Multi-source geographic space big data knowledge graph construction method and system for intelligent city planning

Abstract

The invention discloses a multi-source geographic space big data knowledge graph construction method and system for intelligent city planning, and the method comprises the steps: collecting original data of a vector layer, a grid layer, a text layer and an image layer, and carrying out the preprocessing, and obtaining vector data, grid data, text data and image data; constructing a knowledge graph ontology layer; and according to the set and the alignment mode constructed in the knowledge graph ontology layer, utilizing a DeepLabV3 + semantic segmentation model, an NLP model and a GIS platform to extract respective entity sets, relationship sets, attribute sets and alignment modes in a vector layer, a grid layer, a text layer and an image layer in the target city based on a software platform, and forming the multi-source geographic space big data knowledge graph. The heterogeneous data are efficiently integrated and accurately aligned through the unified reference entity, and the defects of a traditional method in the aspects of data integration efficiency, entity alignment precision and intelligent analysis capacity are overcome.

Description

Technical Field

[0001] The present invention belongs to the field of knowledge graphs and smart city technologies, and in particular, relates to a method and system for constructing a multi-source geospatial big data knowledge graph for smart city planning. Background Art

[0002] With the rapid advancement of global urbanization, the construction of smart cities has put forward higher requirements for urban planning. However, traditional urban planning tools and methods, such as geographic information systems (GIS) and remote sensing technology, face major challenges in the integration of multi-source heterogeneous geographic data. These data sources (such as remote sensing images, sensor networks, and socioeconomic data) have significant differences in spatial resolution, timeliness, and data types, which makes it difficult to effectively apply them in smart city planning with high real-time and accuracy requirements. In addition, the current multi-source geographic data processing still has problems such as low efficiency and high complexity, which to a certain extent limits its actual effectiveness in dynamic decision-making and urban planning. Therefore, how to achieve efficient integration, accurate analysis, and intelligent application of multi-source heterogeneous geographic data has become a technical bottleneck that needs to be urgently solved in smart city planning.

[0003] In response to the above problems, Geographic Knowledge Graph (GeoKG), as an emerging technical means, realizes deep integration and intelligent analysis of data by converting multi-source geographic information into structured knowledge forms. GeoKG can effectively integrate multiple data types such as vector data, raster data, text data and image data, and use semantic modeling technology to accurately describe geographic entities and their relationships, providing a new data support framework for smart city planning. However, GeoKG currently still faces many challenges in practical applications, especially in terms of data accuracy, entity alignment and seamless connection of multimodal data. How to overcome these problems through efficient modeling and analysis methods is a key direction for promoting the intelligent development of smart city planning. Summary of the invention

[0004] The purpose of the present invention is to address the problems existing in the prior art and to provide a method and system for constructing a multi-source geospatial big data knowledge graph for smart city planning, which overcomes the shortcomings of traditional methods in data integration efficiency, entity alignment accuracy and intelligent analysis capabilities by comprehensively integrating and accurately aligning heterogeneous data.

[0005] According to one aspect of the present invention, a method for constructing a multi-source geospatial big data knowledge graph for smart city planning is provided, comprising:

[0006] Collect the original data of vector layer, raster layer, text layer and image layer, and pre-process them to obtain vector data V , Raster Data R, text data Data T and image data Data I ;

[0007] Construct the ontology layer of the knowledge graph. The construction of the ontology layer of the knowledge graph includes: constructing a vector entity set E V based on vector data Data v , a vector attribute set A v and a vector relationship set R v , constructing a raster entity set E R based on raster data Data r , a raster attribute set A r and a raster relationship set R r , constructing a text entity set E T based on text data Data t , a text attribute set A t and a text relationship set R t , constructing an image entity set E I based on image data Data i , an image attribute set A i and an image relationship set R i , aligning the vector entity set E v , the raster entity set E r , the text entity set E t and the image entity set E i using the street name as a reference entity to obtain an entity set E = {E v , E r , E t , E i}, and determining the attribute set A and the relationship set R according to the entity elements in the obtained entity set E, the positional relationships of the entity elements, and the attributes of the entity elements;

[0008] According to the sets and alignment methods constructed in the ontology layer of the knowledge graph, using the DeepLabV3+ semantic segmentation model, the NLP model, and the GIS platform, extract the entity sets, relationship sets, attribute sets, and alignment methods in the vector layer, raster layer, text layer, and image layer in the target city based on the software platform to form a multi-source geospatial big data knowledge graph.

