CN113313835A - Building roof automatic modeling method based on airborne LiDAR point cloud - Google Patents
Building roof automatic modeling method based on airborne LiDAR point cloud Download PDFInfo
- Publication number
- CN113313835A CN113313835A CN202110860546.4A CN202110860546A CN113313835A CN 113313835 A CN113313835 A CN 113313835A CN 202110860546 A CN202110860546 A CN 202110860546A CN 113313835 A CN113313835 A CN 113313835A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- roof
- point
- building
- feature
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/20—Finite element generation, e.g. wire-frame surface description, tesselation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/13—Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Abstract
The invention discloses an automatic building roof modeling method based on airborne LiDAR point cloud, which comprises the following steps: s1 learning and resampling based on the point cloud features that generate the antagonistic neural network; s2 extracting building roof patch elements based on direction regularization constraint and global clustering; s3 building a building roof three-dimensional model based on primitive patch geometric and semantic information. The method can acquire building roof point cloud data with uniform density and moderate magnitude based on the airborne LiDAR point cloud, realize efficient and accurate roof element patch extraction, extract complete roof structure key angular points, complete automatic modeling of a three-dimensional building roof model, and promote application of the airborne LiDAR point cloud in the fields of urban real scene three-dimensional modeling, energy-saving simulation, urban solar energy capture and the like.
Description
Technical Field
The invention relates to an automatic building roof modeling method based on airborne LiDAR point cloud, belongs to the technical field of new generation information, and particularly relates to the technical field of data identification.
Background
The laser radar (LiDAR) technology can rapidly acquire laser point cloud data with three-dimensional coordinates, intensity AND other information on the surface of an object in a non-contact, high-resolution, all-weather AND high-precision manner, AND has important application in the fields of smart cities, three-dimensional real-scene, automatic driving, major infrastructure health monitoring, ecological remote sensing AND the like. The airborne LiDAR point cloud has the advantages of incomparable ratio in the aspect of three-dimensional building roof modeling due to high acquisition speed, wide coverage area and high point location precision, and the airborne LiDAR point cloud is core data infrastructure applied to urban three-dimensional live-action modeling, urban solar energy potential estimation and the like. However, the three-dimensional point cloud has the characteristics of huge data volume, high data density, high redundancy, non-structural property, incompleteness and the like, and the data identification difficulty is high. The intelligent identification and processing of point cloud data is a key between the realization of point cloud data and scientific research and engineering application, and the research core of the intelligent identification and processing method comprises multiple tasks of point cloud data quality enhancement, three-dimensional information intelligent extraction, on-demand three-dimensional reconstruction and the like. The existing building roof element extraction method based on point cloud data has the problems of low robustness, low time efficiency, ambiguous model topological relation, lack of semantic information and the like. In order to fully utilize the advantages of the airborne LiDAR technology and promote the application of airborne LiDAR point cloud data in the fields such as urban real-scene three-dimensional modeling, energy-saving simulation, urban solar energy capture and the like, a method capable of intelligently and quickly reconstructing a building roof based on airborne LiDAR point cloud is urgently needed.
Disclosure of Invention
The invention aims to provide an automatic building roof modeling method based on airborne LiDAR point cloud, and promotes the application of the automatic building roof modeling method in the fields of smart city three-dimensional modeling, city energy conservation, city solar energy capture and the like.
In order to solve the technical problem, the invention provides an automatic building roof modeling method based on airborne LiDAR point cloud, which comprises the following steps:
s1, adopting a designed generation countermeasure neural network to perform up-sampling processing on the building roof point cloud data obtained after filtering processing, and then performing down-sampling processing on the encrypted roof point cloud data through voxel filtering to obtain point cloud data with uniform distribution and moderate magnitude;
s2, adopting the designed element patch fitting method based on direction regular constraint, carrying out regular constraint on the point cloud through the direction constraint, and converting the point cloud into a point cloudL 0 Obtaining a building roof element patch optimization result by a gradient minimization problem and a subset selection problem;
s3 extracts the contour line of the surface patch of the roof element based on the Alphashape algorithm, simplifies the contour line by adopting an angle threshold method to obtain the angular point information, constructs a roof topological graph by constructing a minimum ring structure based on the roof topological graph algorithm, and further obtains the intersection point, the inner point and the outer point of the roof structure to complete the modeling of the roof topology and the geometry.
