the polygon is the digital line drawing map DLG representation of the coarse granularity of the building.
Optionally, step S4 further includes:
for the digital line drawing map DLG representation of the obtained coarse granularity of the building, one line segment is adopted to replace a plurality of continuous line segments which are originally close to collineation in a threshold range, and polygon dense vertexes of the line segments are simplified;
and orthogonalizing adjacent sides in a threshold range by adopting a recursion method by taking the minimum rectangular outer bounding box of the building point cloud as a constraint, so as to generate a digital line drawing map DLG of the building.
Optionally, step S5 of screening the point cloud collection of the building roof component according to the independent building point clouds includes:
and according to the independent building point cloud extracted from the building point cloud example segmentation result, constructing KD-Tree, taking the normal directions of 8 points of nearest well grids of each point according to the nearest algorithm, and removing points of building facades according to the normal threshold value, wherein the rest points form a point combination set of building top surface components.
Optionally, step S5 of fitting the roof element to obtain a projected geometric representation of the roof element of the building comprises:
performing top surface component point cloud segmentation based on the extracted building top surface component point set to obtain point sets of all components of the building segmented by the top surface component point cloud;
according to the point sets of each component of the building segmented by the point cloud of the top surface component, respectively calculating the average point height of each point set as the height characteristic of the component, projecting the point sets onto the ground to form the ground plane point set of the component, expanding the points in the dense ground plane point set into a plane according to resolution, and combining the planes of all the point sets to fit into a polygon, wherein the polygon is the projection geometric representation of the component of the building.
Optionally, the performing the top surface component point cloud segmentation based on the extracted building top surface component point set, and obtaining the point set of each component of the building segmented by the top surface component point cloud comprises:
reconstructing the KD-Tree for the extracted aggregate set of building roof component points such that each point is organized in adjacent sequential order;
clustering according to the Euclidean distance threshold value between points, randomly taking out a point from a point set, adding the point to a candidate component point set, taking the point as a center, and taking out and adding the point to the component point set when the distance from the adjacent point to the point in KD-Tree is smaller than the threshold value;
and sequentially repeating the processes with newly added points as centers, extracting a component from the continuous points smaller than the threshold value, and clustering the rest point sets again until all the points are completely divided according to a clustering method to obtain the point set of each component of the building segmented by the top surface component point cloud.
Optionally, step S5 further includes: the top surface component polygon optimization specifically comprises:
for projection geometric representation of a fitted building top surface component, firstly, adopting a line segment to replace a plurality of continuous line segments which are originally close to collineation within a threshold range, simplifying polygon dense vertexes of the line segment, and then adopting a recursion method to orthogonalize adjacent sides within the threshold range by taking a building DLG as a constraint;
filling gaps among roof elements according to a polygon etching mode, so that all roof element polygons of the same building completely fill DLG interiors, and generating the building roof element polygons by keeping the height characteristic attribute unchanged while optimizing projection geometry.
Optionally, step S6 of merging the digital line map DLG representation of the coarse granularity of the building and the projection geometry representation of the building roof component, generating a multi-level of detail LOD1.3 model of the building comprises:
carrying out polygon combination on components smaller than a certain height difference threshold and an area threshold according to LOD1.3 definition, wherein the height attribute of the components is in a principle of high or low in the combination process, and the combined component polygons represent top surface characteristics on a certain abstract granularity;
after generating a LOD1.3 roof component polygon for each building instance, the component projection polygon is pulled up into a columnar 3D stereoscopic structure according to its height attribute, and the intersecting facades are eroded and exported to form the building LOD1.3 model.
According to the method for automatically generating the building LOD1.3 model, after laser point clouds or multi-view images are collected, a three-dimensional point cloud model is generated by utilizing information such as camera parameters, then a point cloud example is adopted to segment building models in the neural network to segment the point clouds, DLG fitting is carried out after single building point clouds are extracted to project on the ground, and the fitted DLG is obtained, so that the DLG of a single building is obtained. In addition, top surface point clouds are screened out according to the extracted single building point clouds and point normal information, then a clustering method is adopted to conduct clustering segmentation on top surface component point sets, then polygon fitting is conducted after the clustering segmentation of the top surface component point sets is conducted on the ground, meanwhile average height information of the point sets is calculated to serve as the height attribute of the components, the DLG range is combined to conduct post-processing optimization such as normalization on the component fitting polygons, component representation polygons are generated, finally top surface component polygon parallel drawing Gao Jianmo is combined according to the height and area threshold, and finally the building LOD1.3 model is obtained.
