A method for grading building detail levels

Abstract

The present invention discloses a method for grading building detail levels, including obtaining a building to be graded; determining the classification label of the building: assigning a set of classification labels to the building, wherein the first classification label P is used to characterize the internal and external parts of the building, the second classification label C is used to characterize the geometric complexity of the building's features, the third classification label S is used to characterize the building's semantic information, and the fourth classification label D is used to characterize the dimension of the building model. A characterization matrix of the building is constructed based on the classification labels; and based on the characterization matrix, the classification labels and the corresponding buildings to be graded are associated to obtain the compiled names of the buildings to be graded. The present invention significantly improves the definition clarity and ease of operation of the building LOD in the three-dimensional city model, not only providing a clear LOD list for key stages such as data collection and acceptance, but also improving the communication efficiency and consistency of understanding of the building model.

Description

Technical Field

[0001] The invention relates to the technical field of metrology, and in particular to a grading method for building detail levels. Background Art

[0002] With the acceleration of urbanization and the development of information technology, 3D city models play an increasingly important role in urban planning, architectural design, disaster management, virtual reality (VR), and geographic information systems (GIS). These models make the visualization, analysis, and management of urban environments more intuitive and efficient by providing a 3D representation of the real world.

[0003] In the construction of 3D city models, buildings are the core components of urban geographic elements, and the accurate expression of their level of detail (LOD) is crucial to the authenticity and functionality of the model. Existing LOD standards, such as CityGML2.0, divide LOD into five levels from LOD0 to LOD4, aiming to provide models of different fineness for the external representation of buildings. However, in practical applications, this standard has encountered a series of challenges.

[0004] First, the existing LOD definitions are highly coupled in geometry and semantics, which makes it difficult to effectively separate indoor and outdoor models. For example, the details of indoor space are often closely related to the external structure, but the two may require different techniques and methods during data acquisition and modeling. In addition, the lack of refinement of indoor details makes the representation of indoor space often insufficient to meet the needs of specific applications.

[0005] Secondly, there are multiple variations in the existing LOD definition, which has caused misunderstandings in key links such as project bidding, data collection, and acceptance. Different project participants may have different understandings of the definition of LOD, resulting in inconsistent results and even disputes.

[0006] In addition, with the continuous expansion of 3D city model applications, the demand for building models is becoming more and more diverse. Different application scenarios may require different levels of detail expression, from simple outline representation to complex 3D structures and rich semantic information. Therefore, the existing LOD standard is unable to meet these diverse needs.

[0007] Level of Detail (LOD) plays a vital role in the construction of 3D city models. It defines the different levels of detail of objects in the real world. As the core of urban geographic elements, multi-level LOD modeling of buildings is a key task in building virtual 3D cities. The CityGML 2.0 standard divides LOD into five levels from LOD0 to LOD4, but in practice it encounters problems such as high coupling between geometry and semantics, difficulty in separating indoor and outdoor models, and lack of refinement of indoor details. In order to solve these problems, CityGML 3.0 revised LOD, canceled the LOD4 level, retained LOD0, LOD1, LOD2 and LOD3, and recommended that these four LOD levels be used to express both the exterior and interior of urban objects. At the same time, it decoupled the association between geometry and semantics, allowing LOD to focus on spatial representation. Although the CityGML LOD standard plays a key role in the data exchange of virtual 3D city models due to its simplicity, it also raises questions about the fuzzy definition of LOD. In particular, there are multiple variations in the definitions of these LODs, which leads to misunderstandings in project bidding, data collection, and acceptance, which in turn leads to problems where the results do not match expectations and cause disputes. Therefore, establishing a clear and easy-to-operate definition and classification system for building LODs has become a major challenge faced by academia and industry. Therefore, a classification method for building detail levels is needed. Summary of the invention

[0008] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a method for grading building detail levels, aiming to solve the above technical problems.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] The present invention comprises the following steps:

[0011] Get the building to be graded;

[0012] Determine the classification label of the building: assign a classification label to the building, the classification label includes a first classification label P, a second classification label C, a third classification label S and a fourth classification label D, the first classification label P is used to characterize the internal and external parts of the building, the second classification label C is used to characterize the geometric complexity of the building's features, the third classification label S is used to characterize the building semantic information, and the fourth classification label D is used to characterize the dimension of the building model, and construct a characterization matrix of the building based on the classification labels;

[0013] Based on the characterization matrix, the classification labels are associated with the corresponding buildings to be classified to obtain compiled names of the buildings to be classified.