[0009] The above technical solution efficiently integrates and precisely aligns heterogeneous data through a unified reference entity, overcoming the deficiencies of traditional methods in data integration efficiency, entity alignment accuracy, and intelligent analysis capabilities. This method combines the DeepLabV3+ semantic segmentation model, natural language processing technology, and GIS spatial analysis tools to ensure the automation and efficiency of data processing; utilizes the powerful storage and query capabilities of the graph database to provide a new data management and analysis framework for smart city planning, demonstrating significant application effects in aspects such as urban road planning, greening coverage analysis, and environmental quality assessment, providing strong technical support for scientific decision-making in smart cities.

[0010] As a further technical means, the vector data Data V is derived from data extracted from Open Street Map and other open source datasets in the research area according to the vector data extraction principle, and the vector data extraction principle includes: all data containing at least roads and administrative regions in the research area.

[0011] As a further technical means, constructing the vector entity set, vector attribute set, and vector relationship set includes:

[0012] Extracting road names, OSM identifiers, and administrative divisions to form the vector entity set E v ;

[0013] It should be noted that OSM is the abbreviation of OpenStreetMap, which is a global open source mapping project where users can freely edit and use the geographical data. The data structure design of OSM is for flexible and efficient representation of geographical information, mainly composed of three basic elements: points (Node), lines (Way), and relations (Relation). This data can be used for various applications, including but not limited to urban planning, traffic navigation, and geographic information systems. OSM data contains various types of data such as road and railway road networks, buildings, water bodies, land use, points of interest, administrative division boundaries, etc.

[0014] Calculating the topological attributes and geometric attributes of each entity element in the vector entity set E v to form the vector attribute set A v , where the topological attributes at least include adjacent roads and road lengths, and the geometric attributes at least include longitude and latitude;

[0015] Establishing the spatial and attribute relationships of the category, length, nearest road, Euclidean distance, and administrative division of each entity element in the vector entity set E v to form the vector relationship set R v .

[0016] As a further technical means, the raster data Data R is derived from remote sensing data extracted from the study area according to the raster data extraction principles, and the raster data extraction principles include: the requirements of the actual application scenario and the existing data of the platform.

[0017] As a further technical means, constructing the raster entity set, raster attribute set, and raster relationship set includes:

[0018] Generating sampling point IDs along the roads in the study area and combining with the time field to form the raster entity set E r ;

[0019] Extracting the corresponding data types from the remote sensing data in the study area according to the raster data extraction principles to form the raster attribute set A r ;

[0020] Extracting the corresponding data types from the remote sensing data in the study area according to the raster data extraction principles, and establishing the geographical location relationships of the individual entity elements in the raster entity set E r to form the raster relationship set R r .

[0021] As a further technical means, the text data Data T is derived from the data extracted from the store reviews in the study area according to the text data extraction rules, and the text data extraction rules include: at least being able to be segmented into longitude and latitude by natural language processing technology.

[0022] As a further technical means, constructing the text entity set, text relationship set, and text attribute set includes:

[0023] Extracting the store names in the study area from the text data Data T to form the text entity set E t ;

[0024] Extracting the text attributes of the individual text entity elements in the text entity set E T from the text data Data t to form the text attribute set A t ;

[0025] According to the store names in the text data Data T , constructing the geographical relationships between the stores and the nearest roads, administrative divisions where they are located, and longitude and latitude to form the text relationship set R t .

[0026] As a further technical means, the image data Data IData from sampling points along the streets within the research area, where the data from sampling points along the streets within the research area includes the green view rate, building view rate, and vehicle view rate extracted from the 360° panoramic images of the streets.