According to the method for automatically generating the LOD1.3 model of the building, disclosed by the invention, point cloud instance segmentation is adopted to directly separate the point cloud of the single building, fitting and post-treatment optimization are carried out to automatically generate DLG of the single building, so that the complete automation of DLG generation of the building, the complete automation of top surface component generation and optimization can be realized, and the automation of LOD1.3 model generation can be realized, so that the manual workload can be reduced and the working efficiency can be improved. In addition, the method adopts the Euclidean distance clustering algorithm to directly perform cluster fitting on the top surface component, the top component representation can be generated through post-treatment optimization, the LOD1.3 model can be automatically generated through combination according to the threshold requirement of LOD1.3, and each component in the LOD1.3 model building process is discretized in the scheme of the invention, so that manual intervention can be performed in time.
The foregoing description is only an overview of the present invention, and is intended to be implemented in accordance with the teachings of the present invention in order that the same may be more clearly understood and to make the same and other objects, features and advantages of the present invention more readily apparent.
The above, as well as additional objectives, advantages, and features of the present invention will become apparent to those skilled in the art from the following detailed description of a specific embodiment of the present invention when read in conjunction with the accompanying drawings.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to designate like parts throughout the figures. In the drawings:
FIG. 1 shows a general flow diagram of a method for automatically generating a LOD1.3 model of a building in accordance with one embodiment of the invention.
Detailed Description
Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
The embodiment of the invention provides a method for automatically generating a LOD1.3 model of a building, and in combination with fig. 1, the method for automatically generating the LOD1.3 model of the building can comprise the following steps A1 to A11.
A1, data acquisition:
the unmanned aerial vehicle or the unmanned vehicle is used for carrying a multi-view photographic camera or a laser radar, the track is set according to the parameter requirements that the heading of the forward photographic image is overlapped by 80%, the side direction is overlapped by 50%, and the like, the original data of the image or the laser point cloud with the resolution of 0.05m is collected, and the parameters such as camera positioning and the like are recorded.
A2, generating a color point cloud:
and importing the image acquired by the multi-view photographic camera or laser point cloud data acquired by the laser radar into point cloud computing software to generate colorful dense point clouds, recording information of [ x, y, z, i, r, g, b ] (position, intensity and color) of each point, and storing the information as a labs file.
A3, building point cloud instance segmentation:
the method comprises the steps of preprocessing point clouds, roughly separating the point clouds of different planes by adopting a noise density clustering algorithm, sampling each category in an equal ratio, and extracting dense point clouds into sparse point clouds. And then carrying out blocking processing on the point clouds on an [ x, y ] plane according to a coordinate segmentation grid, independently storing each point cloud as a file, and interpolating/extracting the number of points of each point cloud as 20480, wherein the intensity i of the interpolation point is set to 0, and the point intensities i of other normal point clouds are set to 1.
Taking part of the point cloud to be marked as a data set, namely recording the number of each point belonging to a certain building (the number is obtained randomly but the numbers of the same building are consistent), and if the point does not belong to any building, the number is marked as-1. And (3) learning and training by adopting an example segmentation network based on PointNet++, and identifying and reasoning unlabeled point clouds by using a trained neural network model. In the identification process, in order to ensure that a complete result can be obtained by cutting a damaged building, a nine-grid overlapping sampling mode is adopted, grid blocks of 3x3 around each grid block are sequentially combined to form point cloud blocks in a 1500mx1500m area, then a neural network model is loaded to conduct instance segmentation, the non-maximum suppression of [ x, y ] plane projection areas is conducted on the segmentation result, the building is repeatedly identified in a combined mode, the damaged point cloud identification result is removed, and the instance segmentation result of the building point cloud is generated.
A4, building DLG fitting:
and according to the building point cloud example segmentation result, the independent building point cloud can be extracted. Projecting the point cloud of the independent building to the ground to form a ground plane point set of the building, expanding the points in the dense ground plane point set into a plane according to resolution, and combining the planes of all the point sets to fit into a polygon, wherein the polygon is the DLG representation of the coarse granularity of the building.
