I-shaped building group pattern recognition method, device and system and storage medium

Abstract

The invention discloses an I-shaped building group pattern recognition method, device and system and a storage medium, and the method comprises the steps: S1, constructing a building knowledge graph according to the I-shaped arrangement space relation of a building group and the matching relation of corresponding buildings on different scales; and S2, according to the building knowledge graph, through rule-based reasoning, identifying an I-shaped building mode under a single spatial scale and an I-shaped building mode supported by multi-scale data, which conform to a human visual identification principle. By adopting the technical scheme of the invention, the recognition effect of the I-shaped building mode is remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of building recognition, and in particular relates to a method and device, a system, and a storage medium for recognizing an I-shaped building group pattern. Background Art

[0002] Building patterns usually refer to visually significant structures composed of a single building or a group of buildings. In modern urban planning and spatial analysis, they often contain rich spatial information and social significance. In-depth research on building patterns helps to optimize urban spatial structure and improve land use efficiency. By analyzing the morphology and function of building patterns, municipal planners can better identify the bottlenecks of current urban development and future expansion directions. In the construction of smart cities, the automatic recognition and analysis of building patterns provide technical support for intelligent transportation systems, energy management, urban security and other fields, and provide scientific decision-making basis for urban managers. Automatically extracting, simplifying and retaining the core information of building patterns from large-scale data can not only reduce redundant information, but also more intuitively display important building features in geographic space. Therefore, building pattern recognition is also the key to improving the comprehensive efficiency and accuracy of maps.

[0003] Existing research on architectural pattern recognition can be divided into two categories: shape recognition for a single building and arrangement recognition for a group of buildings. Shape recognition mainly relies on the shape features and parameter representation of buildings, and uses methods such as corner functions, Fourier descriptors, and graph convolutional autoencoder models to identify the shape of a single building. Arrangement recognition uses template matching, structural rules, and machine learning methods to identify the layout of a group of buildings. At present, "complex" architectural patterns represented by I-shaped buildings have become an important research object due to their typical geometric structure and significant spatial relationships. The above traditional methods mostly rely on the extraction of geometric features of independent objects. Although they work well in the recognition of the shape of a single building, they still have problems such as insufficient accuracy and poor robustness when dealing with the pattern recognition of I-shaped building groups in complex urban scenes and multi-scale spatial data, and they cannot recognize all visually obvious patterns. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method and device, system, and storage medium for identifying an I-shaped building group pattern, which can significantly improve the recognition effect of the I-shaped building pattern.

[0005] To achieve the above object, the present invention adopts the following technical solution:

[0006] A method for identifying an I-shaped building group pattern, comprising:

[0007] Step S1, constructing a building knowledge graph according to the spatial relationship of the I-shaped arrangement of the building group and the matching relationship of the corresponding buildings at different scales; wherein the spatial relationship of the I-shaped arrangement of the building group includes: adjacent relationship, completely parallel relationship, completely vertical relationship, relative position relationship, and center distance constraint; the matching relationship of the corresponding buildings at different scales includes: one-to-one, one-to-many, many-to-one, many-to-many, one-to-zero and zero-to-one;

[0008] Step S2: Based on the building knowledge graph, rule-based reasoning is used to identify I-shaped building patterns at a single spatial scale that conforms to human visual recognition principles and I-shaped building patterns supported by multi-scale data.

[0009] The present invention also provides an I-shaped building group pattern recognition device, comprising:

[0010] A construction module is used to construct a building knowledge graph based on the spatial relationship of the I-shaped arrangement of building groups and the matching relationship of corresponding buildings at different scales; wherein the spatial relationship of the I-shaped arrangement of building groups includes: proximity relationship, complete parallel relationship, complete vertical relationship, relative position relationship, and center distance constraint; the matching relationship of corresponding buildings at different scales includes: one-to-one, one-to-many, many-to-one, many-to-many, one-to-zero, and zero-to-one;

[0011] The recognition module is used to identify the I-shaped building pattern at a single spatial scale that conforms to the human visual recognition principle and the I-shaped building pattern supported by multi-scale data through rule-based reasoning based on the building knowledge graph.

[0012] An embodiment of the present invention further provides an I-shaped building group pattern recognition system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes an I-shaped building group pattern recognition method when executed by the processor.

[0013] An embodiment of the present invention further provides a storage medium having a computer program stored thereon, and the computer program executes an I-shaped building group pattern recognition method when running.

