CN112712596B - Dense matching point cloud building structured model fine reconstruction method - Google Patents
Dense matching point cloud building structured model fine reconstruction method Download PDFInfo
- Publication number
- CN112712596B CN112712596B CN202110330216.4A CN202110330216A CN112712596B CN 112712596 B CN112712596 B CN 112712596B CN 202110330216 A CN202110330216 A CN 202110330216A CN 112712596 B CN112712596 B CN 112712596B
- Authority
- CN
- China
- Prior art keywords
- data
- plane
- plane element
- standard
- primitive
- 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.)
- Active
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
- 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
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/757—Matching configurations of points or features
Abstract
The invention discloses a dense matching point cloud building structural model fine reconstruction method which comprises the steps of obtaining original point cloud data, generating standard plane element data according to the original point cloud data, then generating candidate plane element data according to the standard plane element data, screening target plane element data from the candidate plane element data, and finally constructing a three-dimensional building model based on the target plane element data. Because the standard plane element data have regular relative position relation, the invention can effectively solve the problem that the relative position relation between the initial plane element data directly extracted from the original point cloud data has deviation in the prior art, and therefore, candidate plane elements generated by intersecting the initial plane elements are not accurate, and the reconstructed three-dimensional model of the building has deviation.
Description
Technical Field
The invention relates to the field of three-dimensional building models, in particular to a dense matching point cloud building structured model fine reconstruction method.
Background
The existing building three-dimensional model method firstly extracts plane elements from point cloud data, and then obtains candidate plane elements for reconstructing a building three-dimensional model through intersection of the plane elements. However, the relative position relationship between the initial plane element data directly extracted from the original point cloud data is deviated, that is, the initial plane element data generally does not have a regular relative position relationship, so that candidate plane elements generated by intersecting the initial plane elements are not accurate, and thus the reconstructed three-dimensional model of the building is deviated.
Thus, there is still a need for improvement and development of the prior art.
Disclosure of Invention
The invention aims to solve the technical problem that a dense matching point cloud building structured model fine reconstruction method is provided aiming at the defects in the prior art, and the method aims to solve the problem that in the prior art, the relative position relation between initial plane element data directly extracted from original point cloud data has deviation, so that candidate plane elements generated by intersecting the initial plane elements are not accurate, and the reconstructed building three-dimensional model has deviation.
The technical scheme adopted by the invention for solving the problems is as follows:
in a first aspect, an embodiment of the present invention provides a dense matching point cloud building structured model fine reconstruction method, where the method includes:
acquiring original point cloud data, and generating standard plane element data according to the original point cloud data;
generating candidate plane element data according to the standard plane element data;
and screening target plane element data from the candidate plane element data, and generating a three-dimensional building model according to the target plane element data.
In one embodiment, the obtaining raw point cloud data from which standard planar primitive data is generated comprises:
acquiring original point cloud data, and extracting a plurality of plane elementary data from the original point cloud data according to a preset algorithm;
and obtaining relative position relation information among the plurality of plane element data, and adjusting the relative position relation information to obtain a plurality of standard plane element data with standard relative position relation.
In one embodiment, the obtaining of the relative position relationship information between the planar primitive data and adjusting the relative position relationship information to obtain the standard planar primitive data with the standard relative position relationship includes:
acquiring a normal vector parameter of point cloud data contained in each of the plurality of planar element data, and acquiring intersection angle data among the plurality of planar element data according to the normal vector parameter;
acquiring relative position relationship information among the plurality of plane element data according to the intersection angle data, and taking the plane element data with the same relative position relationship as element cluster data to obtain a plurality of element cluster data;
taking the number of point clouds contained in each plane element data as a weight value, and calculating weighted average direction angle data corresponding to each element cluster data in the plurality of element cluster data according to the weight value and the intersection angle data;
and performing projection conversion on the point cloud data in the element cluster data according to the weighted average direction angle data to obtain a plurality of standard plane element data with standard relative position relation.
In one embodiment, the generating candidate plane primitive data from the standard plane primitive data comprises:
acquiring boundary straight line data corresponding to each plane element data in the standard plane element data;
acquiring distance information and projection data from each piece of boundary straight line data to each piece of plane element data in the standard plane element data;
determining adjacency relation information between the boundary straight line data and the plane element data according to the distance information and the projection data;
and determining candidate plane element data according to the adjacency relation information and the standard plane element data.
In one embodiment, the determining candidate flat primitive data from the adjacency information and the standard flat primitive data comprises:
acquiring a preset distance threshold, screening basic boundary linear data and basic plane primitive data from the standard plane primitive data and the boundary linear data according to the adjacency relation information and the distance threshold, and constructing newly added plane primitive data according to the basic boundary linear data and the basic plane primitive data;
and taking the standard plane element data and the newly added plane element data as candidate plane element data.
In one embodiment, the base boundary straight line data includes: first boundary line data; the base plane primitive data includes: first standard plane primitive data and second standard plane primitive data; the newly added plane primitive data includes: first newly-added plane primitive data; the acquiring of the preset distance threshold, screening out basic boundary linear data and basic plane primitive data from the standard plane primitive data and the boundary linear data according to the adjacency relation information and the distance threshold, and constructing newly added plane primitive data according to the basic boundary linear data and the basic plane primitive data comprises:
acquiring a preset distance threshold, and screening out first boundary straight line data, first standard plane element data and second standard plane element data from the standard plane element data and the boundary straight line data according to the adjacency relation information and the distance threshold; the first boundary line data is located within the second standard plane primitive data; the first boundary straight line data and the first standard plane primitive data are in an adjacent relation and are separated by a distance smaller than the distance threshold;
generating first newly-added plane element data according to the first boundary straight line data and the second standard plane element data; the first newly added plane element data comprises the first boundary straight line data and is in a mutually perpendicular relation with the second standard plane element data.
In one embodiment, the base boundary straight line data further includes: second boundary line data; the base plane primitive data further includes: third standard plane element data and fourth standard plane element data; the newly added planar primitive data further includes: second newly-added planar primitive data; the acquiring a preset distance threshold, screening basic boundary linear data and basic plane primitive data from the standard plane primitive data and the boundary linear data according to the adjacency relation information and the distance threshold, and constructing newly-added plane primitive data according to the basic boundary linear data and the basic plane primitive data further comprises:
screening out second boundary straight line data, third standard plane element data and fourth standard plane element data from the standard plane element data and the boundary straight line data according to the adjacency relation information and the distance threshold; the second boundary straight line data is positioned in the third standard plane element data, and the second boundary straight line data and the fourth standard plane element data are in an adjacent relation and a mutual parallel relation; the third standard plane element data and the fourth standard plane element data are in a mutually vertical relation;
generating second newly-added plane element data according to the second boundary straight line data, the third standard plane element data and the fourth standard plane element data; the second newly-added planar primitive data includes end point data of the second boundary straight line data, and is in a mutually perpendicular relationship with the third standard planar primitive data and the fourth standard planar primitive data.
In one embodiment, the screening out target planar primitive data from the candidate planar primitive data and generating a three-dimensional building model from the target planar primitive data comprises:
acquiring internal and external orientation element information of image data corresponding to the original point cloud data and empty three-point cloud data corresponding to the image data; the empty three-point cloud data is point cloud data obtained after aerial triangulation is carried out on the image data;
screening candidate elementary data in the candidate plane elementary data according to the internal and external orientation element information and the empty three-point cloud data to obtain target plane elementary data;
and generating a three-dimensional building model according to the target plane element data.
In one embodiment, the screening candidate primitive data in the candidate plane primitive data according to the inside and outside orientation element information and the empty three-point cloud data to obtain target plane primitive data includes:
acquiring camera projection center data according to the internal and external orientation element information;
acquiring three-dimensional point set data of the empty three-point cloud data, and taking the three-dimensional point set data as object space three-dimensional point data corresponding to the empty three-point cloud data;
connecting the object space three-dimensional point data with the camera projection center data to obtain line segment data;
acquiring intersection information of the line segment data and candidate primitive data in the candidate plane primitive data, and obtaining the number of intersection points based on the intersection information;
and screening candidate primitive data in the candidate plane primitive data according to the intersection point number to obtain target plane primitive data.
In one embodiment, the screening candidate planar primitive data from the candidate planar primitive data according to the number of intersections to obtain target planar primitive data includes:
when the number of the intersection points is 0, taking candidate primitive data corresponding to the number of the intersection points as candidate plane primitive data;
when the number of the intersection points is larger than 0, obtaining a distance value from the object space three-dimensional point data to the camera projection center data to obtain a first distance value;
obtaining a distance value from the intersection point to the camera projection center data to obtain a second distance value, and subtracting the first distance value from the second distance value to obtain a distance difference value;
when the distance difference value is smaller than or equal to a preset distance difference threshold value, taking candidate primitive data corresponding to the intersection point number as candidate plane primitive data;
acquiring fitting information of point cloud data in the candidate plane element data, fitting information of boundaries of the candidate plane element data, direction angle data of the boundaries of the candidate plane element data and topological constraint relation information among the candidate plane element data to obtain evaluation index data;
and substituting the evaluation index data into a preset energy equation to obtain a minimum energy solution output by the energy equation based on the evaluation index, and taking the candidate plane element data corresponding to the minimum energy solution as target plane element data.
The invention has the beneficial effects that: according to the embodiment of the invention, the original point cloud data is obtained, the standard plane element data is generated according to the original point cloud data, then the candidate plane element data is generated according to the standard plane element data, the target plane element data is screened out from the candidate plane element data, and finally the three-dimensional building model is constructed based on the target plane element data. Because the standard plane element data have regular relative position relation, the invention can effectively solve the problem that the relative position relation between the initial plane element data directly extracted from the original point cloud data has deviation in the prior art, and therefore, candidate plane elements generated by intersecting the initial plane elements are not accurate, and the reconstructed three-dimensional model of the building has deviation.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flow chart of a dense matching point cloud building structured model fine reconstruction method according to an embodiment of the present invention.
Fig. 2 is a detailed step schematic diagram of a dense matching point cloud building structured model fine reconstruction method provided by an embodiment of the invention.
Fig. 3 is a schematic diagram of generating a first primitive of a newly added plane according to an embodiment of the present invention.
Fig. 4 is a schematic diagram of generating a second newly added planar primitive provided by the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer and clearer, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
It should be noted that, if directional indications (such as up, down, left, right, front, and back … …) are involved in the embodiment of the present invention, the directional indications are only used to explain the relative positional relationship between the components, the movement situation, and the like in a specific posture (as shown in the drawing), and if the specific posture is changed, the directional indications are changed accordingly.