A5, building DLG optimization:
for the DLG characterization of the obtained coarse granularity of the building, firstly, one line segment is adopted to replace a plurality of continuous line segments which are originally close to collineation within a threshold range, polygon dense vertexes of the line segments are simplified, then, a minimum rectangular outer bounding box of the point cloud of the building is taken as a constraint, and adjacent edges are orthogonalized within the threshold range by adopting a recursion method, so that the DLG of the building is generated.
A6, screening a top surface component point cloud:
and according to the independent building point cloud extracted from the building point cloud example segmentation result, constructing KD-Tree, taking the normal directions of 8 points of nearest well grids of each point according to the nearest algorithm, and removing points of building facades according to the normal threshold value, wherein the rest points form a point combination set of building top surface components.
A7, top surface component point cloud segmentation:
reconstructing KD-Tree to make each point organized according to the adjacent sequence, then clustering according to Euclidean distance threshold value, randomly taking out a point from the point set, adding it to the candidate component point set, taking the point as center, taking out the point set added to the component point set when the distance from the adjacent point to the point in KD-Tree is smaller than the threshold value, then repeating the above processes with the newly added point as center, extracting a component from the continuous point smaller than the threshold value, then clustering again in the rest point set until all points are finished to be divided according to the clustering method, judging the point set with the number smaller than 9 well lattice in the component point set as discrete point or fine structure independent surface, and eliminating. The point set of the building roof elements is thus divided into the point set of the individual elements.
A8, fitting a top surface component point cloud:
according to the point sets of each component of the building segmented by the point cloud of the top surface component, respectively calculating the average point height of each point set as the height characteristic of the component, projecting the point sets onto the ground to form the ground plane point set of the component, expanding the points in the intensive ground plane point set into a plane according to resolution, combining the planes of all the point sets to fit into a polygon, wherein the polygon is the projection geometric representation of the component of the building, and finally assigning the height characteristic of the component to the projection geometric as the attribute of the projection geometric representation.
A9, top surface component polygon optimization:
for projection geometric characterization of the top surface components of the building, firstly, adopting a line segment to replace a plurality of continuous line segments which are originally close to collineation within a threshold range, simplifying polygon dense vertexes of the line segments, then adopting a recursion method to orthogonalize adjacent sides within the threshold range by taking the building DLG as a constraint, and then filling gaps among the top surface components in a polygon external erosion mode, so that all the top surface component polygons of the same building are completely filled in the DLG, and in addition, the projection geometric optimization is carried out while the height characteristic attribute is kept unchanged, so that the top surface component polygons of the building are generated.
A10, polygon combination of the top surface components:
after the polygon optimization of the top surface components of the building, each top surface component monopolizes a region in the DLG of the building and is attached with height characteristic attributes, each top surface component polygon characterizes the characteristics of one top surface component, the components smaller than a certain height difference threshold and an area threshold can be combined in a polygon combining way according to the definition of LOD1.3, the height attributes of the components are in a principle of high or low in the combining process, and the combined component polygons characterize the top surface characteristics on a certain abstract granularity.
A11, LOD1.3 modeling:
after generating the LOD1.3 roof component polygon for each building instance, the component projection polygon is pulled up to a columnar 3D stereoscopic structure according to its height attribute, and then the intersecting facade is eroded and exported as the building LOD1.3 model.
The scheme of the embodiment of the invention adopts deep learning and machine learning technology to automatically extract the LOD1.3 model structure of the building from the point cloud data and carry out modeling. And acquiring a point cloud or an image acquired by a multi-view camera by using a laser radar as source data, generating a color point cloud model, and independently dividing a building in the point cloud by using a deep learning point cloud example division technology to extract the point cloud of the independent building.
On the one hand, the independent point clouds of the building are projected onto the ground to form a two-dimensional plane projection point set, then the projection point set is fitted into a vector plane, namely the coarse-granularity DLG of the building, and the coarse-granularity DLG is subjected to post-treatment optimization such as normalization to generate the DLG of the building.