[0014] The present invention applies knowledge graphs to express the multi-scale shape features, spatial relationships and semantic information of buildings, so as to realize the recognition of I-shaped building patterns; the present invention utilizes knowledge graphs to identify the I-shaped arrangement patterns in building complexes, and combines the recognition results of multi-scale data to enhance the recognition effect of potential I-shaped building patterns. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0016] Figure 1 This is a flow chart of a method for identifying an I-shaped building group pattern according to an embodiment of the present invention;

[0017] Figure 2 A schematic diagram for constructing a proximity graph of the present invention; wherein, Figure 2 (a) is a schematic diagram of the triangulated network and skeleton. Figure 2 (b) is a schematic diagram of neighboring graph nodes and edges;

[0018] Figure 3 A schematic diagram of the relative position of buildings in eight-directional neighborhoods;

[0019] Figure 4 It is a schematic diagram of relative position relationship type;

[0020] Figure 5 A schematic diagram of the reasoning of the I-shaped building group pattern combining multiple scales;

[0021] Figure 6 It is a schematic diagram of the process of combining multi-scale enhanced recognition; Figure 6 (a) is a schematic diagram of multi-scale data. Figure 6 (b) is a schematic diagram of the process from large to small scale. Figure 6 (c) is a schematic diagram of the process from small to large scale. DETAILED DESCRIPTION

[0022] 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.

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] Embodiment 1:

[0025] like Figure 1 As shown, an embodiment of the present invention provides a method for identifying an I-shaped building group pattern, comprising:

[0026] Step S1, constructing a building knowledge graph according to the spatial relationship of the I-shaped arrangement of the building group and the matching relationship of the corresponding buildings at different scales; wherein the spatial relationship of the I-shaped arrangement of the building group includes: adjacent relationship, completely parallel relationship, completely vertical relationship, relative position relationship, and center distance constraint; the matching relationship of the corresponding buildings at different scales includes: one-to-one, one-to-many, many-to-one, many-to-many, one-to-zero and zero-to-one;

[0027] Step S2: Based on the building knowledge graph, rule-based reasoning is used to identify I-shaped building patterns at a single spatial scale that conforms to human visual recognition principles and I-shaped building patterns supported by multi-scale data.

[0028] As an implementation of an embodiment of the present invention, in step S1, the neighbor relationship is: in describing a building group, for the neighbor relationship between entities that are spatially adjacent but not connected, the embodiment of the present invention adopts a constrained Delaunay triangulation (CDT) skeleton method, so that the neighbor relationship between two buildings in the map is determined by the shared boundaries or angles between the entities. Figure 2 As shown in (a), buildings that share the same CDT skeleton are defined as neighbors, and then a proximity graph is constructed to represent the proximity relationship between buildings, as shown in Figure 2 (b) Building B i and B j Proximity is represented in the knowledge graph as the existence of a Neighbor type relationship between corresponding entities.

[0029] As an implementation method of an embodiment of the present invention, in step S1, the relative position relationship is: the relative position relationship between two buildings is defined by approximating the building entity with a Smallest Bounding Rectangle. Assume that two buildings Ba (reference building) and Bb (target building) have their Smallest Bounding Rectangles (SBRs) SBRa and SBRb, respectively. The eight-directional neighborhood of SBRa is defined by taking the major axis and minor axis of SBRa as the X-axis and Y-axis, respectively, as follows: Figure 3 The relative position relationship between Ba and Bb can be expressed by judging the position of SBRb in SBRa and its eight-directional neighborhood.

[0030] SBRb can form 13 basic relative position relationships on the long axis or short axis of SBRa. Considering that SBRb may have projections on both the long axis and the short axis of SBRa, the relative position relationship between them can be expressed as r i,j (a,b), that is, 169 one-way relations of Bb relative to Ba, such as Figure 4 As shown, Figure 4 In the figure, the diagonal line frame represents Ba, and the solid color frame represents Bb.

[0031] Since there are basically no overlapping buildings in reality, the overlapping relative position relationship ( Figure 4 In addition, for buildings with too complex shapes, it is generally believed that they do not form a regular pattern. Therefore, before using the minimum enclosing rectangle to approximate the building, the rectangularity (i.e., the ratio of the building area to its minimum enclosing rectangle area) is calculated to determine the complexity of its shape. If the rectangularity is greater than 0.6, it means that its shape is close to a rectangle. Otherwise, it means that its shape is complex, and it is excluded from the determination of the relative position relationship, and it is considered that it does not participate in the formation of the I-shaped building pattern.