[0014] Furthermore, the building shape is segmented, and the image of the single building shape output after the image segmentation is classified and input into the classification judgment model to grade the structure of each building surface, including using the ResNet18 image classification network to classify the labiocheir and occlusal surfaces in the input single building shape data, and then selecting the convolutional network model to grade each building surface and grade the structure of each building surface, and then using the bounding box information generated by the segmentation task to locate in the original image, the classification results are collected in the original input image and returned, and the structural graded judgment result of each building surface is output to realize the judgment of the building shape structure in the input image.

[0015] Furthermore, constructing a representation matrix of a building based on classification labels specifically includes: determining a hierarchical label of the building: determining a hierarchical label for each detail part of the building, constructing a character feature matrix, and determining a label matrix.

[0016] Further, determine the hierarchical label index of each detail part, assign a hierarchical index to each detail part, convert the hierarchical label index of each detail part into a specific hierarchical label, establish an association relationship between the compiled name and the corresponding target name, search the association relationship to determine the corresponding target name; rename the compiled name to the corresponding target name.

[0017] Furthermore, the first classification label P includes: an external LOD, which is used to represent the appearance of the building and its relationship with the surrounding environment; and an internal LOD, which is used to represent the internal space of the building and the architectural elements constituting these spaces.

[0018] Furthermore, the fourth classification label D further includes: 0-dimensional model, represented as points, for the highest level of abstraction; 1-dimensional model, represented as a linear node relationship graph (NRG), for path planning and navigation; 2-dimensional model, represented as a base contour map or floor plan, for urban planning and land management; 3-dimensional model, represented as a three-dimensional building form, for visualization and building energy efficiency analysis.

[0019] Furthermore, the third classification label S further includes: S0, indicating that the model does not distinguish any semantic information, S1 to Sx, indicating that the model at least distinguishes one or more building components.

[0020] Furthermore, the encoding of the third classification label S is binary coded to assign a unique identifier to each building component, the converted grayscale image is segmented into a plurality of square local regions, the local regions are regarded as square kernels and pixel values ​​are obtained;

[0021] Calculate pixel value:

[0022] ic′=ic+αΣb∈sk·(ib-ic)·sgn(e)

[0023] The number of area pixels b is s, the gray value of area pixel b is ib, the gray value of the center pixel is ic, the sign function is k·, the gray factor is α, sgn(e) is the sign function, and returns 1 when 0e≥0, otherwise returns -1 or 0;

[0024] The center pixel is used as the threshold and the grayscale is compared with the domain pixel value. The position with a large domain value is recorded as 1, and the position with a small domain value is recorded as 0, thus obtaining a binary sequence.

[0025] According to the conversion relationship between different bases and combined with the binary weight, the binary is converted into decimal. The converted decimal is the eigenvalue of the local central pixel, and the eigenvalue is output as the feature of the central pixel.

[0026] Furthermore, the second classification label C includes the linear structure complexity and the surface structure complexity of the geometric complexity, and further includes: the linear structure complexity is used to depict the detailed changes of the building outline; the surface structure complexity is used to measure the geometric complexity of a surface area in the 3D model.

[0027] An electronic device,

[0028] The invention comprises: a processor; and a memory arranged to store computer executable instructions, which when executed cause the processor to perform the method.

[0029] In summary, due to the adoption of the above technical solution,

[0030] The beneficial effects of the present invention are:

[0031] The present invention significantly improves the definition clarity and ease of operation of building LOD in three-dimensional urban models, not only providing a clear LOD list for key stages such as data collection and acceptance, but also improving the communication efficiency and consistency of understanding of building models. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a classification of the generalized building model of the present invention.

[0033] Figure 2 Schematic diagrams of different types of linear structures of the present invention.

[0034] Figure 3 Schematic diagrams of different types of planar structures of the present invention.

[0035] Figure 4 It is a simplified schematic diagram of the linear structure of the present invention.

[0036] Figure 5 This is an example schematic diagram of a 1LOD series of embodiments of the present invention.

[0037] Figure 6 This is an example schematic diagram of a 2LOD series of embodiments of the present invention.

[0038] Figure 7 It is an example schematic diagram of a 3LOD series of embodiments of the present invention.

[0039] Figure 8 This is an example schematic diagram of a 4LOD series of embodiments of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] An embodiment of a method for grading building detail levels according to the present invention is further described.

[0042] like Figure 1 As shown, the technical solution of the present invention mainly includes the following aspects:

[0043] The method includes obtaining a building to be classified; determining a classification label of the building: assigning a set of classification labels to the building, wherein a first classification label P is used to characterize the internal and external parts of the building, a second classification label C is used to characterize the geometric complexity of the building's features, a third classification label S is used to characterize the building's semantic information, and a fourth classification label D is used to characterize the dimension of the building model, and constructing a characterization matrix of the building based on the classification labels; and establishing an association between the classification labels and the corresponding building to be classified based on the characterization matrix to obtain a compiled name of the building to be classified.

[0044] Building interior and exterior parts (Part, P): The present invention distinguishes between the exterior (Exter ior) and interior (Interior) of a building so as to select the appropriate model type according to different application requirements and data acquisition technologies. For example, urban planning may focus more on the exterior model of a building, while interior design may require a detailed interior model.

[0045] Model Dimension (D): This invention divides the dimensions of the building model into 0D, 1D, 2D and 3D to evaluate the fineness of LOD. The 0D model is suitable for applications that do not require geometric information, while the 3D model provides the richest details and is suitable for scenes that require stereoscopic representation.

[0046] Semantic information (S): This invention provides accurate classification and identification of the semantic information of building components through option method and coding method. The option method uses a predefined set of options to represent the type and complexity of the building semantic information contained in the model, while the coding method uses binary coding to achieve more detailed semantic distinction.

[0047] Geometric complexity (Completeness, C): This paper evaluates the accuracy of model detail display through the complexity of linear structure and surface structure. Linear structure complexity focuses on the detailed changes of building outlines, while surface structure complexity measures the geometric complexity of surface areas in 3D models.

[0048] Determine the interior and exterior parts of the building (P):

[0049] For a given building model, first determine whether it is the exterior (Exter ior) or the interior (Inter ior) that needs to be expressed. The exterior LOD (ELOD) focuses on the building's appearance and its relationship with the environment, while the interior LOD (ILOD) focuses on the building's internal space and components.

[0050] Select the dimension (D) of the model:

[0051] Choose the appropriate model dimension based on application requirements and data collection feasibility. For example, a 0-dimensional (G0) model is suitable for simplified representation, while a 3-dimensional (G3) model provides richer details. 1-dimensional (G1) and 2-dimensional (G2) models are suitable for specific application scenarios, such as path planning or urban planning.

[0052] Define semantic information(S):

[0053] Depending on the level of semantic detail required, choose the appropriate level of semantic information. Use either the option method (S0 to Sx) or the coding method (binary coding) to define the kind and complexity of the building components included in the model. For example, S0 might mean no semantic distinction, while S1 might mean at least walls and doors are distinguished.

[0054] Set the geometric complexity (C):

[0055] Set the threshold of geometric complexity according to the expected use and performance requirements of the model. Linear complexity (Lr) focuses on the details of the building outline, while surface complexity (Ar) focuses on the complexity of the surface area in the 3D model. N means ignoring geometric accuracy.

[0056] Combined PDSC nomenclature:

[0057] The selected parameters of the above four dimensions are combined to form a complete LOD naming. For example, ELOD-G20-S1-Lr represents an external LOD, which is represented by a 2D base map, distinguishes between walls and doors, and has linear structural complexity.

[0058] In "PDSC", P is used to distinguish whether LOD expresses the exterior or interior of a building. P has two options: {ELOD, ILOD}, where ELOD represents the exterior detail level (Exter ior LOD) of the building and ILOD represents the interior detail level (Inter ior LOD). D is used to distinguish what dimension of model LOD uses to express the building. D has four options: {G0, G1, G20 / 1, G30 / 1 / 10 / 11}. Among them, G0 means using a 0D model to express the building; G1 means using a 1D NRG to express the building; G2 means using a 2D model to express the building, which is further subdivided into G20 and G21. G20 means using a base contour map to express the building. G21 means using the outer contour polygon of each floor plan to express the building. G3 means using a 3D model to express the building, which is further subdivided into G30, G31, G310 and G311. G30 means using a 3D prism model to express the entire building, G31 means using a 3D prism model to express each floor of the building; G310 means using a 3D solid or surface model to express the entire building, G311 means using a 3D solid or surface model to express each floor of the building. If you want to distinguish multiple components at the same time, you can do so through a combination of codes, such as 00000011 means the model distinguishes between walls and doors, and 00000111 means the model distinguishes between walls, doors, and windows at the same time. The coding method can distinguish multiple semantics and is extensible, which can meet the needs of distinguishing multiple semantics of building models in different scenarios.

[0059] In the LOD (Level of Detail) model, the parameter C is used to distinguish the geometric complexity of the building to show the model's expressiveness of architectural features and details at different levels. C contains four options {N, Lr, Ar, LtAs}, where N means ignoring geometric accuracy, that is, the model does not show specific geometric shapes and structural details. Lr represents linear structural complexity, and "r" is the threshold of the main edge length. Figure 2 The edge marked with 1 is the main edge, and the two sides are wing edges. When the length of the main edge of the linear structure is less than this threshold, the related protrusion or concave details will be simplified. Ar represents the complexity of the surface structure, and "r" is the threshold of the main surface area. Figure 3 The face marked with 1 is the main face, and the two side faces are wing faces. If the main face area of ​​the planar structure is smaller than this threshold, the corresponding convex or concave planar structure will also be simplified. Figure 4(a) shows the LOD of the 2D building base map with the main edge length threshold set to 0.4 meters, where linear structures less than 0.4 meters are simplified. Figure 4 In (b), the main edge length threshold is increased to 0.8 meters. It can be seen that the linear structure with a main edge length less than 0.8 meters in (a) is further simplified. Figure 4 In (c), the main edge length threshold is increased to 1.5 meters, which further simplifies the linear structure with a main edge length less than 1.5 meters in (b). LtAs is a combination of the complexity of linear and planar structures, that is, the model needs to consider both aspects at the same time.

[0060] Division of building LOD

[0061] Division of building exterior LOD Based on the first indicator, the building model LOD is divided into ELOD and I LOD. According to the second indicator, the ELOD series is divided into ELOD-G0, ELOD-G1, ELOD-G2 and ELOD-G3 series. The third indicator distinguishes the semantics of ELOD, and the fourth indicator distinguishes the geometric complexity.

[0062] ELOD-G0-S0-N is a semantically insensitive LOD, such as Figure 5 (a); ELOD-G0-S1-N is the LOD that distinguishes semantics, such as Figure 5 (b).

[0063] The LOD-G1 series is an NRG that abstracts buildings as 0D points and the relationships between buildings as edges. Figure 6 (a) (b) are the LODs that are semantically insensitive and semantically sensitive, respectively. The ELOD-G2 series uses a 2D contour map to represent the exterior of a building. The ELOD-G20-Sx-Lr uses a base map to represent the exterior of a building, and it uses linear structure complexity to indicate geometric complexity. Among them, the ELOD-G20-S0-Lr series is the LOD series that is semantically insensitive, such as Figure 7 The r values ​​of (a) and (e) are 0.5 meters and 3 meters respectively; ELOD-G20-S1-Lr is the LOD series for distinguishing semantics, such as Figure 7 (b)(f) Yes Figure 7 (a)(e) The corresponding LODs distinguish the semantics of doors and walls. The ELOD-G21-Sx-Lr series uses 2D floor outlines to express the exterior of a building, which uses linear structures to indicate geometric complexity. Among them, ELOD-G21-S0-Lr is a LOD series that does not distinguish semantics, such as Figure 7 (c)(g), their difference lies in the value of r; ELOD-G21-S1-Lr is a LOD series that distinguishes semantics, such as Figure 7(d)(h), their difference is also the value of r. The ELOD-G3 series uses a 3D model to express the exterior of the building. The ELOD-G30-Sx-Lr uses a prismatic model to express the exterior of the building. It does not distinguish between floors and uses linear structures to indicate geometric complexity. Among them, the ELOD-G30-S0-Lr is a LOD series that does not distinguish semantics, such as Figure 8 The r values ​​of (a) and (e) are 0.5 meters and 3 meters respectively; ELOD-G30-S1-Lr is the LOD series for distinguishing semantics, such as Figure 8 (b)(f) are the LODs corresponding to 8(a)(e) that distinguish the semantics of doors and walls. ELOD-G310-Sx-Lr uses solid models or surface models to express the exterior of the building, and does not distinguish between floors. ELOD-G311-Sx-Lr also uses solid models or surface models to distinguish between floors, and they all use linear structures to indicate geometric complexity. ELOD-G310-Sx-Ar and ELOD-G311-Sx-Ar are similar to the Lr series, but use surface structures to indicate geometric complexity.

[0064] The combination of five types of indicators will produce different LODs inside the building. I LOD-G0 means that the 0D model is used to represent the interior of the building. I LOD-G0-S0-N does not distinguish between semantics, and I LOD-G0-Sx-N (x is greater than 0) distinguishes different semantics.

[0065] Application examples:

[0066] Taking urban planning as an example, the building model of ELOD-G21-S2-Lr may be needed, which provides the outer contour polygon of the floor plan, distinguishes walls, doors and windows, and takes into account the complexity of the linear structure.

[0067] For indoor navigation systems, the ILOD-G1-S1-N model may be needed, which uses 1D NRG to represent indoor structures, distinguishing between walls and doors, but does not focus on geometric accuracy.

[0068] Through the above implementation, the "PDSC" nomenclature of the present invention provides a clear, flexible and easy-to-operate classification system for the LOD of buildings, which is suitable for the construction and application requirements of various three-dimensional city models.

[0069] The above description of the embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for grading building detail levels, characterized in that: The following steps are involved: Get the building to be graded; Determine the classification label of the building: assign a set of classification labels to the building, the classification labels include a first classification label P, a second classification label C, a third classification label S and a fourth classification label D, the first classification label P is used to characterize the internal and external parts of the building, the second classification label C is used to characterize the geometric complexity of the building's features, the third classification label S is used to characterize the building semantic information, and the fourth classification label D is used to characterize the dimension of the building model, and construct a characterization matrix of the building based on the classification labels; Associating the classification labels with the corresponding buildings to be classified based on the representation matrix to obtain compiled names of the buildings to be classified; Constructing a representation matrix of building details based on classification labels specifically includes: determining hierarchical labels of building details: determining a hierarchical label for each detail part of the building, constructing a character feature matrix, and determining a label matrix; Determine the hierarchical label index of each detail part, assign a hierarchical index to each detail part, convert the hierarchical label index of each detail part into a specific hierarchical label, establish an association relationship between the compiled name and the corresponding target name, search the association relationship to determine the corresponding target name; rename the compiled name to the corresponding target name.

2. A method for grading building detail levels according to claim 1, characterized in that: Determining the classification label of the building specifically includes segmenting the building shape and performing image segmentation on it, classifying the image of a single building shape output after image segmentation and inputting it into a classification judgment model to grade the structure of each building surface, including using the ResNet18 image classification network to classify the labial and buccal surfaces and occlusal surfaces in the input single building shape data, and then selecting a convolutional network model to grade each building surface and grade the structure of each building surface, and then using the bounding box information generated by the segmentation task to locate in the original image, the classification results are collected in the original input image and returned, and the structural graded judgment results of each building surface are output to realize the judgment of the classification label of the building shape in the input image.

3. A method for grading building detail levels according to claim 1, characterized in that: The first classification label P includes: external LOD, which is used to represent the appearance of the building and its relationship with the surrounding environment; internal LOD, which is used to represent the interior space of the building and the architectural elements that constitute these spaces.

4. A method for grading building detail levels according to claim 1, characterized in that: The fourth category label D further includes: 0D model, represented as points, for the highest level of abstraction; 1D model, represented as a linear node relationship diagram, for path planning and navigation; 2D model, represented as a base contour map or floor plan, for urban planning and land management; 3D model, represented as a three-dimensional building form, for visualization and building energy efficiency analysis.

5. A method for grading building detail levels according to claim 1, characterized in that: The third classification label S further includes: S0, indicating that the model does not distinguish any semantic information, S1 to Sx, indicating that the model at least distinguishes one or more building components.

6. A method for grading building detail levels according to claim 5, characterized in that: The third classification label S is encoded using binary coding, which is used to assign a unique identifier to each building component. The converted grayscale image is segmented into multiple square local regions, and the local regions are regarded as square kernels and pixel values ​​are obtained; Calculate pixel value: ic′=ic+α∑ b∈S k·(i b -ic)·sgn(e) The number of area pixels b is s, and the grayscale value of area pixels b is i. b , the gray value of the center pixel is ic, the sign function is k·, the gray factor is α, sgn(e) is the sign function, and it returns 1 when 0e≥0, otherwise it returns -1 or 0; The center pixel is used as the threshold and the grayscale is compared with the domain pixel value. The position with a large domain value is recorded as 1, and the position with a small domain value is recorded as 0, thus obtaining a binary sequence. According to the conversion relationship between different bases and combined with the binary weight, the binary is converted into decimal. The converted decimal is the eigenvalue of the local central pixel, and the eigenvalue is output as the feature of the central pixel.

7. A method for grading building detail levels according to claim 1, characterized in that: The second classification label C includes the linear structure complexity and the surface structure complexity of the geometric complexity, and further includes: the linear structure complexity is used to depict the detailed changes of the building outline; the surface structure complexity is used to measure the geometric complexity of a surface area in the 3D model.

8. An electronic device comprising: processor; as well as A memory arranged to store computer executable instructions, which when executed cause the processor to perform the method according to any one of claims 1 to 7.

Images