On the other hand, independent point clouds of the building are filtered according to the normal direction, points of the building facade are removed, so that a top surface point set is extracted, the top surface point set is divided into independent component point sets through clustering calculation, the component point sets are fitted into vector surfaces to be used for representing coarse granularity of the top surface components, meanwhile, the height mean value of the component point sets is recorded to serve as the attribute of the coarse granularity vector surfaces, in addition, the coarse granularity vector surfaces are combined with the DLG of the building to be subjected to post-treatment optimization such as normalization, the representing polygons with the height attribute of the top surface components are generated, the top surface polygons with certain height and area threshold ranges are combined according to LOD1.3 standard specifications, and then the building LOD1.3 model is built through pulling up according to the height attribute.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all technical features thereof can be replaced by others within the spirit and principle of the present invention; such modifications and substitutions do not depart from the scope of the invention.
Claims (10)
1. A method for automatically generating a LOD1.3 model of a building is characterized in that,
s1, acquiring point cloud original data by adopting an unmanned plane or an unmanned vehicle carrying a multi-view camera or a laser radar;
s2, generating a color dense point cloud based on the original point cloud data;
s3, preprocessing the colored dense point cloud, and carrying out building point cloud instance segmentation to obtain an instance segmentation result of the building point cloud;
s4, extracting independent building point clouds according to example segmentation results of the building point clouds, and performing digital line drawing map DLG fitting to obtain digital line drawing map DLG characterization of coarse granularity of the building;
s5, screening according to independent building point clouds to obtain a point cloud integration set of a building top surface component, and fitting the top surface component to obtain a projection geometric representation of the building top surface component;
s6, combining the DLG representation of the coarse-grained digital line drawing map of the building and the projection geometric representation of the top surface component of the building to generate a multi-detail-level LOD1.3 model of the building.
2. The method of claim 1, wherein preprocessing the color dense point cloud in step S3 comprises:
separating point clouds of different planes in the color dense point clouds by adopting a noise density clustering algorithm, and then sampling each category in an equal ratio mode to extract the dense point clouds into sparse point clouds;
partitioning the sparse point cloud on an [ x, y ] plane according to a coordinate partitioning grid, independently storing each block of point cloud as a file, and interpolating or extracting the number of points of each block of point cloud; wherein the intensity of the interpolation point is set to 0, and the point intensity i of the other normal point clouds is 1.
3. The method according to claim 2, wherein performing building point cloud instance segmentation in step S3 to obtain an instance segmentation result of the building point cloud comprises:
for the preprocessed sparse point cloud, marking part of the point cloud as a data set;
performing learning training by using an example segmentation network based on PointNet++, and performing recognition reasoning on unlabeled point clouds by using a trained neural network model; in the identification process, a nine-grid overlapping sampling mode is adopted, grid blocks of 3x3 around each grid block are combined in sequence to form point cloud blocks, then a neural network model is loaded to conduct instance segmentation, the non-maximum suppression of the [ x, y ] plane projection area is conducted on the segmentation result, the repeated identification of the building is combined, the damaged point cloud identification result is removed, and the instance segmentation result of the building point cloud is generated.
4. The method of claim 1, wherein step S4 extracts an independent building point cloud from the example segmentation result of the building point cloud, performs DLG fitting to obtain DLG characterization of the coarse-grained building, and includes:
projecting independent building point clouds to the ground according to example segmentation results of the building point clouds to form a ground plane point set of the building;
and expanding the points in the ground point set into a plane according to resolution, and combining the planes of all the point sets to fit into a polygon, wherein the polygon is the digital line drawing map DLG representation of the coarse granularity of the building.
5. The method of claim 4, further comprising, after step S4:
for the digital line drawing map DLG representation of the obtained coarse granularity of the building, one line segment is adopted to replace a plurality of continuous line segments which are originally close to collineation in a threshold range, and polygon dense vertexes of the line segments are simplified;
and orthogonalizing adjacent sides in a threshold range by adopting a recursion method by taking the minimum rectangular outer bounding box of the building point cloud as a constraint, so as to generate a digital line drawing map DLG of the building.
6. The method of claim 1, wherein step S5 of screening the point cloud collection of building roof components from the individual building point clouds comprises:
and according to the independent building point cloud extracted from the building point cloud example segmentation result, constructing KD-Tree, taking the normal directions of 8 points of nearest well grids of each point according to the nearest algorithm, and removing points of building facades according to the normal threshold value, wherein the rest points form a point combination set of building top surface components.