[0032] As an implementation of an embodiment of the present invention, in step S1, the completely parallel relationship is defined as two buildings Ba and Bb having similar sizes and shapes, being parallel to each other in the main direction, and only having a certain degree of overlap in their long axes. In order to further quantify the degree of overlap in the long axes between the two buildings (Ba and Bb), the long axis overlap ratio (LR) is defined as:

[0033]

[0034] in, represents the projection length of Bb on the major axis of Ba, l a It represents the length of the major axis of building Ba, LR(Ba,Bb) represents the degree of overlap between the projection of Bb on the major axis of Ba and the major axis of Ba, Overlap represents the intersection, and Merge represents the union.

[0035] A completely parallel relationship is defined as:

[0036]

[0037] Among them, a and b represent the serial numbers of buildings, SimA represents the similarity relationship, Aa and Ab represent the areas of buildings Ba and Bb respectively, PalO represents the vertical direction, Oa and Ob represent the directions of SBR of buildings Ba and Bb respectively, LR is the overlap ratio of the long axes between the two buildings, and θ1, θ2, and θ3 are thresholds.

[0038] As an implementation of the embodiment of the present invention, in step S1, the completely perpendicular relationship: if Ba and Bb have mutually perpendicular main directions, and the projection of Bb on Ba completely overlaps with the long axis of Ba, then the building Bb is considered to be completely perpendicular to Ba. Expressed as Full_Per(Ba,Bb), a one-way relationship from Ba to Bb, the embodiment of the present invention defines the completely perpendicular relationship as:

[0039]

[0040] Among them, Per() represents the vertical direction, and Oa-Ob represents the difference in the direction angles of the SBR of the building with serial number a.

[0041] As an implementation of an embodiment of the present invention, in step S1, the center distance constraint is: since the characteristic of the I-shaped building distribution pattern is that the two wing buildings are parallel to each other, the middle building is perpendicular to the two wings and is located at the center of the two wing buildings. Therefore, the degree of centering of the middle building is an important constraint, which affects whether the building group can form an I-shaped building pattern in the recognition principle of human vision. By calculating the relative distance between the center of gravity of the middle building and the center of gravity of the entire building group, a center distance evaluation system is constructed, which is defined as:

[0042]

[0043] Among them, dis(x,y) represents the centroid distance of buildings x and y, A represents the central building, and B1 and B2 represent the buildings on the two wings.

[0044] As an implementation method of an embodiment of the present invention, in step S1, the small-scale map data is usually updated by performing map synthesis on the large-scale map data, and the process of map synthesis is essentially a process of information generalization. When the buildings in the large-scale map are updated to the map of smaller scale, they may be simplified in shape, merged with other buildings, or directly deleted. This will cause the same building to have larger shapes and areas in maps of different scales, or not exist at another scale. Therefore, six relationships are constructed, including one-to-one (1:1), one-to-many (1:M), many-to-one (M:1), many-to-many (M:N), one-to-zero (1:0) and zero-to-one (0:1), to match the entities of the same building at different scales. As shown in Table 1.

[0045] Table 1

[0046] type content 1:1 Indicates that a building at one scale corresponds to only one building at another scale 1:M Indicates that a building at one scale corresponds to multiple buildings at another scale M:1 Indicates that multiple buildings at one scale correspond to one building at another scale M:N Indicates that several buildings at a certain scale correspond to several buildings at another scale 1:0 Indicates that a building at one scale has no corresponding building at another scale 0:1 Indicates that there is no building at one scale that corresponds to a building at another scale

[0047] Furthermore, the attribute graph model in the Neo4j graph database is used to describe the graph structure of the building group with nodes and edges, and the attribute graph of the building group is recorded as: Graph = (B, R),

[0048] Where: B is a node set, B={B1,B2,…B n}; Each node represents an independent building entity and is assigned different labels according to its scale: "BuildingBig" is used for large-scale buildings, and "BuildingSmall" is used for small-scale buildings. Each building entity has four attributes: "bID" represents a unique ID number; "Area" represents the building area; "Orientation" represents the main direction of the building; "Pattern" represents whether the building forms an I-shaped arrangement pattern.