[0027] As a further technical means, constructing the image entity set, image attribute set, and image relationship set includes:

[0028] Based on the sampling point IDs along the roads within the research area, marking the geographical entities in the street view images to form the image entity set E i ;

[0029] Using the deep learning model (DeepLabV3+) to extract the green view rate, building view rate, and vehicle view rate within the research area to form the image attribute set A i ;

[0030] Establishing the geographical location relationships among the sampling points along the streets, the roads they are on, administrative regions, and longitude and latitude to form the image relationship set R i .

[0031] According to one aspect of the specification of the present invention, there is provided a multi-source geospatial big data knowledge graph construction system for smart city planning, including:

[0032] A data collection and processing module for collecting the original data of the vector layer, raster layer, text layer, and image layer and performing preprocessing;

[0033] An ontology layer construction module for constructing a vector entity set E v , a vector relationship set A v , and a vector attribute set R v according to the preprocessed vector layer, constructing a raster entity set E r , a raster relationship set A r , and a raster attribute set R r according to the preprocessed raster layer, constructing a text entity set E t , a text relationship set A t , and a text attribute set R t according to the preprocessed text layer, constructing an image entity set E i , an image relationship set A i , and an image attribute set R i , and aligning the vector entity set E v , the raster entity set E r , the text entity set E t , and the image entity set E i through a unified reference entity to obtain the entity set E = {E v,E r ,E t ,E i}, according to the obtained entity set E, the relationship set A and the attribute set R are obtained to form the ontology layer of the knowledge graph;

[0034] The knowledge graph construction module is used to extract the entity set, relationship set, attribute set and alignment method of each data layer in the target city based on the set and alignment method of the ontology layer of the knowledge graph, using the DeepLabV3+ semantic segmentation model, NLP model and GIS platform based on the ArcGIS Pro, Google Earth Engine and PyCharm platforms, to form a multi-source geospatial big data knowledge graph.

[0035] Compared with the prior art, the beneficial effects of the present invention are:

[0036] It realizes the efficient integration and precise alignment of heterogeneous data, and overcomes the deficiencies of traditional methods in data integration efficiency, entity alignment accuracy and intelligent analysis ability. This method ensures the automation and efficiency of data processing by combining the DeepLabV3+ semantic segmentation model, natural language processing technology and GIS spatial analysis tools; uses the powerful storage and query capabilities of the Neo4j graph database to provide a new data management and analysis framework for smart city planning. Through the actual case verification in Wuhan City, the present invention shows significant application effects in aspects such as urban road planning, green coverage rate analysis and environmental quality assessment, providing strong technical support for the scientific decision-making of smart cities. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is the method flow chart of the embodiment of the present invention; Figure 2 is the research area map of the test data of the invention example; Figure 3 is the data preprocessing process of the first step of the invention example; Figure 4 is the process of constructing the sets of entities, relationships and attributes by combining the characteristics of each data layer and processing the alignment method of each data layer in the second step of the invention example; Figure 5 is a partial screenshot of the knowledge graph constructed in the invention example; Figure 6 is the effect diagram a of applying the knowledge graph constructed in the invention example to the urban heat island effect; Figure 7 is the effect diagram b of applying the knowledge graph constructed in the invention example to the urban heat island effect; Figure 8 is the effect diagram c of applying the knowledge graph constructed in the invention example to the urban heat island effect; Figure 9 Effect diagram of applying the knowledge graph constructed in the embodiment of the present invention to the analysis of urban green view rate and building density; Figure 10 Effect diagram of applying the knowledge graph constructed in the embodiment of the present invention to the urban travel recommendation scenario; Figure 11 Effect diagram of applying the knowledge graph constructed in the embodiment of the present invention to the analysis of road network and traffic flow; Figure 12 Effect diagram of applying the knowledge graph constructed in the embodiment of the present invention to the detection of urban land use change. Detailed implementation manners

[0038] Next, the technical solutions of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0039] As Figure 1 shown, a method for constructing a multi-source geospatial big data knowledge graph for smart city planning includes:

[0040] Step 1: Collect the original data of the vector layer, raster layer, text layer, and image layer, and perform preprocessing to obtain vector data Data V , raster data Data R , text data Data T , and image data Data I , and the processing process is as Figure 3 shown;