7. The method of claim 6, wherein step S5 of fitting the roof element to obtain a projected geometric representation of the roof element of the building comprises:
performing top surface component point cloud segmentation based on the extracted building top surface component point set to obtain point sets of all components of the building segmented by the top surface component point cloud;
according to the point sets of each component of the building segmented by the point cloud of the top surface component, respectively calculating the average point height of each point set as the height characteristic of the component, projecting the point sets onto the ground to form the ground plane point set of the component, expanding the points in the dense ground plane point set into a plane according to resolution, and combining the planes of all the point sets to fit into a polygon, wherein the polygon is the projection geometric representation of the component of the building.
8. The method of claim 7, wherein performing a ceiling element point cloud segmentation based on the extracted aggregate set of building ceiling element points to obtain a set of points for each element of the ceiling element point cloud segmented building comprises:
reconstructing the KD-Tree for the extracted aggregate set of building roof component points such that each point is organized in adjacent sequential order;
clustering according to the Euclidean distance threshold value between points, randomly taking out a point from a point set, adding the point to a candidate component point set, taking the point as a center, and taking out and adding the point to the component point set when the distance from the adjacent point to the point in KD-Tree is smaller than the threshold value;
and sequentially repeating the processes with newly added points as centers, extracting a component from the continuous points smaller than the threshold value, and clustering the rest point sets again until all the points are completely divided according to a clustering method to obtain the point set of each component of the building segmented by the top surface component point cloud.
9. The method of claim 7, further comprising, after step S5: the top surface component polygon optimization specifically comprises:
for projection geometric representation of a fitted building top surface component, firstly, adopting a line segment to replace a plurality of continuous line segments which are originally close to collineation within a threshold range, simplifying polygon dense vertexes of the line segment, and then adopting a recursion method to orthogonalize adjacent sides within the threshold range by taking a building DLG as a constraint;
filling gaps among roof elements according to a polygon etching mode, so that all roof element polygons of the same building completely fill DLG interiors, and generating the building roof element polygons by keeping the height characteristic attribute unchanged while optimizing projection geometry.
10. The method of claim 9, wherein step S6 of merging based on the coarse-grained digital line map DLG representation and the building roof component projection geometry representation, generating a building multi-level of detail LOD1.3 model comprises:
carrying out polygon combination on components smaller than a certain height difference threshold and an area threshold according to LOD1.3 definition, wherein the height attribute of the components is in a principle of high or low in the combination process, and the combined component polygons represent top surface characteristics on a certain abstract granularity;
after generating a LOD1.3 roof component polygon for each building instance, the component projection polygon is pulled up into a columnar 3D stereoscopic structure according to its height attribute, and the intersecting facades are eroded and exported to form the building LOD1.3 model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311094611.2A CN117315184B (en) | 2023-08-29 | 2023-08-29 | Method for automatically generating LOD1.3 model of building |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311094611.2A CN117315184B (en) | 2023-08-29 | 2023-08-29 | Method for automatically generating LOD1.3 model of building |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117315184A true CN117315184A (en) | 2023-12-29 |
| CN117315184B CN117315184B (en) | 2024-03-29 |
Family
ID=89241580
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311094611.2A Active CN117315184B (en) | 2023-08-29 | 2023-08-29 | Method for automatically generating LOD1.3 model of building |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117315184B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090310867A1 (en) * | 2008-06-12 | 2009-12-17 | Bogdan Calin Mihai Matei | Building segmentation for densely built urban regions using aerial lidar data |
| CN103198494A (en) * | 2013-04-03 | 2013-07-10 | 武汉海达数云技术有限公司 | Building elevation extraction method based on building outer contour lines and laser-point cloud |