[0049] R is the edge set, R={Rm(B i ,B j ),B i ∈B,B j ∈B}, Rm(B i ,B j ) represents entity B i and B j The proximity between buildings A and B is represented in the knowledge graph as the existence of a Ba-(Neighbor)-Bb relationship between Ba and Bb. The relative position relationship can be represented as a unidirectional relationship Ba-(Position)->Bb from Ba to Bb. The relative position relationship has two relationship attributes: "P_type" indicates the type of relative position relationship; "LR" indicates the degree of overlap of the long axes of the two buildings. The bidirectional completely parallel relationship and the unidirectional completely perpendicular relationship between Ba and Bb are inferred from the size, main direction, proximity and relative position relationship between the buildings, which are represented by Ba-(Full_Pal)-Bb and Ba-(Full_Per)->Bb respectively. The center distance of the middle building can be inferred from the center of gravity distance between the building entities. The center of gravity distance between Ba and Bb is expressed as dis(Ba,Bb).

[0050] The matching relationship between corresponding buildings at different scales can be represented as a "Match" relationship with the attribute "M_type", where the "M_type" value corresponds to the relationship type, including one-to-one (1:1), one-to-many (1:M), many-to-one (M:1), and many-to-many (M:N). The one-to-zero (1:0) and zero-to-one (0:1) relationships mentioned above do not need to be explicitly represented in the knowledge graph.

[0051] The entities and relationships in the knowledge graph are shown in Table 2.

[0052] Table 2

[0053]

[0054]

[0055] As an implementation method of an embodiment of the present invention, in step S2, based on the constructed knowledge graph, I-shaped building pattern recognition is performed through rule-based reasoning, including two steps: first, the I-shaped building pattern is identified at a specific spatial scale, and second, the I-shaped building pattern recognition result is supported by multi-scale data to enhance the overall pattern recognition effect.

[0056] The I-shaped building pattern recognition at a single spatial scale is as follows:

[0057] The knowledge graph is stored in the Neo4j graph database, and the Cypher language is used to define the rules for reasoning and identifying the I-shaped building pattern. The recognition of the I-shaped building pattern at a specific spatial scale requires obtaining the Neighbor, Full_Pal, Full_Per relationships and Center constraints. Prior to this, from the definitions of equations (2), (3) and (4), it is known that it is necessary to determine the similarity between building entities, the parallelism or perpendicularity of the main directions, and the distance between the centers of gravity, etc., to infer the Full_Pal, Full_Per relationships and Center constraints. As shown in equation (5).

[0058] The specific steps of inferring the Center constraint from the centroid distance between building entities are as follows: (1) Assume there are three buildings B m , B1, B2, based on their proximity, relative position, and completely parallel and completely perpendicular relationship, it is inferred that they may form an I-shaped pattern. (2) By the distance between their centroids dis(B m ,B1), dis(B m ,B2), dis(B1,B2) and formula (4) to calculate the center distance. (3) Compare the calculated center distance with the threshold. If it is less than the threshold, modify the attribute "Pattern" of the three entities to "True", indicating that the building entities form an I-shaped pattern after the center distance constraint; otherwise, it is considered that these buildings do not form an I-shaped pattern under the principle of human visual recognition. The center distance constraint is not explicitly represented in the knowledge graph.

[0059]

[0060] Among them, a, b1, b2 represent buildings a, b1, b2.

[0061] Therefore, the reasoning process of rule-based I-shaped building pattern recognition includes four steps: (1) verifying the SimA, PalO and PerO relationships; (2) reasoning out the Full_Para and Full_Per relationships; (3) verifying the Center constraint; (4) identifying the I-shaped building pattern. Figure 5 shown.

[0062] The I-shaped building pattern recognition supported by multi-scale data is as follows:

[0063] The research on enhancing pattern recognition effect by using multi-scale data can be divided into two types. One is the enhancement from large to small scale, that is, treating a large-scale building group as a decomposition of the corresponding building at a smaller scale. For a certain building or building group, if it is identified as an I-shaped pattern at a large scale, it is inferred that the corresponding building at a smaller scale may also form an I-shaped pattern. For example, at a larger scale, if three buildings (such as Figure 6 In (a), B3, B4, and B5 are identified as an I-shaped pattern. At a smaller scale, these buildings may be decomposed aggregates of the corresponding buildings. Specifically, B5 and B2 have a one-to-one (1:1) correspondence, and B3, B4 and B1 have a many-to-one (M:1) correspondence. Therefore, B1 and B2 may form an I-shaped pattern at a smaller scale. This enhanced recognition process from large to small is shown in Figure 2. Figure 6 (b) As shown. The second is the enhancement from small to large scale, that is, the smaller-scale building group is regarded as the grouping aggregation of the corresponding buildings at a larger scale. For a certain building group, if an I-shaped pattern is identified at a small scale, it is inferred that the corresponding building group at a larger scale may also form an I-shaped pattern. For example, at a smaller scale, if three buildings (such as B3, B4, and B5) are identified as an I-shaped pattern, their corresponding buildings at a larger scale may be decomposed versions of these three buildings, and may still form an I-shaped pattern. Specifically, B3 has a one-to-many (1:M) relationship with B6 and B7, while B4 has a one-to-one (1:1) relationship with B8, and B5 has a one-to-one (1:1) relationship with B9. Then, the larger-scale B6, B7, B8, and B9 may also form an I-shaped pattern. This reflects the enhanced recognition process from small to large. Figure 6 (c) as shown.