[0041] As Figures 4 - 5 shown, step 2: Construct the ontology layer of the knowledge graph. Constructing the ontology layer of the knowledge graph includes: constructing a vector entity set E V , a vector attribute set A v , and a vector relationship set R v according to the vector data Data v , constructing a raster entity set E R , a raster attribute set A r , and a raster relationship set R r according to the raster data Data r , constructing a text entity set E T , a text attribute set A t , and a text relationship set R tand the text relationship set R t , construct the image entity set E based on the image data Data I i , the image attribute set A i and the image relationship set R i , using RoadName (road name) as the reference entity, align the vector entity set E v , the raster entity set E r , the text entity set E t and the image entity set E i to obtain the entity set E = {E v , E r , E t , E i}, determine the attribute set A and the relationship set R according to the entity elements in the obtained entity set E, the positional relationships of the entity elements, and the attributes of the entity elements;

[0042] The alignment method is constructed as follows: ,

[0043] In this embodiment, through a unified reference entity, the efficient integration and precise alignment of heterogeneous data are achieved, overcoming the deficiencies of traditional methods in data integration efficiency, entity alignment accuracy, and intelligent analysis capabilities. This method combines the DeepLabV3+ semantic segmentation model, natural language processing technology, and GIS spatial analysis tools to ensure the automation and efficiency of data processing; utilizes the powerful storage and query capabilities of the graph database to provide a new data management and analysis framework for smart city planning, showing significant application effects in urban road planning, greening coverage analysis, and environmental quality assessment, providing strong technical support for scientific decision-making in smart cities.

[0044] In this embodiment, the vector data Data V is derived from the data extracted from Open Street Map and other open source datasets in the study area according to the vector data extraction principle, and the vector data extraction principle includes: all data containing at least the roads and administrative regions in the study area.

[0045] The raster data Data R is derived from the remote sensing data extracted from the study area according to the raster data extraction principle, and the raster data extraction principle includes: the actual application scenario requirements and the existing data on the platform.

[0046] Specifically, the raster data Data R是 ​Extracted based on Google Earth Engine. The extracted data includes data such as the Normalized Difference Vegetation Index, land use data, radiation data, land surface temperature data, and rainfall data. The specific products to be selected can be determined according to the project requirements and the products available on GEE.

[0047] Text data T Derived from the data extracted from the store reviews within the research area according to the text data extraction rules. The text data extraction rules include: being able to be segmented into longitude and latitude by natural language processing technology at least.

[0048] Image data I Derived from the sampling point data along the streets within the research area. The sampling point data along the streets within the research area includes the green view rate, building view rate, and vehicle view rate extracted from the 360° panoramic images of the streets.

[0049] In this embodiment, the sampling point data along the streets is derived from the 360° panoramic images of the sampling points captured by the Baidu Street View API.

[0050] As Figure 5 shown, in this embodiment, when constructing the ontology layer of the knowledge graph in step two, the first step is to construct the set of entities, relationships, and attributes in the vector data V :

[0051] In this embodiment, constructing the vector entity set, vector attribute set, and vector relationship set includes:

[0052] a) Entity extraction: Extract the road name (RoadName), OSM identifier (ID), and administrative division (District) to form the vector entity set E v ;

[0053] In this embodiment, the vector entity set E v is expressed as follows:

[0054] ;

[0055] b) Attribute extraction: Calculate the topological attributes and geometric attributes of each entity element in the vector entity set E v to form the vector attribute set A v , where the topological attributes include at least adjacent roads and road length, and the geometric attributes include at least longitude and latitude;

[0056] In this embodiment, the vector attribute set A v is expressed as follows:

[0057] ;

[0058] c) Relationship construction: Establish spatial and attribute relationships such as the class, length, nearest road, Euclidean distance, and administrative division of geographical entities, and form a vector relationship set R v .

[0059] In this embodiment, the vector relationship set R v is expressed as follows:

[0060] ={ , , , , , , };

[0061] The second step is to construct the set of entities, relationships, and attributes in the raster data Data R :

[0062] a) Entity extraction: Generate sampling point IDs along the roads in the study area and form a raster entity set E in combination with the time field (T) r ;

[0063] In this embodiment, the raster entity set E r is expressed as follows:

[0064] ;

[0065] b) Attribute extraction: Extract the corresponding data types in the remote sensing data in the study area according to the raster data extraction principle, such as raster attribute values such as land use type, digital elevation model, land surface temperature, precipitation, and albedo, to form a raster attribute set A r ;

[0066] In this embodiment, the raster attribute set A r is expressed as follows:

[0067] ;

[0068] c) Relationship construction: Extract the corresponding data types in the remote sensing data in the study area according to the raster data extraction principle, such as land use type, in-road, digital elevation, land surface temperature, and administrative division where located, etc. relationships, and establish the raster entity set E rThe geographical location relationships of the various entity elements in form a grid relationship set R r .

[0069] In this embodiment, the grid relationship set R r has the following expression:

[0070] = { , , , , , , , , , };

[0071] The third step is the construction of the text data Data T of the text entity set, text relationship set, and text attribute set:

[0072] a) Entity extraction: Extract the store names within the research area from the text data Data T to form a text entity set E t ; The specific method is to extract the store names and related information from the text through natural language processing (NLP) technology to form a text layer entity set Et

[0073] In this embodiment, the text entity set E t has the following expression:

[0074] ;

[0075] b) Attribute extraction: Extract the text attributes of each text entity element in the text entity set E T from the text data Data t such as detailed address (Detaillocation), category (Class), and sentiment analysis result (Sentiment) to form a text attribute set A t ;

[0076] In this embodiment, the text attribute set A t has the following expression:

[0077] ;

[0078] c) Relationship construction: According to the text data Data TFor each store name, construct the geographical relationships between the store and the nearest road (NearRoad), the administrative division where it is located (LocatedIn), and the longitude and latitude (Longitude, Latitude) to form the text relationship set R t 。

[0079] In this embodiment, the text relationship set R t is expressed as follows:

[0080] ={ , , , , , , , , };

[0081] The fourth step is the construction of the image entity set, image attribute set, and image relationship set in the image data Data I :

[0082] a) Entity extraction: Based on the sampling point IDs along the roads in the study area, mark the geographical entities in the street view images to form the image entity set E i ;

[0083] In this embodiment, the text entity set E i is expressed as follows:

[0084] ;

[0085] b) Attribute extraction: Use the deep learning model (DeepLabV3+) to extract the green view index GVI, building view index BVI, and vehicle view index VVI in the study area to form the image attribute set A i ;

[0086] In this embodiment, the image attribute set A i is expressed as follows:

[0087] ;

[0088] c) Relationship construction: Establish the geographical location relationships between the sampling points along the street and the roads where they are located (InRoad), administrative divisions (LocatedIn), and longitude and latitude (Longitude, Latitude) to form the image relationship set R i 。

[0089] In this embodiment, the image relationship set R i is expressed as follows:

[0090] ={Bracket , , , , , , }.

[0091] The fifth step is entity alignment:

[0092] The entity collection of all data layers ( Through the alignment of the unified reference entity (RoadName), a comprehensive entity set E is generated to achieve spatial consistency between the data layers. This alignment lays the foundation for the comprehensive analysis of multi-source heterogeneous data and supports the subsequent relationship and attribute expansion of the knowledge graph.

[0093] Step 3: According to the collection and alignment constructed in the ontology layer of the knowledge graph, the DeepLabV3+ semantic segmentation model, NLP model and GIS platform are used to extract the entity collection, relationship collection, attribute collection and alignment in the vector layer, raster layer, text layer and image layer of the target city based on ArcGIS Pro, Google Earth Engine and PyCharm platforms to form a multi-source geospatial big data knowledge graph, and then realize the storage management of the multi-source geospatial big data knowledge graph based on Neo4j;

[0094] In this embodiment, Wuhan City is taken as an example to extract entities, relationships, and attributes from different data layers of the multi-source geospatial big data knowledge graph and align entities between different data layers. The results of triple extraction are divided into<E, R, A> Extract results and<E, R, E> Extraction results. The next step is to use Python to connect to the Neo4j interface, namely Py2Neo, to visualize the triples of each data layer based on the extracted triples. After the triples are extracted from each data layer, the alignment between multi-source data entities is performed based on RoadName in the Neo4j graph database.

[0095] This implementation is based on the actual data of Wuhan City under the guidance of the knowledge graph ontology layer constructed in step 2. , , Construction of triples, combined with the py2neo interface, stores the extraction results of <Entity, Relation, Attribute> and <Entity, Relation, Entity> in the Neo4j graph database, and aligns multi-source data entities based on RoadName in the Neo4j graph database. For part of the knowledge graph, see the appendix Figure 3 For the extracted <E, R, A> and <E, R, E>, see Tables 1 and 2

[0096] Table 1: Example Table of Knowledge Graph <E, R, A>

[0097]

[0098] Table 2: Example of Knowledge Graph <E, R, E>

[0099]

[0100] The specific operations are as follows:

[0101] 1.1 Vector Layer Entity, Attribute Extraction, and Relationship Construction

[0102] The extraction of entities, attributes, and relationships in the vector layer is based on the ArcGIS Pro platform (here it can also be other GIS platforms, such as QGIS, etc.).

[0103] (1) Entity Extraction

[0104] In the vector layer, it mainly involves three types of entities: RoadName (road name), ID (sampling point), and District (administrative region):

[0105] a. Road vector: It has the RoadName attribute and can be directly extracted.

[0106] b. Sampling point vector: Use the Generate Sampling Points Along Line tool in ArcGIS Pro to extract the entity ID according to the set fixed sampling point interval (500m in this example) in the study area.

[0107] c. Administrative region boundary vector: It contains the administrative region name itself and is extracted as the District entity.

[0108] (2) Attribute Extraction

[0109] a. Road vector: Use the Calculate Geometry tool in ArcGIS Pro to extract the Length (length) attribute; the road classification attribute can be directly obtained from the attribute table of the original data.

[0110] b. Sampling Point Vector: Use computational geometry tools to extract the Latitude and Longitude attributes.

[0111] (3)Relationship Construction

[0112] Relationship construction includes the attribute relationships of entities and the spatial relationships of entities, namely <entity, relationship, attribute> and <entity, relationship, entity>:

[0113] a. Entity-Attribute Relationship (<entity, relationship, attribute>): According to the attribute content, extract the corresponding relationship names, such as Length, Latitude, Longitude, Class, etc. For example:

[0114] <RoadName, Length, 12.5km>

[0115] <ID, Latitude, 32.589>

[0116] <ID, Longitude, 122.65>

[0117] <RoadName, Class, Provincial Road>

[0118] Entity-Entity Relationship (<entity, relationship, entity>): Based on spatial analysis (such as spatial join, proximity analysis), determine the spatial relationships between entities. For example:

[0119] <ID, Located IN, District> (Administrative district where the sampling point belongs)

[0120] <ID, Located IN, RoadName> (Road near the sampling point)

[0121] Finally, all entities, attributes, and relationships form an attribute table in ArcGIS Pro and are exported through Table to Excel. The table header is the entity (head node), Relationship 1, Relationship 2, Relationship 3..., and the table values are the attributes (tail nodes) and entities (tail nodes), forming a preliminary triple table.

[0122] 1.2 Extraction of Entities, Attributes, and Relationship Construction for the Raster Layer

[0123] The extraction of entities, attributes, and relationships for the raster layer is based on the ArcGIS Pro platform and the Google Earth Engine platform.

[0124] (1)Entity Extraction

[0125] In the raster layer, the following entities are mainly extracted:

[0126] a. Sampling point entity (ID): This step has been completed in the vector layer and can be directly reused.

[0127] b. Temporal entity: Set the time nodes as 2000, 2010, and 2020 (the time span can be adjusted according to project requirements).

[0128] (2) Attribute extraction

[0129] The attribute extraction process for raster data is the same, and the main steps are as follows:

[0130] a. Upload the vector boundary of the study area in GEE. According to project requirements, determine the required raster data, which in this example includes: Land Use; Land Surface Temperature (LST); Precipitation; Temperature; Albedo

[0131] b. Write JavaScript code in GEE to extract or invert the raster data within the study area and download it to the local according to the time scales of 2000, 2010, and 2020.

[0132] c. In ArcGIS Pro, use the Extract Values to Points tool to associate the raster data values with the sampling point ID entities.

[0133] (3) Relationship extraction

[0134] The relationship extraction in the raster layer includes the attribute relationships and spatio-temporal relationships of entities, namely <entity, relationship, attribute> and <entity, relationship, entity>:

[0135] a. Entity-attribute relationship (<entity, relationship, attribute>): Extract the corresponding relationship names according to the attribute content, such as LST, LandUse, Temperature, Precipitation, etc. For example:

[0136] <ID, Length, 12.5km>

[0137] <ID, Temperature, 22.5℃>

[0138] <ID, Precipitation, 856.22mm>

[0139] Entity-Entity Relationship (<Entity, Relationship, Entity>): Based on spatial analysis (such as spatial join, proximity analysis), determine the spatial relationship between entities. Construct the temporal relationship between attributes and entities based on the year of raster data. For example:

[0140] <ID, Located IN, District> (Administrative district where the sampling point belongs)

[0141] <ID, InRoad, RoadName> (Road near the sampling point)

[0142] <22.5℃, Year, 2000) (Average temperature in 2000)

[0143] Finally, all entities, attributes, and relationships form an attribute table in ArcGIS Pro and are exported to Excel through Table to Excel. The table header is the entity (head node), Relationship 1, Relationship 2, Relationship 3…, and the table values are the attributes (tail nodes) and entities (tail nodes), forming a preliminary triple table.

[0144] 1.3 Entity, Attribute Extraction and Relationship Construction in the Text Layer

[0145] The extraction of entities, attributes, and relationships in the text layer is based on the ArcGIS Pro and PyCharm platforms and uses natural language processing (NLP) technology for automatic recognition. In this example, the BiLSTM-CRF and BERT models are mainly used to complete the parsing and extraction of text information.

[0146] (1) Entity Extraction

[0147] The extraction of text layer entities mainly involves the store name (StoreName), and the process is as follows:

[0148] a. Data preprocessing: Annotate and vectorize the stored review or description text. Use the pre-trained BERT model to encode the text to generate context representations for each word, enhancing the model's ability to understand entities.

[0149] b. Model training: Train the BiLSTM-CRF model using the annotated training set to enable it to learn to recognize the store name (StoreName). The CRF layer further optimizes the overall consistency of the label sequence, improving the model's accurate recognition rate of the store name.

[0150] c. Entity Recognition: After training, use the BiLSTM-CRF model to predict the store names in the new text. For each input sentence, the model assigns a label to each word, and the words labeled as "StoreName" are recognized as store entities. (2) Attribute Extraction

[0151] The attribute extraction at the text layer mainly involves store categories, sentiment tendencies, location information, etc. Among them, the BERT model can be used for sentiment analysis, and BiLSTM-CRF is used for location information extraction.

[0152] The process of sentiment analysis (ReviewType) is as follows:

[0153] a. Text preprocessing, perform word segmentation (use the WordPiece word segmentation tool for Chinese).

[0154] b. Input the segmented text into the BERT model for encoding.

[0155] c. BERT performs sentiment classification based on the pre-training results and outputs categories such as negative reviews (Negative), positive reviews (Positive), and neutral reviews (Neutral).

[0156] Store attributes extract the specific store address (DetailLocation), store category (Class), and store longitude and latitude coordinates (Longitude, Latitude) based on BiLSTM-CRF.

[0157] (3) Relationship Construction

[0158] The relationship extraction at the text layer includes the attribute relationship and spatial relationship of entities, namely <entity, relationship, attribute> and <entity, relationship, entity>:

[0159] a. Entity-attribute relationship (<entity, relationship, attribute>): According to the attribute content, extract the corresponding relationship name, such as Class, ReviewType, Detail Location, etc. For example:

[0160] <StoreNameA, ReviewType, Negative Review>

[0161] <StoreNameA, Class, Chinese Cuisine>

[0162] Entity-entity relationship (<entity, relationship, entity>): Based on spatial analysis (such as spatial connection, proximity analysis), determine the spatial relationship between entities. For example:

[0163] < StoreNameA, Located IN, District> (Administrative Region of the Store)

[0164] < StoreNameA, InRoad, RoadName> (Road near the Sampling Point)

[0165] Finally, all entities, attributes, and relationships are formed into an attribute table in ArcGIS Pro and exported through Table to Excel. The table header is the entity (head node), Relationship 1, Relationship 2, Relationship 3..., and the table values are the attributes (tail node) and entities (tail node), forming a preliminary triple table.