Preferably, in step S1, the designed antagonistic neural network is used to perform upsampling on the filtered building roof point cloud data, and then the encrypted roof point cloud data is downsampled through voxel filtering to obtain point cloud data with uniform distribution and moderate magnitude, which specifically includes the following steps:
s11, filtering outliers and sparse facade points in the original building point cloud data by adopting a sparse outlier statistical filtering method to obtain building roof point cloud data;
s12 designing a generation countermeasure neural network to extract and expand the characteristics of the roof point cloud data, generating dense point cloud, designing a discriminator network, defining the loss function of the generator, identifying the influence of the generated point set on the distribution of potential points and punishing the output deviating from the target;
s13, the roof dense point cloud data obtained by up-sampling is down-sampled by adopting voxel filtering.
Preferably, in step S2, the designed primitive patch fitting method based on directional regular constraint is adopted, and the point cloud is regularly constrained by the directional constraint and converted into a point cloudL 0 The method specifically comprises the following steps of obtaining a building roof element patch optimization result by a gradient minimization problem and a subset selection problem:
s21, using implicit angle constraint to define a direction constraint model;
s22, solving a point cloud initial normal vector by using a principal component analysis method, and reconstructing the point cloud normal vector by using a direction constraint model;
s23 converting the directional constraint model problem intoL 0 Establishing an implicit relation between disjoint areas by a gradient minimization problem and a subset selection problem to complete global regularity;
s24, solving the problem in S23, dividing the input point cloud into a plurality of connected regions according to the normal vector direction according to the output normal vector set, and performing plane fitting on each region to obtain the result of the primitive patch.
Preferably, in step S3, extracting a roof element patch contour line based on an AlphaShape algorithm, simplifying the contour line by using an angle threshold method to obtain corner point information, constructing a roof topological graph by constructing a minimum ring structure based on a roof topological graph algorithm, and further obtaining a roof structure intersection point, an inner point and an outer point, thereby completing the roof topology and geometric modeling specifically including the steps of:
s31, setting specific geometric shapes and polygonal precision parameters based on an Alphashape algorithm to obtain the initial contour of each roof patch;
s32, sequentially calculating cosine values of included angles between two straight lines formed by adjacent three points on the contour line, and completing contour line simplification by adopting an angle threshold method to obtain angular point information;
s33 defining an intersection point, an inner point and an outer point isocenter, judging the topological relation among planes based on a roof topological graph algorithm, constructing a minimum ring structure, judging and calculating a roof structure key point, a structure intersection point and an inner point according to the geometrical structure rule of a roof patch of a building, and regularizing the outer point;
s34, storing the obtained building roof element by adopting a CityGML layered object to complete building of a building roof model.
The invention has the beneficial effects that: compared with the prior art, the method has the advantages that the countermeasure network and the voxel filtering are designed and generated to perform feature enhancement up-sampling and down-sampling on the point cloud data, and the plane normal vector is reconstructed based on the implicit direction constraint condition, so that the point cloud data of the roof of the building with moderate magnitude and more uniform density can be obtained while the point cloud features are kept; the method has the advantages that the roof element surface patches are extracted in batches more efficiently and accurately, the key corner points and the topological relation of the complete roof structure are generated and constructed, the automatic modeling of the three-dimensional building roof model is completed, the building roof modeling efficiency is greatly improved, and the application of airborne LiDAR point clouds in the fields of urban real-scene three-dimensional modeling, energy-saving simulation, urban solar energy capture and the like is promoted.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention;
FIG. 2 is a schematic diagram of a generator network;
FIG. 3 is a schematic diagram of a feature expansion unit;
FIG. 4 is a schematic diagram of a discriminator network;
FIG. 5 is a schematic view of a self-attention module;
FIG. 6 is a schematic flow chart of a plane adjacent topology relationship judgment algorithm.
Detailed Description
In order to enhance the understanding and comprehension of the present invention, the technical solution is described in detail below with reference to the embodiments.