| CN106600680A (en) * | 2015-10-20 | 2017-04-26 | 星际空间（天津）科技发展有限公司 | Batch type refined three-dimensional modeling method of building frame model |
| CN107967713A (en) * | 2017-11-21 | 2018-04-27 | 泰瑞数创科技(北京)有限公司 | Construction three-dimensional model building method and system based on spatial point cloud data |
| WO2019242174A1 (en) * | 2018-06-21 | 2019-12-26 | 华南理工大学 | Method for automatically detecting building structure and generating 3d model based on laser radar |
| CN111340822A (en) * | 2020-02-24 | 2020-06-26 | 武汉大学 | Multi-scale self-adaptive airborne LiDAR point cloud building single segmentation method |
| CN114612488A (en) * | 2022-03-11 | 2022-06-10 | 中国电建集团中南勘测设计研究院有限公司 | Building-integrated information extraction method, computer device, and storage medium |
| CN116379915A (en) * | 2023-02-28 | 2023-07-04 | 桂林理工大学 | Building mapping method, device, system and storage medium |
-
2023
- 2023-08-29 CN CN202311094611.2A patent/CN117315184B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090310867A1 (en) * | 2008-06-12 | 2009-12-17 | Bogdan Calin Mihai Matei | Building segmentation for densely built urban regions using aerial lidar data |
| CN103198494A (en) * | 2013-04-03 | 2013-07-10 | 武汉海达数云技术有限公司 | Building elevation extraction method based on building outer contour lines and laser-point cloud |
| CN106600680A (en) * | 2015-10-20 | 2017-04-26 | 星际空间（天津）科技发展有限公司 | Batch type refined three-dimensional modeling method of building frame model |
| CN107967713A (en) * | 2017-11-21 | 2018-04-27 | 泰瑞数创科技(北京)有限公司 | Construction three-dimensional model building method and system based on spatial point cloud data |
| WO2019242174A1 (en) * | 2018-06-21 | 2019-12-26 | 华南理工大学 | Method for automatically detecting building structure and generating 3d model based on laser radar |
| CN111340822A (en) * | 2020-02-24 | 2020-06-26 | 武汉大学 | Multi-scale self-adaptive airborne LiDAR point cloud building single segmentation method |
| CN114612488A (en) * | 2022-03-11 | 2022-06-10 | 中国电建集团中南勘测设计研究院有限公司 | Building-integrated information extraction method, computer device, and storage medium |
| CN116379915A (en) * | 2023-02-28 | 2023-07-04 | 桂林理工大学 | Building mapping method, device, system and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 胡思雨 等: "基于 LiDAＲ数据的1∶2000 DLG生产分析", 测绘与空间地理信息, vol. 41, no. 9, 30 September 2018 (2018-09-30), pages 239 - 247 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117315184B (en) | 2024-03-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112489212B (en) | Intelligent building three-dimensional mapping method based on multi-source remote sensing data | |
| CN111915730B (en) | Method and system for automatically generating indoor three-dimensional model by taking semantic slave point cloud into consideration | |
| CN107146280B (en) | Point cloud building reconstruction method based on segmentation | |
| CN111986322B (en) | Point cloud indoor scene layout reconstruction method based on structural analysis | |
| GB2457215A (en) | Automatic 3D Modelling | |
| US20100284607A1 (en) | Method and system for generating a 3d model from images | |
| CN111652241B (en) | Building contour extraction method integrating image features and densely matched point cloud features | |
| CN111667574B (en) | Method for automatically reconstructing regular facade three-dimensional model of building from oblique photography model | |
| CN114332291B (en) | Method for extracting outline rule of oblique photography model building | |
| CN115600307B (en) | Method for generating single building from Mesh model of urban scene | |
| CN112465991B (en) | Method for denoising tunnel point cloud and generating visual model | |
| CN115937461B (en) | Multi-source fusion model construction and texture generation method, device, medium and equipment | |
| Murtiyoso et al. | Point cloud segmentation and semantic annotation aided by GIS data for heritage complexes | |
| CN114463521A (en) | Building target point cloud rapid generation method for air-ground image data fusion | |
| CN114677388A (en) | Room layout dividing method based on unit decomposition and space division | |
| CN117315184B (en) | Method for automatically generating LOD1.3 model of building | |
| CN116071530B (en) | Building roof voxelized segmentation method based on airborne laser point cloud | |
| Yamazaki et al. | Non-manifold implicit surfaces based on discontinuous implicitization and polygonization | |
| CN115033972A (en) | Method and system for unitizing building main body structures in batches and readable storage medium | |
| Xiao | Automatic building outlining from multi-view oblique images | |
| Pu | Generating building outlines from terrestrial laser scanning | |
| Yan et al. | UBMDP: Urban Building Mesh Decoupling and Polygonization | |
| Gkeli et al. | Automatic 3D reconstruction of buildings roof tops in densely urbanized areas | |
| ji Ban et al. | 3D Mesh Reconstruction from Height map and Post Processing | |
| CN117437364B (en) | Method and device for extracting three-dimensional structure of building based on residual defect cloud data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |