CN115661378A - Building model reconstruction method and system - Google Patents
Building model reconstruction method and system Download PDFInfo
- Publication number
- CN115661378A CN115661378A CN202211691743.9A CN202211691743A CN115661378A CN 115661378 A CN115661378 A CN 115661378A CN 202211691743 A CN202211691743 A CN 202211691743A CN 115661378 A CN115661378 A CN 115661378A
- Authority
- CN
- China
- Prior art keywords
- model data
- building
- dimensional model
- dimensional
- semantic object
- 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
Landscapes
- Processing Or Creating Images (AREA)
Abstract
The application discloses a building model reconstruction method and a building model reconstruction system, which are used for solving the technical problem of low model segmentation efficiency. According to the building model reconstruction scheme, the building model data are segmented from the earth surface three-dimensional model data by using the spatial distribution characteristics of the building model, and the segmentation efficiency is improved. Generating a projection view having a mapping relation with the building model data, and performing semantic segmentation on the projection view to generate a semantic object set; generating reconstructed building three-dimensional model data according to the three-dimensional parameters of the semantic object set; therefore, the complex three-dimensional model is segmented and reconstructed by semantic segmentation of the two-dimensional image, the intelligent degree is improved, and the segmentation calculated amount and the reconstruction time consumption are reduced.
Description
Technical Field
The application relates to the technical field of computers, in particular to a building model reconstruction method and a building model reconstruction system.
Background
The oblique photography technology is a high and new technology developed in the international surveying and mapping field in recent years, and earth surface data are automatically generated by mounting a plurality of sensors on the same flight platform and simultaneously acquiring images from five different angles, namely one vertical angle, four oblique angles and the like.
In the process of realizing the prior art, the inventor finds that:
oblique photography techniques export everything on the surface as a whole model. The integral model contains a large amount of invalid information, such as useless soil slopes and garbage can be also included in the integral model for description. However, if the target building model is expected to be segmented by adopting the semantics, the overall data of the model is huge and contains more information dimensions, so that the calculation amount and time consumption for semantic segmentation are large, and the accuracy of semantic segmentation is difficult to ensure.
Therefore, it is necessary to provide a new building model reconstruction scheme to solve the technical problem of low model segmentation efficiency.
Disclosure of Invention
The embodiment of the application provides a new building model reconstruction scheme, which is used for solving the technical problem of low model segmentation efficiency.
Specifically, the building model reconstruction method comprises the following steps:
acquiring three-dimensional model data of the earth surface;
according to the spatial distribution characteristics, segmenting building model data from the three-dimensional model data of the earth surface;
generating a projection view having a mapping relation with the building model data according to the building model data;
performing semantic segmentation on the projection view to generate a semantic object set;
generating three-dimensional parameters of a semantic object set according to the mapping relation between the projection view and the building model data;
and generating reconstructed building three-dimensional model data according to the three-dimensional parameters of the semantic object set.
Further, generating data of a reconstructed building three-dimensional model according to the three-dimensional parameters of the semantic object set, specifically comprising:
constructing an intermediate three-dimensional model data set according to the three-dimensional parameters of the semantic object set;
according to a rare algorithm, carrying out duplication elimination optimization on the intermediate three-dimensional model data set to generate a duplication elimination three-dimensional model data set;
and according to the mapping relation between the projection view and the building model data, aggregating the duplicate-removed three-dimensional model data set to generate reconstructed building three-dimensional model data.
Further, according to the three-dimensional parameters of the semantic object set, constructing an intermediate three-dimensional model data set, specifically comprising:
identifying a first semantic object in the set of semantic objects;
according to the semantics of the first semantic object, a three-dimensional model corresponding to the semantics is retrieved from a preset model library;
and acquiring three-dimensional parameters of the first semantic object, adjusting the size of the three-dimensional model corresponding to the semantics, and generating first intermediate three-dimensional model data.
Further, the building three-dimensional model data is in a TIN format.
Further, the acquiring of the surface three-dimensional model data specifically includes:
and acquiring three-dimensional model data of the earth surface by adopting an oblique photography technology.
The embodiment of the application also provides a building model reconstruction system.
Specifically, a building model reconstruction system includes:
the space segmentation module is used for acquiring three-dimensional model data of the earth surface; the system is also used for segmenting the building model data from the earth surface three-dimensional model data according to the spatial distribution characteristics;
the semantic segmentation module is used for generating a projection view which has a mapping relation with the building model data according to the building model data; the system is also used for carrying out semantic segmentation on the projection view to generate a semantic object set;
the reconstruction module is used for generating three-dimensional parameters of the semantic object set according to the mapping relation between the projection view and the building model data; and generating data of the reconstructed building three-dimensional model according to the three-dimensional parameters of the semantic object set.
Further, the reconstruction module is configured to generate reconstructed building three-dimensional model data according to the three-dimensional parameters of the semantic object set, and specifically is configured to:
constructing an intermediate three-dimensional model data set according to the three-dimensional parameters of the semantic object set;
according to a rare algorithm, carrying out duplication elimination optimization on the intermediate three-dimensional model data set to generate a duplication elimination three-dimensional model data set;
and according to the mapping relation between the projection view and the building model data, aggregating the duplicate-removed three-dimensional model data set to generate reconstructed building three-dimensional model data.
Further, the reconstruction module is configured to construct an intermediate three-dimensional model data set according to the three-dimensional parameters of the semantic object set, and specifically configured to:
identifying a first semantic object in a set of semantic objects;
according to the semantics of the first semantic object, a three-dimensional model corresponding to the semantics is retrieved from a preset model library;
and acquiring three-dimensional parameters of the first semantic object, adjusting the size of the three-dimensional model corresponding to the semantics, and generating first intermediate three-dimensional model data.
Further, the building three-dimensional model data is in a TIN format.
Further, the space segmentation module is configured to obtain surface three-dimensional model data, and specifically configured to:
and acquiring three-dimensional model data of the earth surface by adopting an oblique photography technology.
The technical scheme provided by the embodiment of the application at least has the following beneficial effects:
the building model data are segmented from the earth surface three-dimensional model data by utilizing the spatial distribution characteristics of the building model, so that the segmentation efficiency is improved. Generating a projection view having a mapping relation with the building model data, and performing semantic segmentation on the projection view to generate a semantic object set; generating reconstructed building three-dimensional model data according to the three-dimensional parameters of the semantic object set; therefore, the complex three-dimensional model is segmented and reconstructed by semantic segmentation of the two-dimensional image, the intelligent degree is improved, and the segmentation calculated amount and the reconstruction time consumption are reduced.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
fig. 1 is a flowchart of a building model reconstruction method according to an embodiment of the present disclosure;
fig. 2 is a schematic structural diagram of a building model reconstruction system according to an embodiment of the present application.
100. Building model reconstruction system
11. Space division module
12. Semantic segmentation module
13. And a reconstruction module.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the technical solutions of the present application will be described in detail and completely with reference to the following specific embodiments of the present application and the accompanying drawings. It should be apparent that the described embodiments are only some of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Referring to fig. 1, to solve the technical problem of low efficiency of model segmentation, the present application provides a method for reconstructing a building model, which includes the following steps:
s110: and acquiring three-dimensional model data of the earth surface.
The three-dimensional model data of the earth surface can be understood as a virtual model describing the earth surface by a point cloud. In one embodiment provided herein, the three-dimensional model data of the earth's surface may be acquired using oblique photography. Preferably, the surface three-dimensional model data is in a TIN (Triangulated Irregular Network) format. The TIN is a form of vector-based digital geographic data, and is constructed by forming points into triangles, so that data redundancy can be reduced, and the operation efficiency can be improved.
S120: and segmenting the building model data from the three-dimensional model data of the earth surface according to the spatial distribution characteristics.
It is understood that the surface three-dimensional model data has spatially distributed features based on a spatial coordinate system. Taking the horizontal plane defined by the X axis and the Y axis as an example, the closer to the horizontal plane on the Z axis, the denser and more complex the model space distribution characteristics, and the farther from the horizontal plane on the Z axis, the simpler and more sparse the model space distribution characteristics.
On the basis that the three-dimensional model of the present application is a surface three-dimensional model, data closer to the ground (horizontal plane) is more redundant. This is because the oblique photography technique outputs all objects on the earth surface as an integral model, and a large amount of invalid information existing on the earth surface is also described in the integral model. While data further from the ground (horizontal plane) have less, more significant, description of the target structure.
Therefore, the building model data can be segmented from the surface three-dimensional model data according to the spatial distribution characteristics. The building model data may be understood as a building structure excluding a large amount of invalid information.
S130: and generating a projection view having a mapping relation with the building model data according to the building model data.
S140: and performing semantic segmentation on the projection view to generate a semantic object set.
It will be appreciated by those skilled in the art that the segmented building model data describes a building structure with a large amount of invalid information removed. The building structure comprises a building main body and a building attachment component.
In order to further improve the accuracy of model segmentation, the projection view with the mapping relation with the building model data is generated according to the building model data. In general, the projection view may have a projection of multiple facets. Different projection views may be considered to describe the building model data in different information dimensions.
Semantic segmentation is carried out on the projection view, and a semantic object set can be generated. The semantic segmentation is to identify a building main body and different building attachment components in the projection view. The elements of the semantic object set are semantic objects. On the basis that the attention object of the application is a building three-dimensional model, the semantic object can be represented as a building main body, a window, a balcony, a guardrail, an air conditioner and the like.
Of course, the different projection views have a mapping relationship with the building model data. The mapping relation is expressed in that the position relation of the semantic object in the projection view corresponds to the position relation of the semantic object in the building model one by one.
Therefore, the same semantic object is described by different information dimensions, the limitation of a single semantic object can be overcome, the accuracy of semantic segmentation is improved, and the accuracy of model segmentation is improved. On the other hand, the complex three-dimensional model is segmented by semantic segmentation of the two-dimensional image, so that the calculation amount and time consumption are reduced, and the processing efficiency is improved.
S150: and generating three-dimensional parameters of the semantic object set according to the mapping relation between the projection view and the building model data.
S160: and generating reconstructed building three-dimensional model data according to the three-dimensional parameters of the semantic object set.
It can be understood that when the semantic object represents a two-dimensional image (i.e. the semantic object in the projection view), two-dimensional parameters of the semantic object, such as any two of the length, width and height of the semantic object, can be obtained.
When the semantic object represents the three-dimensional model, the semantic object has three-dimensional parameters. And obtaining at least two projection views containing the same semantic object according to the mapping relation between the projection views and the building model data. According to the at least two projection views, the three-dimensional parameters of the semantic object can be acquired in a supplementary mode.
Three-dimensional model data corresponding to the semantic object may be generated, typically based on three-dimensional parameters of the semantic object. According to the three-dimensional parameters of the semantic object set, a three-dimensional model data set corresponding to the semantic object set can be generated. In other words, the elements in the three-dimensional model data set are three-dimensional model data of the corresponding semantic object.
And aggregating the three-dimensional model data set according to the mapping relation between the projection view and the building model data to generate the reconstructed building three-dimensional model data.
In view of the fact that repeated data still exist in the generated three-dimensional model data of the reconstructed building according to the three-dimensional parameters of the semantic object set, in order to further reduce the storage amount and improve the operation efficiency, in a specific embodiment provided by the present application, the generating of the three-dimensional model data of the reconstructed building according to the three-dimensional parameters of the semantic object set specifically includes:
constructing an intermediate three-dimensional model data set according to the three-dimensional parameters of the semantic object set;
according to a rare algorithm, carrying out duplication elimination optimization on the intermediate three-dimensional model data set to generate a duplication elimination three-dimensional model data set;
and according to the mapping relation between the projection view and the building model data, aggregating the duplicate-removed three-dimensional model data set to generate reconstructed building three-dimensional model data.
It will be appreciated that the three-dimensional model data generated from the three-dimensional parameters of the semantic object is considered in this embodiment as an intermediate three-dimensional model. And taking the generated three-dimensional model data set as an intermediate three-dimensional model data set according to the three-dimensional parameters of the semantic object set. In other words, the elements in the intermediate three-dimensional model data set are intermediate three-dimensional model data of the corresponding semantic object.
And then, carrying out duplication elimination optimization on the intermediate three-dimensional model data by adopting a rare algorithm to generate duplication elimination three-dimensional model data. The principal point of the extraction algorithm is to approximate the curve as a series of points and reduce the number of points. Therefore, the de-duplicated three-dimensional model data has better fidelity, and a large amount of data can be compressed under the condition of reflecting the basic shape characteristics of the original graph or curve.
And aggregating the duplication-removing three-dimensional model data subjected to duplication-removing optimization by adopting a Chirsh algorithm according to the mapping relation between the projection view and the building model data to generate the reconstructed building three-dimensional model data. Therefore, the storage capacity of the data of the three-dimensional model of the generated and reconstructed building is small, and the storage resources are saved.
Of course, when the semantic object is a building attachment member, most building attachment members have a low requirement for simulation. In order to improve modeling efficiency, in a specific embodiment provided in the present application, a middle three-dimensional model data set is constructed according to three-dimensional parameters of a semantic object set, which specifically includes:
identifying a first semantic object in a set of semantic objects;
according to the semantics of the first semantic object, a three-dimensional model corresponding to the semantics is retrieved from a preset model library;
and acquiring three-dimensional parameters of the first semantic object, adjusting the size of the three-dimensional model corresponding to the semantics, and generating first intermediate three-dimensional model data.
The preset model library stores the three-dimensional model data of the building attachment part and records the associated semantics of the three-dimensional model data of the building attachment part. In this way, the corresponding building attached component three-dimensional model data can be retrieved through the semantics of the semantic object. Specifically, taking a first semantic object in the semantic object set as an example, according to the semantics of the first semantic object, a three-dimensional model corresponding to the semantics can be retrieved from a preset model library. Further, the three-dimensional model corresponding to the semantics may adjust a size of the three-dimensional model corresponding to the semantics according to the three-dimensional parameters of the first semantic object. And taking the three-dimensional model after the size adjustment as first intermediate three-dimensional model data. Therefore, the calculation amount of the reconstruction model is reduced, and the reconstruction efficiency of the building model is improved.
The concrete implementation process of the building model reconstruction method is described as follows:
firstly, acquiring earth surface three-dimensional model data by adopting an oblique photography technology, and constructing the earth surface three-dimensional model data in a TIN format.
And then, dividing a plurality of building model data from the whole TIN grid model one by utilizing the spatial distribution characteristics of the building.
And performing multi-view rendering on the building model according to the building model data to generate a plurality of projection views which have a mapping relation with the building model data. And 2D AI semantic segmentation is carried out on each projection view to obtain a semantic object set with elements of semantic objects such as wall surfaces, windows, air conditioners and the like.
And generating three-dimensional parameters of a plurality of semantic objects according to the mapping relation between the projection view and the building model data. And assigning the three-dimensional parameters of the semantic objects to the building three-dimensional model, and reversely generating the TIN building three-dimensional model, namely generating the data of the reconstructed building three-dimensional model.
And (4) carrying out duplicate removal optimization on the TIN three-dimensional models corresponding to the same semantic objects by adopting a rare-extraction algorithm, so as to reduce the number of faces and generate a duplicate removal three-dimensional model.
And searching a universal three-dimensional model with corresponding semantics from a preset model library for the semantic object with the semantics as the building attachment component. And adjusting the size of the universal three-dimensional model according to the three-dimensional parameters of the corresponding building attachment parts.
And finally, aggregating the de-duplicated three-dimensional model data and the size-adjusted general three-dimensional model data to generate reconstructed building three-dimensional model data.
In summary, the building model reconstruction method provided by the application segments the building model data from the earth surface three-dimensional model data by using the spatial distribution characteristics of the building model, and improves the segmentation efficiency. Generating a projection view having a mapping relation with the building model data, and performing semantic segmentation on the projection view to generate a semantic object set; generating reconstructed building three-dimensional model data according to the three-dimensional parameters of the semantic object set; therefore, the complex three-dimensional model is segmented and reconstructed by semantic segmentation of the two-dimensional image, the intelligent degree is improved, and the segmentation calculated amount and reconstruction time consumption are reduced.
Referring to fig. 2, in order to support the building model reconstruction method, the present application further provides a building model reconstruction system 100, which includes:
the space segmentation module 11 is used for acquiring three-dimensional model data of the earth surface; the building model data is segmented from the earth surface three-dimensional model data according to the spatial distribution characteristics;
the semantic segmentation module 12 is used for generating a projection view having a mapping relation with the building model data according to the building model data; the system is also used for carrying out semantic segmentation on the projection view to generate a semantic object set;
the reconstruction module 13 is configured to generate three-dimensional parameters of the semantic object set according to a mapping relationship between the projection view and the building model data; and generating data of the reconstructed building three-dimensional model according to the three-dimensional parameters of the semantic object set.
The space segmentation module 11 acquires three-dimensional model data of the earth's surface.
The three-dimensional model data of the earth surface can be understood as a virtual model describing the earth surface by a point cloud. In one embodiment provided herein, the space segmentation module 11 may acquire the three-dimensional model data of the earth surface by using an oblique photography technique. Preferably, the surface three-dimensional model data is in a TIN (Triangulated Irregular Network) format. The TIN is a form of vector-based digital geographic data, and is constructed by forming points into triangles, so that data redundancy can be reduced, and the operation efficiency can be improved.
The space division module 11 divides the building model data from the earth surface three-dimensional model data according to the space distribution characteristics.
It is understood that the surface three-dimensional model data has spatially distributed features based on a spatial coordinate system. Taking the horizontal plane defined by the X axis and the Y axis as an example, the closer to the horizontal plane on the Z axis, the denser and more complex the model space distribution characteristics, and the farther from the horizontal plane on the Z axis, the simpler and more sparse the model space distribution characteristics.
On the basis that the three-dimensional model of the present application is a three-dimensional model of the earth's surface, data closer to the ground (horizontal plane) is more redundant. This is because the oblique photography technique outputs all objects on the earth surface as an integral model, and a large amount of invalid information existing on the earth surface is also described in the integral model. While data further from the ground (horizontal) have less invalid information and describe the target structure more predominantly.
Therefore, the space division module 11 can divide the building model data from the surface three-dimensional model data according to the space distribution characteristics. The building model data may be understood as a building structure excluding a large amount of invalid information.
The semantic segmentation module 12 generates a projection view having a mapping relation with the building model data according to the building model data; and performing semantic segmentation on the projection view to generate a semantic object set.
It will be appreciated by those skilled in the art that the building model data partitioned by the space partitioning module 11 describes a building structure from which a large amount of invalid information is removed. The building structure comprises a building main body and a building attachment component.
In order to further improve the accuracy of model segmentation, the semantic segmentation module 12 generates a projection view having a mapping relationship with the building model data according to the building model data. In general, the projection view may have a projection of multiple facets. Different projection views may be considered to describe the building model data in different information dimensions.
The semantic segmentation module 12 performs semantic segmentation on the projection view, and may generate a semantic object set. The semantic segmentation is to identify a building main body and different building attachment components in the projection view. The elements of the semantic object set are semantic objects. On the basis that the attention object of the application is a building three-dimensional model, the semantic object can be represented as a building main body, a window, a balcony, a guardrail, an air conditioner and the like.
Of course, the different projection views have a mapping relationship with the building model data. The mapping relation is expressed in that the position relation of the semantic object in the projection view corresponds to the position relation of the semantic object in the building model one by one.
Therefore, the same semantic object is described by different information dimensions, the limitation of a single semantic object can be overcome, the accuracy of semantic segmentation is improved, and the accuracy of model segmentation is improved. On the other hand, the complex three-dimensional model is segmented by semantic segmentation of the two-dimensional image, so that the calculation amount and time consumption are reduced, and the processing efficiency is improved.
The reconstruction module 13 generates three-dimensional parameters of a semantic object set according to the mapping relation between the projection view and the building model data; and generating reconstructed building three-dimensional model data according to the three-dimensional parameters of the semantic object set.
It is understood that when the semantic object represents a two-dimensional image (i.e. the semantic object in the projection view), the reconstruction module 13 can obtain two-dimensional parameters of the semantic object, for example, any two parameters of the length, the width and the height of the semantic object.
When a semantic object represents a three-dimensional model, the semantic object has three-dimensional parameters. The reconstruction module 13 may obtain at least two projection views containing the same semantic object according to the mapping relationship between the projection views and the building model data. The reconstruction module 13 may additionally obtain three-dimensional parameters of the semantic object according to the at least two projection views.
Generally, the reconstruction module 13 may generate three-dimensional model data corresponding to the semantic object according to the three-dimensional parameters of the semantic object. The reconstruction module 13 may generate a three-dimensional model data set corresponding to the semantic object set according to the three-dimensional parameters of the semantic object set. In other words, the elements in the three-dimensional model data set are three-dimensional model data of the corresponding semantic object.
The reconstruction module 13 aggregates the three-dimensional model data set according to the mapping relationship between the projection view and the building model data, and can generate reconstructed building three-dimensional model data.
Considering that the reconstruction module 13 still has repeated data in the generated reconstructed building three-dimensional model data according to the three-dimensional parameters of the semantic object set, in order to further reduce the storage amount and improve the operation efficiency, in a specific embodiment provided by the present application, the reconstruction module 13 generates the reconstructed building three-dimensional model data according to the three-dimensional parameters of the semantic object set, which specifically includes:
constructing an intermediate three-dimensional model data set according to the three-dimensional parameters of the semantic object set;
according to a rare algorithm, carrying out duplication elimination optimization on the intermediate three-dimensional model data set to generate a duplication elimination three-dimensional model data set;
and according to the mapping relation between the projection view and the building model data, aggregating the duplicate-removed three-dimensional model data set to generate reconstructed building three-dimensional model data.
It will be appreciated that in this embodiment the reconstruction module 13 treats the three-dimensional model data generated from the semantic object three-dimensional parameters as an intermediate three-dimensional model. The reconstruction module 13 regards the three-dimensional model data set generated from the three-dimensional parameters of the semantic object set as an intermediate three-dimensional model data set. In other words, the elements in the intermediate three-dimensional model data set are intermediate three-dimensional model data of the corresponding semantic object.
And then the reconstruction module 13 performs deduplication optimization on the intermediate three-dimensional model data by adopting a rare-extraction algorithm to generate deduplication three-dimensional model data. The principal point of the extraction algorithm is to approximate the curve as a series of points and reduce the number of points. Therefore, the de-duplicated three-dimensional model data has better fidelity, and a large amount of data can be compressed under the condition of reflecting the basic shape characteristics of the original graph or curve.
The reconstruction module 13 aggregates the duplication-elimination three-dimensional model data subjected to duplication elimination optimization by using the extraction-and-extraction algorithm according to the mapping relation between the projection view and the building model data to generate reconstructed building three-dimensional model data. Therefore, the storage capacity of the data of the three-dimensional model of the generated and reconstructed building is small, and the storage resources are saved.
Of course, when the semantic object is a building adhering component, most building adhering components have low requirements for simulation. In order to improve modeling efficiency, in a specific embodiment provided in the present application, the reconstructing module 13 constructs an intermediate three-dimensional model data set according to three-dimensional parameters of a semantic object set, which specifically includes:
identifying a first semantic object in a set of semantic objects;
according to the semantics of the first semantic object, a three-dimensional model corresponding to the semantics is retrieved from a preset model library;
and acquiring three-dimensional parameters of the first semantic object, adjusting the size of the three-dimensional model corresponding to the semantics, and generating first intermediate three-dimensional model data.
The preset model library stores the three-dimensional model data of the building attachment part and records the associated semantics of the three-dimensional model data of the building attachment part. In this way, the reconstruction module 13 can retrieve the corresponding three-dimensional model data of the building attachment part by the semantics of the semantic object. Specifically, taking a first semantic object in the semantic object set as an example, the reconstruction module 13 may retrieve a three-dimensional model corresponding to the semantic meaning from a preset model library according to the semantic meaning of the first semantic object. Further, the reconstruction module 13 may adjust the size of the three-dimensional model corresponding to the semantic meaning according to the three-dimensional parameter of the first semantic object. The reconstruction module 13 takes the resized three-dimensional model as first intermediate three-dimensional model data. Therefore, the calculation amount of the reconstruction model is reduced, and the reconstruction efficiency of the building model is improved.
The following describes a specific implementation of the building model reconstruction system 100:
firstly, the space segmentation module 11 acquires the three-dimensional model data of the earth surface by adopting an oblique photography technology, and constructs the three-dimensional model data of the earth surface in a TIN format.
And then the space division module 11 divides a plurality of building model data from the whole TIN grid model one by utilizing the space distribution characteristics of the building.
The semantic segmentation module 12 performs multi-view rendering on the building model according to the building model data to generate a plurality of projection views having a mapping relationship with the building model data. And 2D AI semantic segmentation is carried out on each projection view to obtain a semantic object set with elements of semantic objects such as wall surfaces, windows, air conditioners and the like.
The reconstruction module 13 generates three-dimensional parameters of a plurality of semantic objects according to the mapping relationship between the projection view and the building model data. The reconstruction module 13 assigns the three-dimensional parameters of the semantic object to the building three-dimensional model, and reversely generates a TIN building three-dimensional model, that is, generates data of the reconstructed building three-dimensional model.
The reconstruction module 13 performs deduplication optimization on the TIN three-dimensional models corresponding to the same semantic object by using a rare-extraction algorithm, so as to reduce the number of faces and generate a deduplication three-dimensional model.
The reconstruction module 13 searches a general three-dimensional model with corresponding semantics from a preset model library for a semantic object with semantics being a building attachment component. And adjusting the size of the universal three-dimensional model according to the three-dimensional parameters of the corresponding building attachment parts.
And the final reconstruction module 13 aggregates the de-duplicated three-dimensional model data and the size-adjusted general three-dimensional model data to generate reconstructed building three-dimensional model data.
To sum up, the building model reconstruction system 100 provided by the present application segments the building model data from the earth surface three-dimensional model data by using the spatial distribution characteristics of the building model, thereby improving the segmentation efficiency. Generating a projection view having a mapping relation with the building model data, and performing semantic segmentation on the projection view to generate a semantic object set; generating reconstructed building three-dimensional model data according to the three-dimensional parameters of the semantic object set; therefore, the complex three-dimensional model is segmented and reconstructed by semantic segmentation of the two-dimensional image, the intelligent degree is improved, and the segmentation calculated amount and reconstruction time consumption are reduced.
It is to be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, having an element defined by the phrase "comprising a … …" does not exclude the presence of another like element in a process, method, article, or apparatus that comprises the element.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The above description is only an example of the present application and is not intended to limit the present application. Various modifications and changes may occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims (10)
1. A method of building model reconstruction, comprising the steps of:
acquiring three-dimensional model data of the earth surface;
according to the spatial distribution characteristics, segmenting building model data from the three-dimensional model data of the earth surface;
generating a projection view having a mapping relation with the building model data according to the building model data;
performing semantic segmentation on the projection view to generate a semantic object set;
generating three-dimensional parameters of a semantic object set according to the mapping relation between the projection view and the building model data;
and generating reconstructed building three-dimensional model data according to the three-dimensional parameters of the semantic object set.
2. The building model reconstruction method according to claim 1, wherein generating reconstructed building three-dimensional model data from the three-dimensional parameters of the semantic object sets specifically comprises:
constructing an intermediate three-dimensional model data set according to the three-dimensional parameters of the semantic object set;
according to a rare algorithm, carrying out duplication elimination optimization on the intermediate three-dimensional model data set to generate a duplication elimination three-dimensional model data set;
and according to the mapping relation between the projection view and the building model data, aggregating the duplicate-removed three-dimensional model data set to generate reconstructed building three-dimensional model data.
3. The building model reconstruction method according to claim 2, wherein constructing the intermediate three-dimensional model data set from the three-dimensional parameters of the semantic object sets comprises:
identifying a first semantic object in a set of semantic objects;
according to the semantics of the first semantic object, a three-dimensional model corresponding to the semantics is retrieved from a preset model library;
and acquiring three-dimensional parameters of the first semantic object, adjusting the size of the three-dimensional model corresponding to the semantics, and generating first intermediate three-dimensional model data.
4. The building model reconstruction method according to claim 1, wherein the building three-dimensional model data is in a TIN format.
5. The building model reconstruction method according to claim 1, wherein the acquiring of the surface three-dimensional model data specifically comprises:
and acquiring three-dimensional model data of the earth surface by adopting an oblique photography technology.
6. A building model reconstruction system, comprising:
the space segmentation module is used for acquiring three-dimensional model data of the earth surface; the building model data is segmented from the earth surface three-dimensional model data according to the spatial distribution characteristics;
the semantic segmentation module is used for generating a projection view which has a mapping relation with the building model data according to the building model data; the system is also used for carrying out semantic segmentation on the projection view to generate a semantic object set;
the reconstruction module is used for generating three-dimensional parameters of the semantic object set according to the mapping relation between the projection view and the building model data; and generating data of the reconstructed building three-dimensional model according to the three-dimensional parameters of the semantic object set.
7. The building model reconstruction system of claim 6, wherein the reconstruction module is configured to generate reconstructed building three-dimensional model data from the three-dimensional parameters of the semantic object sets, and is specifically configured to:
constructing an intermediate three-dimensional model data set according to the three-dimensional parameters of the semantic object set;
according to a rare algorithm, carrying out duplication elimination optimization on the intermediate three-dimensional model data set to generate a duplication elimination three-dimensional model data set;
and according to the mapping relation between the projection view and the building model data, aggregating the duplicate-removed three-dimensional model data set to generate reconstructed building three-dimensional model data.
8. The building model reconstruction system of claim 7, wherein the reconstruction module is configured to construct the intermediate three-dimensional model data set based on three-dimensional parameters of the semantic object sets, and in particular to:
identifying a first semantic object in a set of semantic objects;
according to the semantics of the first semantic object, a three-dimensional model corresponding to the semantics is retrieved from a preset model library;
and acquiring three-dimensional parameters of the first semantic object, adjusting the size of the three-dimensional model corresponding to the semantics, and generating first intermediate three-dimensional model data.
9. The building model reconstruction system of claim 6 wherein said building three dimensional model data is in TIN format.
10. The building model reconstruction system of claim 6, wherein the spatial segmentation module is configured to obtain surface three-dimensional model data, and is specifically configured to:
and acquiring three-dimensional model data of the earth surface by adopting an oblique photography technology.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211691743.9A CN115661378B (en) | 2022-12-28 | 2022-12-28 | Building model reconstruction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211691743.9A CN115661378B (en) | 2022-12-28 | 2022-12-28 | Building model reconstruction method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115661378A true CN115661378A (en) | 2023-01-31 |
| CN115661378B CN115661378B (en) | 2023-03-21 |
Family
ID=85023511
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211691743.9A Active CN115661378B (en) | 2022-12-28 | 2022-12-28 | Building model reconstruction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115661378B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150178988A1 (en) * | 2012-05-22 | 2015-06-25 | Telefonica, S.A. | Method and a system for generating a realistic 3d reconstruction model for an object or being |
| CN108389251A (en) * | 2018-03-21 | 2018-08-10 | 南京大学 | The full convolutional network threedimensional model dividing method of projection based on fusion various visual angles feature |
| US20190197311A1 (en) * | 2017-12-26 | 2019-06-27 | Harbin Institute Of Technology | Evaluation Method of Solar Energy Utilization Potential in Urban High-density Areas Based on Low-altitude Photogrammetry |
| CN110009727A (en) * | 2019-03-08 | 2019-07-12 | 深圳大学 | A kind of indoor threedimensional model automatic reconfiguration method and system with structure semantics |
| CN111815776A (en) * | 2020-02-04 | 2020-10-23 | 山东水利技师学院 | Three-dimensional building fine geometric reconstruction method integrating airborne and vehicle-mounted three-dimensional laser point clouds and streetscape images |
| CN113128405A (en) * | 2021-04-20 | 2021-07-16 | 北京航空航天大学 | Plant identification and model construction method combining semantic segmentation and point cloud processing |
| US20210248816A1 (en) * | 2019-03-28 | 2021-08-12 | Southeast University | Embedded urban design scene emulation method and system |
| CN113313832A (en) * | 2021-05-26 | 2021-08-27 | Oppo广东移动通信有限公司 | Semantic generation method and device of three-dimensional model, storage medium and electronic equipment |
| CN114387417A (en) * | 2022-03-25 | 2022-04-22 | 北京飞渡科技有限公司 | Three-dimensional building modeling method and device and three-dimensional building group modeling method |
| CN115033967A (en) * | 2022-06-27 | 2022-09-09 | 北京航空航天大学 | Building template real-time modeling method based on point cloud data |
| CN115457208A (en) * | 2022-09-19 | 2022-12-09 | 上海方联技术服务有限公司 | Three-dimensional modeling method and device for MEP equipment, storage medium and electronic device |
-
2022
- 2022-12-28 CN CN202211691743.9A patent/CN115661378B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150178988A1 (en) * | 2012-05-22 | 2015-06-25 | Telefonica, S.A. | Method and a system for generating a realistic 3d reconstruction model for an object or being |
| US20190197311A1 (en) * | 2017-12-26 | 2019-06-27 | Harbin Institute Of Technology | Evaluation Method of Solar Energy Utilization Potential in Urban High-density Areas Based on Low-altitude Photogrammetry |
| CN108389251A (en) * | 2018-03-21 | 2018-08-10 | 南京大学 | The full convolutional network threedimensional model dividing method of projection based on fusion various visual angles feature |
| CN110009727A (en) * | 2019-03-08 | 2019-07-12 | 深圳大学 | A kind of indoor threedimensional model automatic reconfiguration method and system with structure semantics |
| US20210248816A1 (en) * | 2019-03-28 | 2021-08-12 | Southeast University | Embedded urban design scene emulation method and system |
| CN111815776A (en) * | 2020-02-04 | 2020-10-23 | 山东水利技师学院 | Three-dimensional building fine geometric reconstruction method integrating airborne and vehicle-mounted three-dimensional laser point clouds and streetscape images |
| CN113128405A (en) * | 2021-04-20 | 2021-07-16 | 北京航空航天大学 | Plant identification and model construction method combining semantic segmentation and point cloud processing |
| CN113313832A (en) * | 2021-05-26 | 2021-08-27 | Oppo广东移动通信有限公司 | Semantic generation method and device of three-dimensional model, storage medium and electronic equipment |
| CN114387417A (en) * | 2022-03-25 | 2022-04-22 | 北京飞渡科技有限公司 | Three-dimensional building modeling method and device and three-dimensional building group modeling method |
| CN115033967A (en) * | 2022-06-27 | 2022-09-09 | 北京航空航天大学 | Building template real-time modeling method based on point cloud data |
| CN115457208A (en) * | 2022-09-19 | 2022-12-09 | 上海方联技术服务有限公司 | Three-dimensional modeling method and device for MEP equipment, storage medium and electronic device |
Non-Patent Citations (1)
| Title |
|---|
| 韩姗姗;黄远程;白穆;: "基于DGCNN语义分割的倾斜摄影测量三维点云建筑物分类" * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115661378B (en) | 2023-03-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bauchet et al. | Kinetic shape reconstruction | |
| Cignoni et al. | External memory management and simplification of huge meshes | |
| Gobbetti et al. | Layered point clouds: a simple and efficient multiresolution structure for distributing and rendering gigantic point-sampled models | |
| US6747651B1 (en) | System and method for creating bounding volume hierarchies utilizing model simplification | |
| KR101555426B1 (en) | Method and apparatus for rendering terrain | |
| CN113012063B (en) | Dynamic point cloud repairing method and device and computer equipment | |
| Faust et al. | Real-time global data model for the digital earth | |
| Liu et al. | Real-time, dynamic level-of-detail management for three-axis NC milling simulation | |
| Yang et al. | A hybrid spatial index for massive point cloud data management and visualization | |
| Azri et al. | Review of spatial indexing techniques for large urban data management | |
| CN115661378B (en) | Building model reconstruction method and system | |
| Glander et al. | Techniques for generalizing building geometry of complex virtual 3D city models | |
| KR100715669B1 (en) | Device and method for representation of multi-level lod 3-demension image | |
| Platings et al. | Compression of Large‐Scale Terrain Data for Real‐Time Visualization Using a Tiled Quad Tree | |
| CN113495935A (en) | Terrain slicing method and system based on irregular triangularization mesh generation | |
| WO2000004433A2 (en) | System and method for creating bounding volume hierarchies utilizing model simplification | |
| Abderrahim et al. | Interactive multiresolution visualization of 3D Mesh | |
| Ponchio | Multiresolution structures for interactive visualization of very large 3D datasets | |
| CN116758243B (en) | Scene grid division generation and rendering display method based on real-time point cloud flow | |
| CN116310086A (en) | LOD virtual complex scene simplification method based on quadtree | |
| Li et al. | Automatic 3D reconstruction of complex buildings from incomplete point clouds with topological-relation constraints | |
| CN117974899A (en) | Three-dimensional scene display method and system based on digital twinning | |
| Zhang et al. | Exploiting frame-to-frame coherence for rendering terrain using continuous LOD | |
| Zink et al. | Cloth simulation and collision detection using geometry images | |
| Song et al. | a Paralleled Delaunay Triangulation Algorithm for Processing Large LIDAR Points |
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 | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: Method and System for Building Model Reconstruction Effective date of registration: 20230613 Granted publication date: 20230321 Pledgee: Beijing ustron Tongsheng financing Company limited by guarantee Pledgor: Beijing Daoyi Shuhui Technology Co.,Ltd. Registration number: Y2023990000295 |