An automatic identification method of urban shadow area based on knowledge graph

Abstract

The application discloses a kind of city shadow area automatic identification method based on knowledge graph, including shadow area case base construction, shadow area feature recognition model construction, target city graph construction, target city shadow area graph subset search, target city shadow area identification result output and show five parts.This method integrates present city data through knowledge graph, accurately identifies city shadow area and carries out fine display to identification result.The application can cope with city shadow area automatic identification in the field of city planning, the identification accuracy is higher based on super large city case base, the identification speed is faster based on shadow area graph subset search, the three-dimensional interactive display formed is more three-dimensional and intuitive, and the guidance to city planning project is stronger.

Description

Technical Field

[0001] The present invention relates to the field of urban planning, and specifically to a method for automatically identifying urban shadow areas based on a knowledge graph. Background Art

[0002] Urban shadow zones are the product of unbalanced and inadequate development within cities. They are characterized by low-rise, disorderly buildings, fragmented and inadequate public facilities, outdated and fragmented architectural features, and a low-end, scattered presence of businesses. Research on urban shadow zones can serve as a valuable case study for exploring issues such as unbalanced urban development, social spatial differentiation, and the failure of urban planning. Further in-depth research can reveal the inherent logic and mechanisms of urban development, providing theoretical support and empirical evidence for future urban planning.

[0003] Currently, most shadow zone identification methods still rely on the calculation and summary of indicators based on geographic environment, economic development level, and social and cultural conditions. However, due to the complex and diverse causes of shadow zones, it is difficult to accurately judge them based on only a few indicators. Furthermore, the selection of specific indicators is subject to individual subjective perceptions. Consequently, shadow zone identification suffers from limited summary efficiency and accuracy, making traditional shadow zone identification methods less scientific and efficient, and hindering the further development and optimization of shadow zone areas. Summary of the Invention

[0004] In order to address the deficiencies mentioned in the above-mentioned background technology, the purpose of the present invention is to provide a method for automatic identification of urban shadow areas based on knowledge graphs. The present invention can integrate current urban data through knowledge graphs, accurately identify urban shadow areas and display the identification results in a refined manner.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A method for automatically identifying urban shadow areas based on a knowledge graph includes the following steps:

[0007] Step 1: Construction of the shadow area case library

[0008] Spatial morphological data, business data, and population flow data were obtained for seven megacities nationwide. These megacities, including Shanghai, Beijing, Shenzhen, Chongqing, Guangzhou, Chengdu, and Tianjin, were obtained from the "Statistical Charts on Economic and Social Development: Basic Population Situation of Megacities and Supercities from the Seventh National Population Census," published by the National Bureau of Statistics. A quantitative urban shadow area identification method was used to identify shadow areas in the central areas of the seven megacities. Boundary vector data for each central shadow area was obtained for each megacity. Corresponding spatial morphological data, business data, and population data were extracted based on the boundaries of each shadow area to construct a shadow area case database.

[0009] Step 2: Construction of shadow area feature recognition model

[0010] Combined with the shadow area case library obtained in step S1, spatial morphological elements (including blocks, roads, and open spaces), business elements, and pedestrian flow elements are used as entities in the knowledge graph. A knowledge graph operating system is used to construct each shadow area map. Using a supervised clustering learning algorithm and a deep learning system, the entities and attributes, as well as relationships and attributes, of the shadow area map are used as machine learning labels to generate a shadow area feature recognition model, which is then further identified to obtain a shadow area feature map.

[0011] Step 3: Target city map construction

[0012] Retrieve the current status data of the planning bureau of the target city, integrate the spatial form data, business data and pedestrian flow data respectively, and use the knowledge graph operating system described in step S2 to construct the target city graph.

[0013] Step 4: Search for a subset of the target city's shadow area atlas

[0014] A graph convolutional network model was selected and trained using graph embedding algorithms on both the target city graph and the shadow feature graph. This algorithm learned how to map graph structure data into a low-dimensional vector space, generating an embedded representation of each node in the low-dimensional vector space. A similarity calculation method was used to calculate the feature similarity and relationship similarity between each subgraph of the target city graph and the shadow feature graph in the low-dimensional vector space. The closer the similarity is to 1, the more similar the target city graph subgraph and the shadow feature graph are in the embedding space, meaning their graph structures and relationships are more similar to those of the shadow feature graph. First, the feature similarity between the target city graph and the shadow feature graph was calculated to identify potential shadow subgraphs with similar attributes to the shadow feature graph. Next, the relationship similarity between the potential shadow subgraphs and the shadow feature graph was calculated to identify shadow subgraphs with similar network structures to the shadow feature graph and label them as the target city shadow map subset.

[0015] Step 5: Output and display of target city shadow area identification results

[0016] The target city shadow area recognition results are output to the 3D interactive display device, and the above content is integrated into a recognition report and printed in Word / Excel format.

[0017] Further, the urban shadow area quantitative identification method in step one refers to taking a block as the basic unit of the spatial boundary of the urban shadow area, and calculating the indexes of each block unit in terms of spatial location, construction intensity, industry layout, and crowd vitality according to a table (Table 1). The geographic information system is used to spatially position and superimpose the above-mentioned element analysis results to obtain a comprehensive index distribution. The relative low-value blocks located in the last 10% of the comprehensive index distribution form a preliminary boundary definition result. The adjacent blocks outside each block area are identified, and if the block is located in the last 20% of the identification result, it is included in the shadow area range. Further, the blocks in each block area are field surveyed, and the blocks with natural elements and public buildings are removed according to the survey results.

[0018] Table 1 Identification of spatial boundary of shadow area

[0019]

[0020] Further, the step two of constructing a shadow area graph through a knowledge graph operation system is the innovative step of the patent, which refers to first constructing a graph mode layer. Defining the entity elements and their attributes, entity relationships and their attributes (Table 2), and defining the structure and hierarchical relationship between entities, attributes, and relationships (Table 3); secondly, constructing a graph data layer. The spatial form data, industry data, and crowd data in the database are data cleaned, and the cleaned data is arranged into a structured data format of triplets; finally, the structured data of the data layer is imported into a graph database management system, the relationship between entities is established according to the definition of the mode layer, and thus the graph construction is completed.

[0021] Table 2 Graph entities and their attributes

[0022]

[0023]

[0024] Table 3 Graph relationships and their attributes

[0025]

[0026] Furthermore, the shadow area feature recognition model in step 2 is an innovative step of this patent. It refers to using the entities and attributes, relationships, and properties of the shadow area map as machine learning labels, dividing the shadow area map into training, validation, and test sets in a ratio of 6:2:2. The operator selects entities and relationships related to the characteristics of each shadow area on the knowledge graph visualization platform, and uses an eye tracker with 0.4° gaze tracking accuracy to assist in obtaining the operator's selection tendency. A supervised clustering learning algorithm is used to perform shadow area map feature extraction learning and training through a 512GB video memory deep learning system. A machine learning model with strong generalization performance is selected as the shadow area feature recognition model, and the identified entity elements and attributes, entity relationships, and attributes are output to obtain the shadow area feature map.

[0027] Furthermore, the feature similarity calculation in step 4 is an innovative step of this patent, which refers to extracting the attributes of the target city map subgraph and the shadow feature map, where the attributes are the pattern layer attributes of the shadow feature map and the target city map subgraph defined in steps 1 and 4. The cosine similarity measurement method is used to calculate and compare the feature similarity between the two maps. The formula is as follows:

[0028]

[0029] Where A is the embedded representation of the target city map subgraph in the low-dimensional vector space, and B is the embedded representation of the shadow feature map in the low-dimensional vector space.

[0030] Calculate the feature similarity between all n sub-graphs of the target city graph and the shadow feature graph, that is, calculate Similarity(A1,B), Similarity(A2,B)…Similarity(A n ,B). In the interval [-1,1], sort the feature similarities from large to small, select the top 10% feature similarities, if the top 10% feature similarities are all greater than 0, then mark the corresponding sub-graph as a potential shadow area sub-graph with similar attribute features to the shadow area feature map; otherwise, remove the sub-graphs with feature similarities less than or equal to 0, and mark the remaining corresponding sub-graphs as potential shadow area sub-graphs with similar attribute features to the shadow area feature map.

[0031] Furthermore, the calculation of the relationship similarity in step 4 is an innovative step of this patent, which refers to converting the potential shadow area subgraph and the shadow feature map into the form of an adjacency matrix, and using the subgraph isomorphism algorithm to calculate and compare the relationship similarity between the two maps. The formula is as follows:

[0032]

[0033] Among them G iis the potential shadow subgraph, G is the shadow feature map, V1 is the node set of the potential shadow subgraph, Similarity(v,G(v)) is the similarity between node v and the corresponding node G(v) in the shadow feature map. The higher the score, the more similar the potential shadow subgraph is to the shadow feature map.

[0034] Calculate the similarity between all m sub-images of the potential shadow area and the shadow feature map, that is, calculate score(G1,G), score(G2,G)…score(G m ,G), sort all relationship similarities from large to small, select the top 10% similarities, and mark their corresponding subgraphs as the shaded area graph subset of the target city.

[0035] Furthermore, the display of the target city shadow area identification results in step five refers to the display of the graph structure of three graphs, namely the shadow area feature graph, the target city graph, and the target city shadow area graph subset, in the form of a graph network, wherein the graph structure includes the network structure of each entity, attribute and relationship, and displays the feature similarity and relationship similarity between the shadow area graph subset of each target city and the shadow area feature graph.

[0036] Beneficial effects of the present invention:

[0037] 1. Obtain spatial morphological data, business data, and pedestrian flow data for megacities. Utilize a quantitative urban shadow zone identification method to identify the boundary vector data of shadow zones in each central area. Combine these four types of data to construct a shadow zone case library, laying the foundation for the subsequent construction of a shadow zone feature recognition model.

[0038] 2. Through the knowledge graph operating system, a map of each shadow area is constructed. A supervised clustering learning algorithm and a deep learning system are used to train shadow area features, thereby establishing a shadow area feature recognition model. This model provides a standard and reliable reference for shadow area feature recognition.

[0039] 3. Combining feature similarity and relationship similarity algorithms, we can realize the search function of the target city's shadow area map subset, screen and identify the subgraphs in the target city map whose spatial form and relationship network are similar to the shadow area feature map, thereby realizing the automatic identification of the target city's shadow area. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flow chart of the method of the present invention.

[0041] Figure 2 It is the flow chart of shadow area feature recognition. DETAILED DESCRIPTION

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making any creative efforts shall fall within the scope of protection of the present invention.

[0043] A method for automatic identification of urban shadow areas based on knowledge graph, such as Figure 1 and Figure 2 As shown, the following steps are included:

[0044] 1. Construction of a Shadow Area Case Library. Spatial morphological data, business data, and population flow data were obtained for seven megacities nationwide. These megacities were sourced from the "Statistical Charts on Economic and Social Development: Basic Population Situation of Megacities and Supercities from the Seventh National Population Census," published by the National Bureau of Statistics. These cities include Shanghai, Beijing, Shenzhen, Chongqing, Guangzhou, Chengdu, and Tianjin. Using a quantitative urban shadow area identification method, shadow areas in the central areas of the seven megacities were identified. Boundary vector data for the shadow areas in each megacity's central area were obtained. Corresponding spatial morphological data, business data, and population data were extracted based on the boundaries of each shadow area to construct a shadow area case library.

[0045] The quantitative identification method for urban shadow zones is as follows: blocks are used as the basic units for defining the spatial boundaries of urban shadow zones. Indicators for each block unit in terms of spatial location, construction intensity, business layout, and population vitality are calculated according to a table (Table 1). The results of the above factor analysis are spatially aligned and superimposed using a geographic information system to obtain a comprehensive index distribution. The area consisting of blocks with relatively low values ​​in the bottom 10% of the comprehensive index is used as the preliminary boundary definition result. Adjacent blocks outside each area are identified. If the block is in the bottom 20% of the identification result, it is included in the shadow zone. Further field surveys are conducted on each area, and based on the survey results, natural elements and blocks containing public buildings are eliminated.

[0046] Table 1 Recognition of spatial boundaries of shadow areas

[0047]

[0048]

[0049] 2. Construction of a shadow area feature recognition model. Combining the shadow area case library obtained during the construction of the shadow area case library, spatial morphological elements, business elements, and pedestrian flow elements are used as entities in the knowledge graph. Each shadow area map is constructed through the knowledge graph operating system. Using a supervised clustering learning algorithm and a deep learning system, the entities and attributes, as well as relationships and attributes of the shadow area map, are used as machine learning labels to generate a shadow area feature recognition model, which is then further identified to obtain a shadow area feature map.

[0050] The detailed process of constructing the shaded area maps using the knowledge graph operating system is as follows: First, the graph schema layer is constructed. The graph's entity elements and their attributes, entity relationships and their attributes are defined (Table 2), and the structure and hierarchical relationships between entities, attributes, and relationships are defined (Table 3). Second, the graph data layer is constructed. The spatial morphology data, business data, and population data in the database are cleaned and organized into a structured data format of triples. Finally, the structured data in the data layer is imported into the graph database management system, and the relationships between entities are established according to the definitions in the schema layer, thus completing the graph construction.

[0051] Table 2 Graph entities and their attributes

[0052]

[0053]

[0054] Table 3 Graph relationships and their attributes

[0055]

[0056] The shadow area feature recognition model is specifically designed as follows: The entities and attributes, as well as relationships and properties, of the shadow area map are used as machine learning labels. The shadow area map is divided into training, validation, and test sets in a 6:2:2 ratio. Operators select entities and relationships related to each shadow area's characteristics on a knowledge graph visualization platform. An eye tracker with 0.4° gaze tracking accuracy is used to assist in capturing the operator's selection preferences. A supervised clustering learning algorithm is used to train and extract shadow area map features using a deep learning system with 512GB of video memory. A machine learning model with strong generalization performance is selected as the shadow area feature recognition model. The identified entity elements and attributes, entity relationships, and properties are output to form a shadow area feature map.

[0057] 3. Target city map construction: Retrieve the current status data of the target city planning bureau, integrate the spatial form data, business data, and pedestrian flow data, and use the knowledge graph operating system described in the construction of the shadow area feature recognition model to construct the target city map.

[0058] 4. Finding subsets of the target city's shadow area graph. A graph convolutional network model is selected and a graph embedding algorithm is applied to both the target city graph and the shadow area feature graph. These models are trained to learn how to map graph structure data into a low-dimensional vector space, generating an embedded representation of each node in the low-dimensional vector space. Using a similarity calculation method, the feature similarity and relationship similarity between each subgraph of the target city graph and the shadow area feature graph are calculated in the low-dimensional vector space. The closer the similarity is to 1, the more similar the target city graph subgraph and the shadow area feature graph are in the embedding space, i.e., their graph structures and relationships are more similar to the shadow area feature graph. First, the feature similarity between the target city graph and the shadow feature graph is calculated to screen out potential shadow area subgraphs with similar attributes to the shadow area feature graph. Next, the relationship similarity between the potential shadow area subgraphs and the shadow feature graph is calculated to screen out shadow area subgraphs with similar network structures to the shadow area feature graph and mark them as subsets of the target city's shadow area graph.

[0059] The method for calculating feature similarity screening is as follows: for the target city map subgraph and the shadow feature map, their attributes are extracted, where the attributes are the pattern layer attributes of the shadow feature map and the target city map subgraph defined in steps S2 and S41. The cosine similarity measurement method is used to calculate and compare the feature similarity between the two maps. The formula is as follows:

[0060]

[0061] Where A is the embedding representation of the target city map subgraph in the low-dimensional vector space, and B is the embedding representation of the shadow feature map in the low-dimensional vector space. Calculate the feature similarity between all n subgraphs of the target city map and the shadow feature map, that is, calculate Similarity(A1,B), Smilarity(A2,B)…Smilarity(A n ,B). In the interval [-1,1], sort the feature similarities from large to small, select the top 10% feature similarities, if the top 10% feature similarities are all greater than 0, then mark the corresponding sub-graph as a potential shadow area sub-graph with similar attribute features to the shadow area feature map; otherwise, remove the sub-graphs with feature similarities less than or equal to 0, and mark the remaining corresponding sub-graphs as potential shadow area sub-graphs with similar attribute features to the shadow area feature map.

[0062] The method for calculating relationship similarity screening is as follows: the potential shadow area subgraph and the shadow feature map are converted into the form of an adjacency matrix, and the subgraph isomorphism algorithm is used to calculate and compare the relationship similarity between the two maps. The formula is as follows:

[0063]

[0064] Among them Gi is the potential shadow subgraph, G is the shadow feature map, V1 is the node set of the potential shadow subgraph, Similarity(v,G(v)) is the similarity between node v and the corresponding node G(v) in the shadow feature map. The higher the score, the more similar the potential shadow subgraph is to the shadow feature map. Calculate the similarity between all m subgraphs of the potential shadow subgraph and the shadow feature map, that is, calculate score(G1,G), score(G2,G)…score(G m ,G), sort all relationship similarities from large to small, select the top 10% similarities, and mark their corresponding subgraphs as the shaded area graph subset of the target city.

[0065] 5. Output and display of target city shadow area identification results. Output the target city shadow area identification results to a 3D interactive display device, and integrate the above content into an identification report and print it in Word / Excel format.

[0066] The target city shadow area identification results are displayed specifically as follows: the graph structure of three graphs, namely the shadow area feature graph, the target city graph, and the target city shadow area graph subset, is displayed in the form of a graph network, wherein the graph structure includes the network structure of each entity, attribute and relationship, and the feature similarity and relationship similarity between the shadow area graph subset of each target city and the shadow area feature graph are displayed.

Claims

1. A method for automatically identifying urban shadow areas based on knowledge graph, characterized in that: The method comprises the following steps: Step S1: Construction of shadow area case library Obtain spatial morphological data, business data, and crowd flow data for seven megacities across China: Shanghai, Beijing, Shenzhen, Chongqing, Guangzhou, Chengdu, and Tianjin. Use the quantitative urban shadow area identification method to identify shadow areas in the central areas of these seven megacities, obtaining boundary vector data for each central shadow area of ​​each megacity. Combined with the boundaries of each shadow area, extract corresponding spatial morphological data, business data, and crowd data to construct a shadow area case library. Step S2: Shadow area feature recognition model construction Combined with the shadow area case library obtained in step S1, spatial morphological elements, business elements, and pedestrian flow elements are used as entities of the knowledge graph, and each shadow area graph is constructed through the knowledge graph operating system; using a supervised clustering learning algorithm and a deep learning system, the entities and attributes, relationships, and attributes of the shadow area graph are used as machine learning labels to generate a shadow area feature recognition model, and further identify and obtain the shadow area feature graph; Step S3: Target city map construction Retrieve the current status data of the target city planning bureau, integrate the spatial form data, business data and pedestrian flow data respectively, and use the knowledge graph operating system described in step S2 to construct the target city graph; Step S4: Searching for a subset of the target city's shadow area atlas Select a graph convolutional network model, apply a graph embedding algorithm to the target city map and the shadow area feature map, and map the graph network structure into a computable low-dimensional vector space. Use a feature similarity algorithm to perform a similarity test between the target city map obtained in step S3 and the shadow area feature map obtained in step S2, and screen out potential shadow area subgraphs with similar attribute features to the shadow area feature map. Then, use a relational similarity algorithm to screen out shadow area subgraphs with similar network structures to the shadow area feature map obtained in step S2 from the potential shadow area subgraphs, and mark them as the target city shadow area map subset. Step S5: Output and display of target city shadow area identification results Output the target city shadow area recognition results to the 3D interactive display device, and integrate the above content into a recognition report and print it in Word / Excel format; The specific method for quantitatively identifying urban shadow areas in step S1 is as follows: Using blocks as the basic unit for defining the spatial boundaries of the urban shadow zone, we calculated indicators for each block unit in terms of spatial location, construction intensity, business layout, and population vitality (see Table 1). We used a geographic information system to spatially align and overlay the above factor analysis results to obtain a comprehensive index distribution. The districts consisting of blocks with relatively low values ​​in the bottom 10% of the comprehensive index were used as preliminary boundary demarcation results. We identified the adjacent blocks outside each district, and if the block was in the bottom 20% of the identified results, it was included in the shadow zone. We further conducted on-site surveys of each district, and based on the survey results, we eliminated natural elements and blocks containing public buildings. Table 1 Recognition of spatial boundaries of shadow areas 2. The method for automatically identifying urban shadow areas based on knowledge graph according to claim 1 is characterized in that: The specific steps of constructing the graphs of each shadow area through the knowledge graph operating system in step S2 are as follows: First, construct the graph model layer; define the graph's entity elements and their attributes, entity relationships and their attributes, see Table 2; define the structure and hierarchical relationship between entities, attributes and relationships, see Table 3; Secondly, we build a graph data layer; we clean the spatial morphology data, business data, and population data in the database, and organize the cleaned data into a structured data format of triples; Finally, the structured data in the data layer is imported into the graph database management system, and the relationships between entities are established according to the definition of the schema layer, thus completing the graph construction; Table 2 Graph entities and their attributes Table 3 Graph relationships and their attributes 3. The method for automatically identifying urban shadow areas based on knowledge graph according to claim 2 is characterized in that: The shadow area feature recognition model described in step S2 is specifically as follows: The entities and attributes, relationships and properties of the shadow area map are used as machine learning labels, and the shadow area map is divided into training set, validation set, and test set in a ratio of 6:2:

2. The operator selects entities and relationships related to the characteristics of each shadow area on the knowledge graph visualization platform, and the operator's selection tendency is obtained with the assistance of an eye tracker with 0.4° gaze tracking accuracy; a supervised clustering learning algorithm is used to perform shadow area map feature extraction learning and training through a 512GB video memory deep learning system, and a machine learning model with strong generalization performance is selected as the shadow area feature recognition model, and the identified entity elements and attributes, entity relationships and attributes are output to obtain the shadow area feature map.