[0064] Therefore, the use of multi-scale data to enhance recognition includes: first, constructing a cross-scale building matching relationship and searching for buildings that have been identified as I-shaped patterns at a specific scale. Secondly, according to the relationship between the identified buildings and buildings at other scales (smaller or larger), determine the applicable enhanced recognition type. Finally, based on the enhanced recognition process from large to small or from small to large, identify potential I-shaped patterns. Through this cross-scale matching and pattern conversion, more accurate building pattern recognition can be achieved between different scales.

[0065] The embodiment of the present invention abstracts the spatial relationship between building entities and the matching relationship across scales into nodes and edges in the knowledge graph, and constructs center distance constraints to make the recognition results more consistent with human visual principles; and identifies the I-shaped pattern of the building and the potential pattern features based on rule reasoning in the knowledge graph.

[0066] Embodiment 2:

[0067] The embodiment of the present invention further provides an I-shaped building group pattern recognition device, comprising:

[0068] A construction module is used to construct a building knowledge graph based on the spatial relationship of the I-shaped arrangement of building groups and the matching relationship of corresponding buildings at different scales; wherein the spatial relationship of the I-shaped arrangement of building groups includes: proximity relationship, complete parallel relationship, complete vertical relationship, relative position relationship, and center distance constraint; the matching relationship of corresponding buildings at different scales includes: one-to-one, one-to-many, many-to-one, many-to-many, one-to-zero, and zero-to-one;

[0069] The recognition module is used to identify the I-shaped building pattern at a single spatial scale that conforms to the human visual recognition principle and the I-shaped building pattern supported by multi-scale data through rule-based reasoning based on the building knowledge graph.

[0070] Embodiment 3:

[0071] An embodiment of the present invention further provides an I-shaped building group pattern recognition system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes an I-shaped building group pattern recognition method when executed by the processor.

[0072] Embodiment 4:

[0073] An embodiment of the present invention further provides a storage medium having a computer program stored thereon, and the computer program executes an I-shaped building group pattern recognition method when running.

[0074] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for identifying an I-shaped building group pattern, characterized in that: include: Step S1, constructing a building knowledge graph according to the spatial relationship of the I-shaped arrangement of the building group and the matching relationship of the corresponding buildings at different scales; wherein the spatial relationship of the I-shaped arrangement of the building group includes: adjacent relationship, completely parallel relationship, completely vertical relationship, relative position relationship, and center distance constraint; the matching relationship of the corresponding buildings at different scales includes: one-to-one, one-to-many, many-to-one, many-to-many, one-to-zero and zero-to-one; Step S2: Based on the building knowledge graph, rule-based reasoning is used to identify I-shaped building patterns at a single spatial scale that conforms to human visual recognition principles and I-shaped building patterns supported by multi-scale data.

2. A device for identifying an I-shaped building group pattern, characterized in that: include: A construction module is used to construct a building knowledge graph based on the spatial relationship of the I-shaped arrangement of building groups and the matching relationship of corresponding buildings at different scales; wherein the spatial relationship of the I-shaped arrangement of building groups includes: proximity relationship, complete parallel relationship, complete vertical relationship, relative position relationship, and center distance constraint; the matching relationship of corresponding buildings at different scales includes: one-to-one, one-to-many, many-to-one, many-to-many, one-to-zero, and zero-to-one; The recognition module is used to identify the I-shaped building pattern at a single spatial scale that conforms to the human visual recognition principle and the I-shaped building pattern supported by multi-scale data through rule-based reasoning based on the building knowledge graph.

3. A I-shaped building group pattern recognition system, characterized in that: include: A memory and a processor, wherein the memory stores a computer program executed by the processor, and when the computer program is executed by the processor, the I-shaped building group pattern recognition method as claimed in claim 1 is executed.

4. A storage medium, characterized in that: The storage medium stores a computer program, which executes the I-shaped building group pattern recognition method as claimed in claim 1 when running.

Images