Example 1: as shown in FIG. 1, a method for automatically modeling a building roof based on airborne LiDAR point clouds includes the following steps:
s1: designing a generation countermeasure neural network, carrying out up-sampling processing on building roof point cloud data obtained after filtering processing, and then carrying out down-sampling processing on encrypted roof point cloud data through voxel filtering to obtain point cloud data with uniform distribution and moderate magnitude. The method comprises the following steps and contents:
s11: filtering outliers and sparse elevation points in original building point cloud data by adopting a sparse outlier statistical filtering method to obtain building roof point cloud data;
s12: a method for generating an antagonistic neural network to extract and expand the features of the point cloud data of the roof is designed, the antagonistic neural network is used for generating dense point clouds, a discriminator network is designed, the loss function of a generator is defined, the influence of a generated point set on the distribution of potential points is identified, and punishment is carried out on the output deviating from a target. Wherein the content of the first and second substances,
designing and generating a Generic Adaptive Network (GAN), using an antagonistic learning strategy of a GAN framework to perform global prediction on an input point cloud, and penalizing an output deviating from a target includes the following steps:
(1) firstly, an Up-Down-Up unit is constructed to expand point characteristics, the difference value of the point characteristics is Up-sampled to achieve the self-correcting purpose, and the characteristic expansion capability of a generator is improved. The design generator network is shown in fig. 2 and comprises a feature extraction unit, a feature expansion unit and a point cloud generation unit;
(2) feature extraction unit extracts feature points from an original point cloudPMid-extraction of bottom layer geometric featuresFIntegrating different hierarchical features in a dense connection mode by adopting a Multi-step Patch-based progressive Upsampling (MPU) method;
(3) the feature expansion unit extracts the point cloud features as shown in FIG. 3FPerforming expansion to obtain featuresF up A feature increasing module (Up) and a feature reducing module (Down) are designed in the unit, and the diversity of the acquired features is increased by an Up-Down-Up mode. Firstly, the extracted feature F is processed by a multilayer perceptron to obtain a feature dimension increasingF 1 Then, the feature increasing module is used for carrying out feature upsampling to obtain initial extended features
Then performing feature reduction to obtainF 1 Features of the same dimensionF 2 Then will beF 1 AndF 2 difference of (2)ΔPerforming feature expansion to obtainΔupThe final extended feature Fup is composed ofΔupExtended features from the original
Adding to obtain;
(4) a discriminator network based on the MUP baseline network is designed as shown in fig. 4. The global features and the final confidence values are respectively obtained by using a group of shared multilayer perceptrons, maximum pooling and full-connection layer regression, if the confidence values are close to 1, the prediction input comes from a target interval, otherwise, the prediction input comes from a generator;
(5) a self-attention module is added in a generator for both feature expansion and a discriminator, as shown in fig. 5, input features are converted into G, H and K through three independent multi-layer perceptrons, attention weight W is obtained through G and H, weight feature WTK is obtained, and final output features are the sum of the input features and the weight features, wherein a loss function of the generator is defined as composite loss and comprises least square loss, uniform loss and reconstruction loss.
S13: and performing downsampling on the roof dense point cloud data acquired by upsampling by adopting voxel filtering. Wherein the content of the first and second substances,
the voxel meshing method is as follows:
(1) setting the voxel size according to the average distance of the point cloud after up-sampling and the average error of the original point cloud data, and carrying out voxel grid division on the whole up-sampled point cloud data;
(2) replacing the point cloud in the original voxel grid with a random point in each voxel grid;
s2: the designed element patch fitting method based on the direction regular constraint is adopted to carry out the regular constraint on the point cloud through the direction constraint and convert the point cloud into the point cloudL 0 The gradient minimization problem and the subset selection problem are solved, and the step of obtaining the optimization result of the roof element surface patch of the building specifically comprises the following steps：
S21: using implicit angle constraints, a direction constraint model is defined. Wherein the content of the first and second substances,
the orientation constraint model is defined as follows:
given a set of planes P, V is represented as a set of different normal vectors for P, if and only if P and V satisfy
S22: and solving a point cloud initial normal vector by using a principal component analysis method, and reconstructing the point cloud normal vector by using a direction constraint model. Wherein the content of the first and second substances,
(1) giving a point cloud data set D, solving a point cloud initial normal vector I by using a principal component analysis method, and reconstructing the point cloud initial normal vector I to meet the requirement
Wherein:
v represents a reconstructed point cloud normal vector set;
e is an energy function used for measuring the matching degree between V and the original normal vector I;
and m is the number of normal vectors in different directions of the constraint reconstruction plane.
(2) By usingL 0 The energy function is solved to obtain E, and the formula 2 is converted into
Wherein:
v and m have the same meaning as formula (2);
n represents the number of point clouds;
λthe parameters are used to balance the effects of both terms.
(3) Equation 3 is divided into two sub-problem solutions, where,
sub-problem 1:
sub-problem 2:
the numerical meaning of the formula is the same as above.