The three-dimensional fine model of the building is a single building reconstruction method based on point cloud data, and a structured polygonal model corresponding to the single building is recovered from a discrete three-dimensional point set on the surface of the building. The building model is an indispensable component for the construction of a novel smart city, and provides key basic information for various applications such as city management, planning, simulation, safety, emergency response and the like. The existing building three-dimensional model method firstly extracts initial plane elements from point cloud data, and then obtains candidate plane elements for reconstructing a building three-dimensional model through intersection of the initial plane elements. However, the relative position relationship between the initial plane element data directly extracted from the original point cloud data is deviated, that is, the initial plane element data generally does not have a regular relative position relationship, so that candidate plane elements generated by intersecting the initial plane elements are not accurate, and thus the reconstructed three-dimensional model of the building is deviated.
Aiming at the defects in the prior art, the invention provides a dense matching point cloud building structured model fine reconstruction method. Firstly, original point cloud data is obtained, and standard plane element data is generated according to the original point cloud data. Candidate plane primitive data is then generated from the standard plane primitive data. And finally, screening out target plane element data from the candidate plane element data, and generating a three-dimensional building model according to the target plane element data. In short, after the original point cloud data is acquired, standard plane element data is generated according to the original point cloud data, then candidate plane element data is generated according to the standard plane element data, target plane element data is screened out from the candidate plane element data, and finally a three-dimensional building model is constructed based on the target plane element data. Because the standard plane element data have regular relative position relation, the invention can effectively solve the problem that the relative position relation between the initial plane element data directly extracted from the original point cloud data has deviation in the prior art, and therefore, candidate plane elements generated by intersecting the initial plane elements are not accurate, and the reconstructed three-dimensional model of the building has deviation.
As shown in fig. 1, the method comprises the steps of:
s100, acquiring original point cloud data, and generating standard plane elementary data according to the original point cloud data.
Specifically, in order to reconstruct a three-dimensional model of a building, the present embodiment needs to first acquire original point cloud data obtained by performing laser scanning on the building. In order to ensure that correct candidate plane element data are generated subsequently and further a fine three-dimensional building model is constructed, the present embodiment also needs to recover the regular relative position relationship between the plane elements existing in the original point cloud data to obtain standard plane element data.
In one implementation, the step S100 specifically includes the following steps:
step S110, acquiring original point cloud data, and extracting a plurality of plane elementary data from the original point cloud data according to a preset algorithm;
step S120, obtaining relative position relation information among the plurality of plane element data, and adjusting the relative position relation information to obtain a plurality of standard plane element data with standard relative position relation.
Specifically, in this embodiment, a plurality of plane primitive data are extracted from the original point cloud data according to a preset algorithm, for example, the original point cloud data may be extracted by using a random sampling consistency algorithm or a region growing algorithm. Then, relative position relationship information between the plurality of plane element data is obtained, and it can be understood that the extracted relative position relationship is biased, for example, the relative position relationship between two plane element data should be a mutual perpendicular relationship, but the included angle between two plane elements is not a regular 90 degrees. Therefore, this embodiment needs to adjust the relative position relationship information to obtain a plurality of standard plane primitive data with standard relative position relationship.
In one implementation, in order to obtain the standard plane element data, the present embodiment may obtain the normal vector parameter of the point cloud data included in each of the plurality of plane element data, obtain the intersection angle data between the plurality of plane element data according to the normal vector parameter, and obtain the relative position relationship information between the plurality of plane element data according to the intersection angle data. For example, the present embodiment designs a plane equation for calculating point cloud data included in each plane element in advance:
wherein the content of the first and second substances,
is as follows
The plane equation parameters corresponding to each plane element,
are normal vector parameters.And also can calculate the gravity center position of each plane element
Number of dots
. Then, the intersection angle between each plane element is calculated through the normal vector parameters, and the elements with a certain threshold range tending to 0 degrees or 90 degrees are regarded as the plane elements which are mutually parallel (coplanar) or perpendicular, so that the relative position relation information between each plane element is obtained.
Then, in this embodiment, the planar primitive data with the same relative position relationship is used as a primitive cluster data to obtain a plurality of primitive cluster data, the number of point clouds included in each planar primitive data is used as a weight value, weighted average direction angle data corresponding to each primitive cluster data in the plurality of primitive cluster data is calculated according to the weight value and the intersection angle data, and finally, projection conversion is performed on the point cloud data in the primitive cluster data according to the weighted average direction angle data to obtain a plurality of standard planar primitive data with a standard relative position relationship. Specifically, in the conversion process, the center of gravity position of each plane element is taken as a pivot, and the normal vector of each plane element is rotated to the weighted average direction angle of the element cluster corresponding to the plane element, so that the plane elements of the building are regularized, and a standard plane element is obtained.
In one implementation, after the standard plane primitive is obtained, the boundary straight line of the plane primitive may be further regularized. Specifically, the original point cloud data is projected into the standard plane primitive data, then a preset algorithm (for example, an alpha-shape algorithm) is used for obtaining boundary points of each standard primitive plane primitive data in a 2D space, and then an existing boundary simplifying method (for example, a Douglas Puck method) is used for obtaining a simplified boundary straight line. After simplifying the boundary straight line, in order to further obtain the regularized boundary straight line, this embodiment further needs to abstract the boundary straight line into a central point and a direction angle, and convert the regularization problem of the boundary straight line into a labeling problem of the direction angle in the parameter space by using the following formula:
wherein the content of the first and second substances,
is composed of straight line segments
The set of boundary points of (a) is,
is the midpoint of the straight line segment,
is the direction angle. In order to improve the regularity of the boundary lines of the planar elements in the same building, the boundary lines should have the same direction angle as much as possible.
In one implementation, in order to make the boundary lines have the same direction angle, the present embodiment constructs a direction angle pool, and then selects the direction angle of each boundary line from the direction angle pool. For example, for a certain edge in a plane element of a certain building, the direction angle pool is composed of two parts: 1) the direction angle of other boundary lines in parallel (coplanar) planar primitives; 2) the direction angle of the projection of a non-parallel planar element in the planar element. The present embodiment sets the value range of the direction angle to be the range of the normal angle according to the perpendicular relation to the boundary straight line
The labels with the same direction angle have two expression modes of parallel or perpendicular. How to select the direction angle specifically is shown in the following first energy equation formula:
wherein, L represents the set of all straight line segments of the building, and L represents a boundary straight line formed by a plurality of optimized boundary points; s represents a boundary straight line pair with similar initial direction angles; k represents one element in S. The first term in the formula quantifies the degree of data fitting of the point set to each candidate direction angle through the vertical distance; the second punishment selects different direction angles from the straight line segments with similar direction angles. And finally, solving the direction angle of each straight-line segment by minimizing the first energy equation, and carrying out merging operation on continuous straight-line segments with the same direction angle to obtain the globally regularized building boundary straight line with higher data fidelity.
After the plane primitive data in the point cloud data is regularized and standard plane primitive data is obtained, as shown in fig. 1, the method further includes the following steps:
and step S200, generating candidate plane element data according to the standard plane element data.
Specifically, since there is a deviation in the relative positional relationship of primitive planes directly extracted from the original point cloud data, candidate plane primitive data generated from the intersections of these primitive planes also has a large error rate. In order to improve the accuracy of the three-dimensional model of the building, the present embodiment selects to generate candidate plane primitive data by using standard plane primitive data with regular relative position relationship.
In one implementation, the step S200 specifically includes the following steps:
step S210, acquiring boundary straight line data corresponding to each standard plane element data in the standard plane element data;
step S220, obtaining distance information and projection data from each piece of boundary straight line data to each piece of standard plane element data in the boundary straight line data;
step S230, determining adjacency relation information between each piece of boundary straight line data and each piece of standard plane elementary data according to the distance information and the projection data;
step S240, determining candidate plane element data according to the adjacency relation information and the standard plane element data.
Specifically, the present embodiment provides an algorithm for determining adjacency information between boundary straight line data and standard plane primitive data:
step 1: a boundary straight line l in a certain standard plane element A is selected, a boundary point set P forming the boundary straight line l is extracted, and the number n of adjacent points returns to zero;
step 2: arbitrarily taking a certain standard plane element B, and calculating a buffer plane area B + corresponding to the standard plane element B according to a preset range;
step 3: taking a certain point P in the boundary point set P, and calculating the vertical distance s from the point P to the plane of the standard plane primitive B;
step 4: if s is smaller than a given threshold value ts, entering Step 5; otherwise, returning to Step 3;
step 5: calculating the projection pB from the point p to the plane of the standard plane element B, and if the pB falls in a buffer plane area B +, increasing the number n of adjacent points by 1;
step 6; repeating Step 3-Step 5, if n exceeds a given threshold tn, the boundary straight line l is considered to be adjacent to the standard plane primitive B.
Step 7; repeating Step 1-Step 6 until all standard plane primitive data participate in judgment, and obtaining all adjacent boundary straight lines and standard plane primitive combinations, namely determining the adjacency relation information between the boundary straight line data and the standard plane primitives. Candidate flat primitive data is then determined from the adjacency information and the standard flat primitive data.
In one implementation, in order to determine candidate plane primitive data according to the adjacency information and the standard plane primitive data, a distance threshold is preset in this embodiment, then basic boundary straight line data and basic plane primitive data are screened out from the standard plane primitive data and the boundary straight line data according to the adjacency information and the distance threshold, and new plane primitive data are constructed according to the basic boundary straight line data and the basic plane primitive data.
In one implementation, the base boundary straight line data includes: first boundary line data; the base plane primitive data includes: first standard plane primitive data and second standard plane primitive data; the newly added plane primitive data includes: the first newly added plane primitive data. The present embodiment screens out first boundary straight line data, first standard plane primitive data, and second standard plane primitive data among the standard plane primitive data and the boundary straight line data according to the adjacency information and the distance threshold. Wherein the first boundary line data is located within the second standard plane primitive data, the first boundary line data and the first standard plane primitive data are in an adjacent relationship and are separated by a distance less than the distance threshold. And then generating first newly-added plane element data according to the first boundary straight line data and the second standard plane element data. Wherein the first newly added planar primitive data comprises the first boundary straight line data and is in a mutually perpendicular relationship with the second standard planar primitive data.
In short, the present embodiment provides a method of generating candidate planar primitive data that is different from the conventional method (conventional method: generating candidate planar primitive data based on intersection of planar primitive data). As shown in FIG. 3, if the boundary in the standard plane element A is straight
The distance between the standard plane primitive B and the boundary straight line is less than the distance threshold value
As a first boundary straight line, the standard plane primitive B is taken as a first standard plane primitive, the standard plane primitive A is taken as a second standard plane primitive, and then the straight line passing through the boundary is taken
And a plane Z perpendicular to the standard plane element a serves as a first added plane.
In one implementation, the base boundary straight line data further includes: second boundary line data; the base plane primitive data further includes: third standard plane element data and fourth standard plane element data; the newly added planar primitive data further includes: second newly added plane primitive data. The embodiment may further screen out second boundary straight line data, third standard plane primitive data, and fourth standard plane primitive data from the standard plane primitive data and the boundary straight line data according to the adjacency relation information and the distance threshold, where the second boundary straight line data is located in the third standard plane primitive data, and the second boundary straight line data and the fourth standard plane primitive data are in an adjacency relation and a parallel relation. The third standard plane element data and the fourth standard plane element data are in a mutually perpendicular relationship. Then, generating second newly-added planar primitive data according to the second boundary straight line data, the third standard planar primitive data and the fourth standard planar primitive data; the second newly-added planar primitive data includes end point data of the second boundary straight line data, and is in a mutually perpendicular relationship with the third standard planar primitive data and the fourth standard planar primitive data.
For example, as shown in FIG. 4, if there is a boundary line in the standard plane primitive C
Adjacent to the standard plane element D and satisfying the condition that the standard plane element C is perpendicular to the standard plane element D and the boundary straight line
Parallel to the standard plane element D, then a straight line will pass through the boundary
Planes with endpoints and perpendicular to both the standard plane primitive C, D
，
As a second newly added plane.
And finally, taking the standard plane element data and the newly added plane element data as candidate plane element data. In addition, the three types of plane element data can be sequentially intersected with the space bounding box where the original point cloud data is located, and then an element pool consisting of a plurality of bounded plane polygons and a common-edge relation among the polygons are obtained. In summary, because the original point cloud data inevitably has a deficiency, plane elements are directly extracted from the original point cloud data in the traditional method, and candidate plane element data generated by intersecting the plane elements easily has a deficiency, so that the fineness of a reconstructed building model is insufficient, details are lost, and the subsequent application value of the building three-dimensional model is greatly reduced. In addition to the original standard plane elements serving as candidate plane elements, some newly added planes participating in building three-dimensional models can be generated in the embodiment, and the newly added planes can compensate for the influence caused by the missing point cloud data to a certain extent.
To construct a three-dimensional building model, as shown in FIG. 1, the method further comprises the steps of:
s300, screening target plane element data from the candidate plane element data, and generating a three-dimensional building model according to the target plane element data.
Specifically, because there may be a large number of redundant plane primitives, and there may also be a small number of irregular plane primitives or plane primitives with incorrect topological relations in the candidate plane primitives, the present embodiment needs to screen the candidate plane primitives to obtain a target plane primitive with optimal geometry and topology, so as to construct an accurate three-dimensional building model.
In one implementation, the step S300 specifically includes the following steps:
step S310, obtaining internal and external orientation element information of image data corresponding to the original point cloud data and empty three-point cloud data corresponding to the image data; the empty three-point cloud data is point cloud data obtained after aerial triangulation is carried out on the image data;
step S320, screening candidate elementary data in the candidate plane elementary data according to the internal and external orientation element information and the empty three-point cloud data to obtain target plane elementary data;
and S330, generating a three-dimensional building model according to the target plane elementary data.
Specifically, the inside and outside orientation element information includes inside orientation element information and outside orientation element information, where the inside orientation element information is an optical center measurement coordinate that determines a projection center of the film and a camera focal length that determines a ratio of the film. The external orientation element information refers to the celestial coordinates of the optical center and the included angle between the y axis of the measurement coordinate system and the right ascension circle passing through the optical center. The aerial triangulation refers to a measurement method for encrypting control points indoors according to a small number of field control points in stereo photogrammetry to obtain the elevation and the plane position of the encrypted points. The primary purpose of aerial triangulation is to provide absolutely directed control points for mapping of regions lacking field control points. In the embodiment, the candidate primitive data in the candidate plane primitive data are screened through the internal and external orientation element information and the empty three-point cloud data, so that the plane primitives with depth conflicts obviously existing in the candidate plane primitives can be eliminated, redundant plane primitives are reduced, and the efficiency of subsequent operation is improved.
Specifically, in this embodiment, camera projection center data needs to be acquired according to the internal and external orientation element information, then three-dimensional point set data of the empty three-point cloud data is acquired, and the three-dimensional point set data is used as object three-dimensional point data corresponding to the empty three-point cloud data. And then, connecting the object space three-dimensional point data with the camera projection center data to obtain line segment data. Acquiring intersection information of the line segment data and candidate primitive data in the candidate plane primitive data, acquiring intersection point quantity based on the intersection information, and screening the candidate primitive data in the candidate plane primitive data according to the intersection point quantity to acquire target plane primitive data.
In one implementation, the specific process of screening candidate primitive data in the candidate plane primitive data according to the number of intersections to obtain target plane primitive data is as follows: and when the number of the intersection points is 0, taking the candidate primitive data corresponding to the number of the intersection points as candidate plane primitive data. And when the number of the intersection points is more than 0, obtaining a distance value from the object space three-dimensional point data to the camera projection center data to obtain a first distance value. And then obtaining a distance value from the intersection point to the camera projection center data to obtain a second distance value, and subtracting the first distance value from the second distance value to obtain a distance difference value. And when the distance difference value is smaller than or equal to a preset distance difference threshold value, taking the candidate primitive data corresponding to the intersection point number as candidate plane primitive data. In short, in order to construct an accurate three-dimensional building model, as shown in fig. 2, after removing plane primitives with depth conflicts obviously existing in the candidate plane primitives, the remaining plane primitives also need to be screened again as candidate plane primitives, and then the target plane primitive data is preferentially selected from the candidate plane primitives.
In an implementation manner, in this embodiment, it is required to obtain fitting information of point cloud data in the candidate plane element data, fitting information of a boundary of the candidate plane element data, direction angle data of the boundary of the candidate plane element data, and topological relation information between the candidate plane element data to obtain evaluation index data, then substitute the evaluation index data into a preset energy equation, obtain a minimum energy solution output by the energy equation based on the evaluation index, and use candidate plane element data corresponding to the minimum energy solution as target plane element data. And finally, generating a three-dimensional building model according to the target plane element data.
Specifically, in order to obtain fitting information of point cloud data in candidate plane element data, in this embodiment, original point cloud data included in a certain buffer range of each candidate plane element needs to be extracted, an average distance value from the point cloud data to the corresponding candidate plane element and a projection coverage ratio generated when the point cloud data is projected to the corresponding candidate plane element are counted, and the average distance value and the projection coverage ratio are used as the fitting information of the point cloud data in the candidate plane element data.
Specifically, in order to obtain the fitting information of the boundary of the candidate plane element data, the present embodiment needs to calculate the coincidence distance value and the straight line sag value between the intersection line of the candidate plane element and the feature line of the boundary of the adjacent building in the same direction, and take the coincidence distance value and the straight line sag value as the fitting information of the boundary of the candidate plane element data.
Specifically, in order to obtain the direction angle data of the boundary of the candidate plane element data, the present embodiment needs to calculate the direction angle of the intersection line of the candidate plane elements, group the intersection lines of the candidate plane elements according to the mutual perpendicular relationship, and count the cumulative length of the intersection lines in each group. It will be appreciated that the alternative planar element intersections have a length value, and thus the cumulative length is the sum of the length values of all intersections in the same group. Thereby achieving the purpose of punishing the appearance of redundant direction angles.
Specifically, in order to obtain the topology constraint relationship information between the candidate plane primitive data, the present embodiment needs to constrain the number of plane primitives connected by the same intersection line in the candidate plane primitives, avoid the topology conflict between the finally generated target plane primitives, and generate the topology constraint relationship information between the candidate plane primitive data according to the constraint relationship.
Finally, in the embodiment, the fitting information of point cloud data in the candidate plane element data, the fitting information of the boundary of the candidate plane element data, the direction angle data of the boundary of the candidate plane element data, and the topological constraint relation information between the candidate plane element data are used as evaluation index data, the evaluation index data are substituted into a preset energy equation, a minimum energy solution output by the energy equation based on the evaluation index is obtained, and the candidate plane element data corresponding to the minimum energy solution are used as target plane element data.
The formula of the energy equation employed in this embodiment is as follows:
wherein the content of the first and second substances,
fitting information for point cloud data in the candidate plane primitive data,
as fitting information of the boundary of the alternative plane primitive data,
for topology constraint relationship information between alternative plane primitive data,
direction angle data which is a boundary of the candidate plane primitive data. And calculating the minimum energy solution of the energy equation by a binary marking method to determine the target plane elementary data.
In summary, the invention discloses a dense matching point cloud building structured model fine reconstruction method, which comprises the steps of obtaining original point cloud data, generating standard plane element data according to the original point cloud data, then generating candidate plane element data according to the standard plane element data, screening target plane element data from the candidate plane element data, and finally constructing a three-dimensional building model based on the target plane element data. Because the standard plane element data have regular relative position relation, the invention can effectively solve the problem that the relative position relation between the initial plane element data directly extracted from the original point cloud data has deviation in the prior art, and therefore, candidate plane elements generated by intersecting the initial plane elements are not accurate, and the reconstructed three-dimensional model of the building has deviation.
It is to be understood that the invention is not limited to the examples described above, but that modifications and variations may be effected thereto by those of ordinary skill in the art in light of the foregoing description, and that all such modifications and variations are intended to be within the scope of the invention as defined by the appended claims.
Claims (8)
1. A dense matching point cloud building structured model fine reconstruction method is characterized by comprising the following steps:
acquiring original point cloud data, and generating standard plane element data according to the original point cloud data;
generating candidate plane element data according to the standard plane element data;
screening target plane element data from the candidate plane element data, and generating a three-dimensional building model according to the target plane element data;
the acquiring of the original point cloud data and the generating of the standard plane element data according to the original point cloud data comprise:
acquiring original point cloud data, and extracting a plurality of plane elementary data from the original point cloud data according to a preset algorithm;
acquiring relative position relation information among the plurality of plane element data, and adjusting the relative position relation information to obtain a plurality of standard plane element data with standard relative position relation;
the obtaining of the relative position relationship information between the plurality of planar primitive data and the adjusting of the relative position relationship information to obtain a plurality of standard planar primitive data having a standard relative position relationship includes:
acquiring a normal vector parameter of point cloud data contained in each of the plurality of planar element data, and acquiring intersection angle data among the plurality of planar element data according to the normal vector parameter;
acquiring relative position relationship information among the plurality of plane element data according to the intersection angle data, and taking the plane element data with the same relative position relationship as element cluster data to obtain a plurality of element cluster data;
taking the number of point clouds contained in each plane element data as a weight value, and calculating weighted average direction angle data corresponding to each element cluster data in the plurality of element cluster data according to the weight value and the intersection angle data;
and performing projection conversion on the point cloud data in the element cluster data according to the weighted average direction angle data to obtain a plurality of standard plane element data with standard relative position relation.
2. The method for fine reconstruction of dense matching point cloud building structural model according to claim 1, wherein the generating candidate plane element data according to the standard plane element data comprises:
acquiring boundary straight line data corresponding to each plane element data in the standard plane element data;
acquiring distance information and projection data from each piece of boundary straight line data to each piece of plane element data in the standard plane element data;
determining adjacency relation information between the boundary straight line data and the plane element data according to the distance information and the projection data;
and determining candidate plane element data according to the adjacency relation information and the standard plane element data.
3. The method of claim 2, wherein the determining candidate plane primitive data according to the adjacency information and the standard plane primitive data comprises:
acquiring a preset distance threshold, screening basic boundary linear data and basic plane primitive data from the standard plane primitive data and the boundary linear data according to the adjacency relation information and the distance threshold, and constructing newly added plane primitive data according to the basic boundary linear data and the basic plane primitive data;
and taking the standard plane element data and the newly added plane element data as candidate plane element data.
4. The method for finely reconstructing the dense matching point cloud building structural model according to claim 3, wherein the basic boundary straight line data comprises: first boundary line data; the base plane primitive data includes: first standard plane primitive data and second standard plane primitive data; the newly added plane primitive data includes: first newly-added plane primitive data; the acquiring of the preset distance threshold, screening out basic boundary linear data and basic plane primitive data from the standard plane primitive data and the boundary linear data according to the adjacency relation information and the distance threshold, and constructing newly added plane primitive data according to the basic boundary linear data and the basic plane primitive data comprises:
acquiring a preset distance threshold, and screening out first boundary straight line data, first standard plane element data and second standard plane element data from the standard plane element data and the boundary straight line data according to the adjacency relation information and the distance threshold; the first boundary line data is located within the second standard plane primitive data; the first boundary straight line data and the first standard plane primitive data are in an adjacent relation and are separated by a distance smaller than the distance threshold;
generating first newly-added plane element data according to the first boundary straight line data and the second standard plane element data; the first newly added plane element data comprises the first boundary straight line data and is in a mutually perpendicular relation with the second standard plane element data.
5. The method for finely reconstructing the dense matching point cloud building structural model according to claim 4, wherein the base boundary straight line data further comprises: second boundary line data; the base plane primitive data further includes: third standard plane element data and fourth standard plane element data; the newly added planar primitive data further includes: second newly-added planar primitive data; the acquiring a preset distance threshold, screening basic boundary linear data and basic plane primitive data from the standard plane primitive data and the boundary linear data according to the adjacency relation information and the distance threshold, and constructing newly-added plane primitive data according to the basic boundary linear data and the basic plane primitive data further comprises:
screening out second boundary straight line data, third standard plane element data and fourth standard plane element data from the standard plane element data and the boundary straight line data according to the adjacency relation information and the distance threshold; the second boundary straight line data is positioned in the third standard plane element data, and the second boundary straight line data and the fourth standard plane element data are in an adjacent relation and a mutual parallel relation; the third standard plane element data and the fourth standard plane element data are in a mutually vertical relation;
generating second newly-added plane element data according to the second boundary straight line data, the third standard plane element data and the fourth standard plane element data; the second newly-added planar primitive data includes end point data of the second boundary straight line data, and is in a mutually perpendicular relationship with the third standard planar primitive data and the fourth standard planar primitive data.
6. The method for finely reconstructing the dense matching point cloud building structural model according to claim 1, wherein the step of screening out target plane element data from the candidate plane element data and generating the three-dimensional building model according to the target plane element data comprises:
acquiring internal and external orientation element information of image data corresponding to the original point cloud data and empty three-point cloud data corresponding to the image data; the empty three-point cloud data is point cloud data obtained after aerial triangulation is carried out on the image data;
screening candidate elementary data in the candidate plane elementary data according to the internal and external orientation element information and the empty three-point cloud data to obtain target plane elementary data;
and generating a three-dimensional building model according to the target plane element data.
7. The method as claimed in claim 6, wherein the step of screening candidate element data in the candidate plane element data according to the inside and outside orientation element information and the empty three-point cloud data to obtain target plane element data comprises:
acquiring camera projection center data according to the internal and external orientation element information;
acquiring three-dimensional point set data of the empty three-point cloud data, and taking the three-dimensional point set data as object space three-dimensional point data corresponding to the empty three-point cloud data;
connecting the object space three-dimensional point data with the camera projection center data to obtain line segment data;
acquiring intersection information of the line segment data and candidate primitive data in the candidate plane primitive data, and obtaining the number of intersection points based on the intersection information;
and screening candidate primitive data in the candidate plane primitive data according to the intersection point number to obtain target plane primitive data.
8. The method of claim 7, wherein the step of screening candidate plane element data from the candidate plane element data according to the number of intersection points to obtain target plane element data comprises:
when the number of the intersection points is 0, taking candidate primitive data corresponding to the number of the intersection points as candidate plane primitive data;
when the number of the intersection points is larger than 0, obtaining a distance value from the object space three-dimensional point data to the camera projection center data to obtain a first distance value;
obtaining a distance value from the intersection point to the camera projection center data to obtain a second distance value, and subtracting the first distance value from the second distance value to obtain a distance difference value;
when the distance difference value is smaller than or equal to a preset distance difference threshold value, taking candidate primitive data corresponding to the intersection point number as candidate plane primitive data;
acquiring fitting information of point cloud data in the candidate plane element data, fitting information of boundaries of the candidate plane element data, direction angle data of the boundaries of the candidate plane element data and topological constraint relation information among the candidate plane element data to obtain evaluation index data;
and substituting the evaluation index data into a preset energy equation to obtain a minimum energy solution output by the energy equation based on the evaluation index, and taking the candidate plane element data corresponding to the minimum energy solution as target plane element data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110330216.4A CN112712596B (en) | 2021-03-29 | 2021-03-29 | Dense matching point cloud building structured model fine reconstruction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110330216.4A CN112712596B (en) | 2021-03-29 | 2021-03-29 | Dense matching point cloud building structured model fine reconstruction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112712596A CN112712596A (en) | 2021-04-27 |
| CN112712596B true CN112712596B (en) | 2021-07-06 |
Family
ID=75550390
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110330216.4A Active CN112712596B (en) | 2021-03-29 | 2021-03-29 | Dense matching point cloud building structured model fine reconstruction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112712596B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023164933A1 (en) * | 2022-03-04 | 2023-09-07 | 华为技术有限公司 | Building modeling method and related apparatus |
| CN114693879B (en) * | 2022-03-16 | 2023-06-13 | 哈尔滨理工大学 | Auxiliary alignment method for large-sized forging part based on three-dimensional reconstruction technology |
| CN114821013B (en) * | 2022-07-01 | 2022-10-18 | 深圳大学 | Element detection method and device based on point cloud data and computer equipment |
| CN116129076B (en) * | 2023-04-17 | 2023-06-27 | 深圳大学 | Building Mesh model simplification method with rule feature maintained |
| CN116258820B (en) * | 2023-05-15 | 2023-09-08 | 深圳大学 | Large-scale urban point cloud data set and building individuation construction method and related device |
| CN117315160A (en) * | 2023-10-31 | 2023-12-29 | 重庆市规划和自然资源信息中心 | Building three-dimensional live-action modeling working method |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6914601B2 (en) * | 2001-06-12 | 2005-07-05 | Minolta Co., Ltd. | Method, apparatus, and computer program for generating three-dimensional shape data or volume data |
| CN102855663B (en) * | 2012-05-04 | 2015-04-01 | 北京建筑工程学院 | Method for building CSG (Constructive Solid Geometry) model according to laser radar grid point cloud |
| CN108257213B (en) * | 2018-01-17 | 2021-08-24 | 视缘(上海)智能科技有限公司 | Point cloud lightweight polygonal curved surface reconstruction method |
| CN112070787B (en) * | 2020-08-10 | 2022-06-07 | 武汉大学 | Aviation three-dimensional point cloud plane segmentation method based on opponent reasoning theory |
| CN112381947B (en) * | 2021-01-18 | 2021-05-04 | 深圳大学 | Method and terminal for extracting three-dimensional change information of building and recognizing semantics |
| CN112444188B (en) * | 2021-02-01 | 2021-05-11 | 深圳大学 | Multi-view InSAR sea wall high-precision three-dimensional deformation measurement method |
-
2021
- 2021-03-29 CN CN202110330216.4A patent/CN112712596B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN112712596A (en) | 2021-04-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112712596B (en) | Dense matching point cloud building structured model fine reconstruction method | |
| CN107025685B (en) | Airborne building roof point cloud modeling method under topology perception | |
| US8711143B2 (en) | System and method for interactive image-based modeling of curved surfaces using single-view and multi-view feature curves | |
| US9942535B2 (en) | Method for 3D scene structure modeling and camera registration from single image | |
| CN108171780A (en) | A kind of method that indoor true three-dimension map is built based on laser radar | |
| WO2018061010A1 (en) | Point cloud transforming in large-scale urban modelling | |
| CN113916130B (en) | Building position measuring method based on least square method | |
| CN114066960B (en) | Three-dimensional reconstruction method, point cloud fusion method, device, equipment and storage medium | |
| US8600713B2 (en) | Method of online building-model reconstruction using photogrammetric mapping system | |
| CN105184857A (en) | Scale factor determination method in monocular vision reconstruction based on dot structured optical ranging | |
| CN106570905A (en) | Initial attitude verification method of non-cooperative target point cloud | |
| CN114332291A (en) | Oblique photography model building outer contour rule extraction method | |
| CN115564926B (en) | Three-dimensional patch model construction method based on image building structure learning | |
| CN107610219A (en) | The thick densification method of Pixel-level point cloud that geometry clue perceives in a kind of three-dimensional scenic reconstruct | |
| WO2023124676A1 (en) | 3d model construction method, apparatus, and electronic device | |
| CN113177974A (en) | Point cloud registration method and device, electronic equipment and storage medium | |
| CN114332125A (en) | Point cloud reconstruction method and device, electronic equipment and storage medium | |
| CN113393577B (en) | Oblique photography terrain reconstruction method | |
| CN115631317B (en) | Tunnel lining ortho-image generation method and device, storage medium and terminal | |
| CN115421509B (en) | Unmanned aerial vehicle flight shooting planning method, unmanned aerial vehicle flight shooting planning device and storage medium | |
| CN116704144A (en) | Live-action modeling method and system based on BIM collaborative design platform | |
| CN115496908A (en) | Automatic layering method and system for high-rise building oblique photography model | |
| CN113420460B (en) | Urban building height limit rapid analysis method and system based on OSG data astronomical line | |
| CN112991436B (en) | Monocular vision SLAM method based on object size prior information | |
| CN114119891A (en) | Three-dimensional reconstruction method and reconstruction system for robot monocular semi-dense map |
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 |