4. The method for automatically identifying urban shadow areas based on knowledge graph according to claim 3 is characterized in that: Step S4 includes the following steps: S41 target city map and shadow area map embedding A graph convolutional network model was selected and a graph embedding algorithm was applied to the target city map and the shadow area feature map. The model was trained to learn how to map graph structure data into a low-dimensional vector space, thereby generating an embedded representation of each node in the low-dimensional vector space. S42 Find the shadow area atlas subset of the target city Using a similarity calculation method, the feature similarity and relationship similarity between each subgraph of the target city graph and the feature graph of the shadow area are calculated in a low-dimensional vector space. The closer the similarity is to 1, the more similar the directions of the target city graph subgraph and the feature graph of the shadow area in the embedding space are, that is, the more similar their graph structures and relationships are to the feature graph of the shadow area. First, the feature similarity between the target city map and the shadow feature map is calculated, and the potential shadow area subgraphs with similar attribute features to the shadow area feature map are screened out; then, the relationship similarity between the potential shadow area subgraphs and the shadow feature map is calculated, and the shadow area subgraphs with similar network structures to the shadow area feature map are screened out and marked as the shadow area map subset of the target city.

5. The method for automatically identifying urban shadow areas based on knowledge graph according to claim 4 is characterized in that: The method for calculating feature similarity screening in step S42 is as follows: For the target city map subgraph and the shadow feature map, their attributes are extracted, where the attributes are the pattern layer attributes of the shadow feature map and the target city map subgraph defined in step S2 and step S41; the cosine similarity measurement formula is used to calculate and compare the feature similarity between the two maps; the formula is as follows: Where A is the embedded representation of the target city map subgraph in the low-dimensional vector space, and B is the embedded representation of the shadow feature map in the low-dimensional vector space; Calculate the feature similarity between all n sub-graphs of the target city graph and the shadow feature graph, that is, calculate Similarity(A1,B), Similarity(A2,B)…Similarity(A n ,B); In the interval [-1,1], sort the feature similarities from large to small, select the top 10% feature similarities, if the top 10% feature similarities are all greater than 0, then mark the corresponding subgraph as a potential shadow area subgraph with similar attribute features to the shadow area feature map; otherwise, remove the feature similarities less than or equal to 0, and mark the remaining corresponding subgraphs as potential shadow area subgraphs with similar attribute features to the shadow area feature map.

6. The method for automatically identifying urban shadow areas based on knowledge graph according to claim 5 is characterized in that: The method for calculating the relationship similarity screening in step S42 is as follows: The potential shadow area subgraph and the shadow feature map are converted into the form of an adjacency matrix, and the subgraph isomorphism algorithm is used to calculate and compare the similarity between the two maps; the formula is as follows: Among them G i is the potential shadow subgraph, G is the shadow feature map, V1 is the node set of the potential shadow subgraph, Similarity(v,G(v)) is the similarity between node v and the corresponding node G(v) in the shadow feature map. The higher the score, the more similar the potential shadow subgraph is to the shadow feature map. Calculate the similarity between all m sub-images of the potential shadow area and the shadow feature map, that is, calculate score(G1,G), score(G2,G)…score(G m ,G), sort all relationship similarities from large to small, select the top 10% similarities, and mark their corresponding subgraphs as the shaded area graph subset of the target city.

7. The method for automatically identifying urban shadow areas based on knowledge graph according to claim 6 is characterized in that: The target city shadow area identification result in step S5 is shown as follows: The graph structures of the shadow area feature map, the target city map, and the shadow area map subset of the target city are displayed in the form of a graph network, wherein the graph structure includes the network structure of each entity, attribute, and relationship, and displays the feature similarity and relationship similarity between the shadow area map subset of each target city and the shadow area feature map.

Images