S23: transforming the problem of the directional constraint model intoL 0 And establishing an implicit relation between disjoint areas by a gradient minimization problem and a subset selection problem to complete global regularity. Wherein the content of the first and second substances,
(1) the global regular constraint algorithm is designed as follows:
inputting: the method comprises the following steps of point cloud D, an initial normal vector I, a neighborhood set xi, a local regular parameter lambda and a global regular parameter m;
and (3) outputting: resetting the normal vector V;
3: repeating;
(2) the subset selection algorithm is as follows:
inputting: disjoint sets of regions epsilon, region-shared normal vectors
And (3) outputting: a subset normal vector V;
1: the aggregate epsilon is sorted according to the descending order of points;
9：else；
11：end if；
12：end for；
13：return V；
s24: and (6) solving the problem in the S23, dividing the input point cloud into a plurality of connected regions according to the normal vector direction according to the output normal vector set, and performing plane fitting on each region to obtain the result of the element patch.
S3: the method comprises the following steps of extracting contour lines of roof element surface patches based on an Alphashape algorithm, simplifying the contour lines by adopting an angle threshold method to obtain angular point information, constructing a roof topological graph by constructing a minimum ring structure based on a roof topological graph algorithm, further solving intersection points, inner points and outer points of the roof structure, and completing roof topology and geometric modeling, wherein the method specifically comprises the following steps:
s31: and setting specific geometric shapes and polygonal precision parameters based on an Alphashape algorithm to obtain the initial contour of each roof patch.
S32: and sequentially calculating cosine values of included angles between two straight lines formed by adjacent three points on the contour line, and finishing contour line simplification by adopting an angle threshold method to obtain angular point information. The angle threshold is set according to the density and quality of the point cloud, and is set to 30 degrees in this embodiment.
S33: defining the isocenter of the intersection point, the inner point and the outer point, judging the topological relation among planes based on a roof topological graph algorithm, constructing a minimum ring structure, judging and calculating the key point of the roof structure and the intersection point and the inner point of the structure according to the geometrical structure rule of a roof patch of the building, and regularizing the outer point. Wherein the content of the first and second substances,
(1) defining the intersection, the inner point and the outer point as follows:
intersection point: a point resulting from the intersection of three non-coplanar intersecting surfaces;
interior point: points on the intersection line of two adjacent roof surfaces which are not coplanar;
exterior point: the outer corner points on the contour lines of the adjacent roof surfaces are absent;
(2) based on a theoretical algorithm of a roof topological graph, abandoning concepts such as step edges and the like, constructing the roof topological graph by constructing a minimum ring structure, and setting the minimum number of link points as 3 to obtain intersection points; determining adjacent roof surfaces through the edge of each ring, solving an intersection line and projecting the simplified contour line to the intersection line to obtain an inner point;
(3) judging the adjacent topological relation among all planes of the roof, and designing an algorithm flow chart as shown in figure 6: firstly, selecting any plane P1Then any other plane P is selected2Setting a distance threshold D to represent P2Point cloud and P in plane1Simplifying the distance threshold point by point of the contour line, setting a number threshold N to represent P2Point cloud and P in plane1The number of the point-by-point distances D of the simplified contour line is smaller than the distance threshold value D. Calculating P1Point-by-point and P in simplified contour2Judging whether the distance of the inner point cloud meets a set distance threshold D and a number threshold N, and if D is smaller than the number N of D and larger than N, adding P2Is recorded as P1An abutment surface of (a); sequentially selecting the rest planes and repeating the operations until the judgment of the relation of all the adjacent planes is completed;
(4) and searching each plane in sequence by using a breadth-first search algorithm, and judging whether adjacent topological relations exist between the planes and the rest planes or not so as to construct a minimum roof topological graph structure.
(5) By utilizing parameters and topological relations of all surface patches of the roof, the intersection point of the roof structure is directly obtained through the geometrical structure relation, the inner point is obtained through projection of the intersection line of the adjacent surface patches and the simplified structure line, and the outer point is obtained through contour line regularization optimization.
S34: and storing the obtained roof element of the building by adopting a CityGML hierarchical object to complete building of a roof model of the building. The geometric object model defined by GML3 represents a building model with a level of detail, extracted roof element patches are stored by using a hierarchical object of ctygml, the building roof is composed of a plurality of patches, the patches are composed of lines, and the lines are composed of key points such as intersection points, interior points, exterior points and the like obtained in each step S33. And combining the topological characteristics of the key points to form different primitive patches, and storing the primitive patches as a three-dimensional model in a CityGML format.
It should be noted that the above-mentioned embodiments are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention, and all equivalent substitutions or substitutions made on the basis of the above-mentioned technical solutions belong to the scope of the present invention.
Claims (8)
1. An automatic building roof modeling method based on airborne LiDAR point cloud is characterized by comprising the following steps:
s1: adopting a designed generation countermeasure neural network to perform up-sampling processing on the building roof point cloud data obtained after filtering processing, and then performing down-sampling processing on the encrypted roof point cloud data through voxel filtering to obtain point cloud data with uniform distribution and moderate magnitude;
s2: the designed element patch fitting method based on the direction regular constraint is adopted to carry out the regular constraint on the point cloud through the direction constraint and convert the point cloud into the point cloudL 0 Obtaining a building roof element patch optimization result by a gradient minimization problem and a subset selection problem;
s3: the method comprises the steps of extracting contour lines of roof element surface patches based on an Alphashape algorithm, simplifying the contour lines by adopting an angle threshold method to obtain angular point information, constructing a roof topological graph by constructing a minimum ring structure based on a roof topological graph algorithm, further solving intersection points, inner points and outer points of the roof structure, and completing roof topological and geometric modeling.
2. The method of claim 1 for automatic building rooftop modeling based on airborne LiDAR point clouds, comprising: in step S1, the method specifically includes the following steps:
s11, filtering outliers and sparse facade points in the original building point cloud data by adopting a sparse outlier statistical filtering method to obtain building roof point cloud data;
s12, designing a generation confrontation neural network to extract and expand the characteristics of the roof point cloud data, generating dense point clouds, designing a discriminator network to identify the influence of the generated point set on the distribution of potential points and punishing the output deviating from the target;
s13, the roof dense point cloud data obtained by up-sampling is down-sampled by adopting voxel filtering.
3. The method of claim 1 for automatic building rooftop modeling based on airborne LiDAR point clouds, comprising: in step S2, the method specifically includes the following steps:
s21: using implicit angle constraint to define a direction constraint model;
s22: solving a point cloud initial normal vector by using a principal component analysis method, and reconstructing the point cloud normal vector by using a direction constraint model;
s23: transforming the problem of the directional constraint model intoL 0 Establishing an implicit relation between disjoint areas by a gradient minimization problem and a subset selection problem to complete global regularity;
s24: and (6) solving the problem in the S23, dividing the input point cloud into a plurality of connected regions according to the normal vector direction according to the output normal vector set, and performing plane fitting on each region to obtain the result of the element patch.
4. The method of claim 1 for automatic building rooftop modeling based on airborne LiDAR point clouds, comprising: in step S3, the method specifically includes the following steps:
s31: setting specific geometric shapes and polygonal precision parameters based on an Alphashape algorithm to obtain the initial contour of each roof surface patch;
s32: calculating cosine values of included angles between two straight lines formed by adjacent three points on the contour line in sequence, and completing contour line simplification by adopting an angle threshold method to obtain angular point information;
s33: defining an intersection point, an inner point and an outer point isocenter, judging the topological relation among planes based on a roof topological graph algorithm, constructing a minimum ring structure, judging and calculating a key point of a roof structure and the intersection point and the inner point of the structure according to the geometrical structure rule of a roof patch of a building, and regularizing the outer point;
s34: and storing the obtained roof element of the building by adopting a CityGML hierarchical object to complete building of a roof model of the building.
5. The method of claim 2 for automatic building rooftop modeling based on airborne LiDAR point clouds, comprising: designing and generating a confrontation Network generic adaptive Network, GAN in step S12, performing global prediction on the input point cloud by using a confrontation learning strategy of a GAN framework, and penalizing the output deviating from a target, wherein the steps include the following steps:
(1) firstly, constructing an Up-Down-Up unit to expand point characteristics, Up-sampling a difference value of the point characteristics to achieve the purpose of self-correction, and improving the characteristic expansion capability of a generator, wherein the Up-Down-Up unit comprises a characteristic extraction unit, a characteristic expansion unit and a point cloud generation unit;
(2) feature extraction unit extracts feature points from an original point cloudPMid-extraction of bottom layer geometric featuresFIntegrating different hierarchical features in a dense connection mode by adopting a Multi-step Patch-based gradually up-sampling Multi-step upscaling and MPU method based on patches;
(3) the extracted point cloud characteristicsFPerforming expansion to obtain featuresF up A feature increasing module (Up) and a feature reducing module Down are designed in the unit, the diversity of acquired features is increased in an Up-Down-Up mode, firstly, the extracted features F are processed by a multilayer perceptron to obtain feature dimension increasingF 1 Then, the feature increasing module is used for carrying out feature upsampling to obtain initial extensionFeature(s)
(ii) a Then performing feature reduction to obtainF 1 Features of the same dimensionF 2 Then will beF 1 AndF 2 difference of (2)ΔPerforming feature expansion to obtainΔupThe final extended feature Fup is composed ofΔupExtended features from the original
Adding to obtain;
(4) designing a discriminator network based on an MUP baseline network, respectively solving the global characteristics and the final confidence value by using a group of shared multilayer perceptrons, maximum pooling and full-link layer regression, if the confidence value is close to 1, predicting input from a target interval, and otherwise, generating the input from a generator;
(5) a self-attention module is added in a generator for feature expansion and a discriminator, input features are converted into G, H and K through three independent multilayer perceptrons, attention weight W is obtained through G and H, weight feature WTK is obtained, and finally output features are the sum of the input features and the weight features; the loss function of the generator is defined as a composite loss, including least square loss, uniform loss and reconstruction loss.
6. The method of claim 2 for automatic building rooftop modeling based on airborne LiDAR point clouds, comprising: s13: adopting voxel filtering to carry out downsampling on roof dense point cloud data obtained by upsampling, wherein the voxel grid division method comprises the following steps:
(1) setting the voxel size according to the average distance of the point cloud after up-sampling and the average error of the original point cloud data, and carrying out voxel grid division on the whole up-sampled point cloud data;
(2) and replacing the point cloud in the original voxel grid by a random point in each voxel grid.
7. The method of claim 3 for automatic building rooftop modeling based on airborne LiDAR point clouds, wherein:
s21: using implicit angle constraints, a direction constraint model is defined, as follows,
the orientation constraint model is defined as follows:
given a set of planes P, V is represented as a set of different normal vectors for P, if and only if P and V satisfy
8. The method of claim 3 for automatic building rooftop modeling based on airborne LiDAR point clouds, wherein:
s22: the method comprises the steps of solving a point cloud initial normal vector by using a principal component analysis method, and reconstructing the point cloud normal vector by using a direction constraint model, wherein,
(1) giving a point cloud data set D, solving a point cloud initial normal vector I by using a principal component analysis method, and reconstructing the point cloud initial normal vector I to meet the requirement
V is a reconstructed point cloud normal vector set, and E is used for measuring the matching degree between V and an original normal vector I;
(2) by usingL 0 The energy function is solved to obtain E, and the formula 2 is converted into
Wherein N represents the number of point clouds,
a neighborhood set representing the ith point is shown,
representing an index of the ith point cloud of the reconstruction method vector set;
(3) equation 3 is divided into two sub-problem solutions, where,
sub-problem 1:
sub-problem 2:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110860546.4A CN113313835B (en) | 2021-07-29 | 2021-07-29 | Building roof automatic modeling method based on airborne LiDAR point cloud |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110860546.4A CN113313835B (en) | 2021-07-29 | 2021-07-29 | Building roof automatic modeling method based on airborne LiDAR point cloud |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113313835A true CN113313835A (en) | 2021-08-27 |
| CN113313835B CN113313835B (en) | 2021-11-09 |
Family
ID=77382133
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110860546.4A Active CN113313835B (en) | 2021-07-29 | 2021-07-29 | Building roof automatic modeling method based on airborne LiDAR point cloud |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113313835B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113822999A (en) * | 2021-09-22 | 2021-12-21 | 杭州趣村游文旅集团有限公司 | Building segmentation method of digital country three-dimensional model |
| CN114004871A (en) * | 2022-01-04 | 2022-02-01 | 山东大学 | Point cloud registration method and system based on point cloud completion |
| CN114332366A (en) * | 2021-12-24 | 2022-04-12 | 西运才 | Digital city single house point cloud facade 3D feature extraction method |
| CN115115806A (en) * | 2022-08-30 | 2022-09-27 | 北京飞渡科技有限公司 | Roof parameterization reconstruction method and system based on single-element analysis |
| CN115393542A (en) * | 2022-07-13 | 2022-11-25 | 南京林业大学 | Generalized building three-dimensional geometric reconstruction method |
| CN115564926A (en) * | 2022-12-06 | 2023-01-03 | 武汉大学 | Three-dimensional patch model construction method based on image building structure learning |
| CN116933359A (en) * | 2023-06-26 | 2023-10-24 | 武汉峰岭科技有限公司 | Building complex roof modeling method and system |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106228539A (en) * | 2016-07-12 | 2016-12-14 | 北京工业大学 | Multiple geometric primitive automatic identifying method in a kind of three-dimensional point cloud |
| CN110458939A (en) * | 2019-07-24 | 2019-11-15 | 大连理工大学 | The indoor scene modeling method generated based on visual angle |
| US20200090357A1 (en) * | 2018-09-14 | 2020-03-19 | Lucas PAGÉ-CACCIA | Method and system for generating synthetic point cloud data using a generative model |
| CN110941996A (en) * | 2019-11-04 | 2020-03-31 | 深圳市唯特视科技有限公司 | Target and track augmented reality method and system based on generation of countermeasure network |
| CN111080671A (en) * | 2019-12-27 | 2020-04-28 | 深圳大学 | Motion prediction method based on deep neural network and intelligent terminal |
| CN111899353A (en) * | 2020-08-11 | 2020-11-06 | 长春工业大学 | Three-dimensional scanning point cloud hole filling method based on generation countermeasure network |
| CN112241997A (en) * | 2020-09-14 | 2021-01-19 | 西北大学 | Three-dimensional model establishing and repairing method and system based on multi-scale point cloud up-sampling |
| KR20210009198A (en) * | 2019-07-16 | 2021-01-26 | 연세대학교 산학협력단 | Method and Apparatus for Generating Point Cloud Using Feature of Object Acquired from Image |
| CN112561796A (en) * | 2020-12-02 | 2021-03-26 | 西安电子科技大学 | Laser point cloud super-resolution reconstruction method based on self-attention generation countermeasure network |
| CN112884886A (en) * | 2021-03-17 | 2021-06-01 | 南通大学 | Three-dimensional point cloud pipeline extraction and modeling method capable of self-adapting to search radius |
-
2021
- 2021-07-29 CN CN202110860546.4A patent/CN113313835B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106228539A (en) * | 2016-07-12 | 2016-12-14 | 北京工业大学 | Multiple geometric primitive automatic identifying method in a kind of three-dimensional point cloud |
| US20200090357A1 (en) * | 2018-09-14 | 2020-03-19 | Lucas PAGÉ-CACCIA | Method and system for generating synthetic point cloud data using a generative model |
| KR20210009198A (en) * | 2019-07-16 | 2021-01-26 | 연세대학교 산학협력단 | Method and Apparatus for Generating Point Cloud Using Feature of Object Acquired from Image |
| CN110458939A (en) * | 2019-07-24 | 2019-11-15 | 大连理工大学 | The indoor scene modeling method generated based on visual angle |
| CN110941996A (en) * | 2019-11-04 | 2020-03-31 | 深圳市唯特视科技有限公司 | Target and track augmented reality method and system based on generation of countermeasure network |
| CN111080671A (en) * | 2019-12-27 | 2020-04-28 | 深圳大学 | Motion prediction method based on deep neural network and intelligent terminal |
| CN111899353A (en) * | 2020-08-11 | 2020-11-06 | 长春工业大学 | Three-dimensional scanning point cloud hole filling method based on generation countermeasure network |
| CN112241997A (en) * | 2020-09-14 | 2021-01-19 | 西北大学 | Three-dimensional model establishing and repairing method and system based on multi-scale point cloud up-sampling |
| CN112561796A (en) * | 2020-12-02 | 2021-03-26 | 西安电子科技大学 | Laser point cloud super-resolution reconstruction method based on self-attention generation countermeasure network |
| CN112884886A (en) * | 2021-03-17 | 2021-06-01 | 南通大学 | Three-dimensional point cloud pipeline extraction and modeling method capable of self-adapting to search radius |
Non-Patent Citations (2)
| Title |
|---|
| DONGWOOK SHU等: "3D Point Cloud Generative Adversarial Network Based on Tree Structured Graph Convolutions", 《2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV)》 * |
| 庞月勇 等: "面向水文分析的公路微地形三维建模方法", 《地理与地理信息科学》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113822999A (en) * | 2021-09-22 | 2021-12-21 | 杭州趣村游文旅集团有限公司 | Building segmentation method of digital country three-dimensional model |
| CN113822999B (en) * | 2021-09-22 | 2024-02-23 | 杭州趣村游文旅集团有限公司 | Building segmentation method of digital rural three-dimensional model |
| CN114332366A (en) * | 2021-12-24 | 2022-04-12 | 西运才 | Digital city single house point cloud facade 3D feature extraction method |
| CN114004871A (en) * | 2022-01-04 | 2022-02-01 | 山东大学 | Point cloud registration method and system based on point cloud completion |
| CN114004871B (en) * | 2022-01-04 | 2022-04-15 | 山东大学 | Point cloud registration method and system based on point cloud completion |
| CN115393542A (en) * | 2022-07-13 | 2022-11-25 | 南京林业大学 | Generalized building three-dimensional geometric reconstruction method |
| CN115393542B (en) * | 2022-07-13 | 2023-04-18 | 南京林业大学 | Generalized building three-dimensional geometric reconstruction method |
| CN115115806A (en) * | 2022-08-30 | 2022-09-27 | 北京飞渡科技有限公司 | Roof parameterization reconstruction method and system based on single-element analysis |
| CN115564926A (en) * | 2022-12-06 | 2023-01-03 | 武汉大学 | Three-dimensional patch model construction method based on image building structure learning |
| CN115564926B (en) * | 2022-12-06 | 2023-03-10 | 武汉大学 | Three-dimensional patch model construction method based on image building structure learning |
| CN116933359A (en) * | 2023-06-26 | 2023-10-24 | 武汉峰岭科技有限公司 | Building complex roof modeling method and system |
| CN116933359B (en) * | 2023-06-26 | 2024-02-02 | 武汉峰岭科技有限公司 | Building complex roof modeling method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113313835B (en) | 2021-11-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113313835B (en) | Building roof automatic modeling method based on airborne LiDAR point cloud | |
| CN110120097B (en) | Semantic modeling method for airborne point cloud of large scene | |
| Sun et al. | Aerial 3D building detection and modeling from airborne LiDAR point clouds | |
| CN111815776A (en) | Three-dimensional building fine geometric reconstruction method integrating airborne and vehicle-mounted three-dimensional laser point clouds and streetscape images | |
| CN102004922B (en) | High-resolution remote sensing image plane extraction method based on skeleton characteristic | |
| CN103703490B (en) | For generation of the equipment of three-dimensional feature data and the method for generation of three-dimensional feature data | |
| CN110263111B (en) | Priori knowledge-based space-time monitoring method for land utilization/coverage information | |
| CN104574512A (en) | Multi-scale DEM (digital elevation model) construction method considering topographical semantic information | |
| CN115861619A (en) | Airborne LiDAR (light detection and ranging) urban point cloud semantic segmentation method and system of recursive residual double-attention kernel point convolution network | |
| CN112288667A (en) | Three-dimensional target detection method based on fusion of laser radar and camera | |
| CN111028335B (en) | Point cloud data block surface patch reconstruction method based on deep learning | |
| Alidoost et al. | Y-shaped convolutional neural network for 3d roof elements extraction to reconstruct building models from a single aerial image | |
| CN109242810A (en) | A kind of City expending remote sensing monitoring and cities and towns quality evaluating method | |
| Quackenbush et al. | Road extraction: A review of LiDAR-focused studies | |
| CN116129118B (en) | Urban scene laser LiDAR point cloud semantic segmentation method based on graph convolution | |
| CN117315169A (en) | Live-action three-dimensional model reconstruction method and system based on deep learning multi-view dense matching | |
| Wang et al. | Methods of As-is BIM reconstruction using point cloud data for existing buildings | |
| CN112488117B (en) | Point cloud analysis method based on direction-induced convolution | |
| Chen et al. | Ground 3D Object Reconstruction Based on Multi-View 3D Occupancy Network using Satellite Remote Sensing Image | |
| Xiang et al. | Integrating Inverse Photogrammetry and a Deep Learning–Based Point Cloud Segmentation Approach for Automated Generation of BIM Models | |
| Costa et al. | Three-Dimensional Reconstruction of Satellite images using Generative Adversarial Networks | |
| CN112633376A (en) | Point cloud data ground feature classification method and system based on deep learning and storage medium | |
| Fang et al. | Semantic supported urban change detection using ALS point clouds | |
| CN112365456A (en) | Transformer substation equipment classification method based on three-dimensional point cloud data | |
| Yao et al. | Semantic Segmentation of Indoor 3d Point Clouds by Joint Optimization of Geometric Features and Neural Networks |
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 |