[0166] 1.4 Entity, Attribute Extraction, and Relationship Construction of the Image Layer

[0167] The extraction of entities, attributes, and relationships in the image layer is based on the ArcGIS Pro and PyCharm platforms, and the DeepLabV3+ semantic segmentation model is used for automatic recognition.

[0168] (1) Entity Extraction

[0169] In the image layer, the following entities are mainly extracted:

[0170] a. Sampling Point Entity (ID): This step has been completed in the vector layer and can be directly reused.

[0171] (2) Attribute Extraction

[0172] The attribute extraction of the image layer is mainly determined according to the results of semantic segmentation and the requirements of the project. In this example, GVI (Green View Index), BVI (Building View Index), and VVI (Vehicle View Index) are used as the specific attributes of the image layer. The specific extraction process is as follows:

[0173] a. Use the sampling point data to crawl 360° panoramic images in recent years (the time is determined according to the project requirements and street view image data) based on the Baidu Street View platform;

[0174] b. Download the DeepLabV3+ model to the local, perform semantic segmentation on the crawled street view images of the research area, and output the semantically segmented images;

[0175] c. Based on PyCharm, calculate the percentage of different pixel values in the whole image to obtain the GVI, BVI, and VVI of each sampling point ID.

[0176] (3) Relationship Construction

[0177] The relationship extraction in the image layer includes the attribute relationship and spatial relationship of entities, namely <Entity, Relationship, Attribute> and <Entity, Relationship, Entity>:

[0178] a. Entity-attribute relationship (<entity, relationship, attribute>): Extract the corresponding relationship name according to the attribute content, such as GVI, BVI, VVI, etc. For example:

[0179] <ID, GVI, 22.1>

[0180] <ID, VVI, 12.5>

[0181] Entity-entity relationship (<entity, relationship, entity>): Based on spatial analysis (such as spatial join, proximity analysis), determine the spatial relationship between entities. For example:

[0182] <ID, Located IN, District> (the administrative region where the sampling point belongs)

[0183] <ID, InRoad, RoadName> (the road where the sampling point is located)

[0184] Finally, all entities, attributes, and relationships form an attribute table in ArcGIS Pro and are exported to Excel through Table to Excel. The table header is the entity (head node), Relationship 1, Relationship 2, Relationship 3..., and the table values are the attributes (tail nodes) and entities (tail nodes), forming a preliminary triple table.

[0185] 1.5 Entity alignment of data layers

[0186] In the four data layers (vector layer, raster layer, text layer, relationship layer), the entities of each data layer are associated with RoadName. Therefore, use RoadName as the core entity to align the data of the four data layers on the PyCharm platform, and finally construct a knowledge representation in triple format. The core of data alignment is to ensure that all entities in the data layers can be associated through RoadName to construct a unified structured table. Since the core entities (ID, StoreName, etc.) of the vector layer, raster layer, and text layer are all associated with RoadName, data alignment is performed using RoadName as the main index.

[0187] During the data integration process, RoadName is used as the core entity to align the text layer, image layer, vector layer, and raster layer in order to construct a consistent relationship dataset and finally convert it into triple format. The following are the specific data alignment steps:

[0188] (1): Establish RoadName as the main index

[0189] Since entities in all four data layers can be associated with RoadName, RoadName is used as the main index for alignment.

[0190] Unify the naming format of RoadName to ensure that road names from different data sources are consistent and avoid matching failures due to spelling differences. Therefore, all RoadNames are extracted from the vector layer as the reference index table, and information from other data layers is associated with this table.

[0191] (2): Alignment of data in the text layer (Text Layer)

[0192] Match the corresponding RoadName through address resolution. If the store has longitude and latitude information, use spatial nearest neighbor analysis to associate it with the nearest RoadName.

[0193] (3): Alignment of data in the image layer (Image Layer)

[0194] Align RoadName based on geographical coordinates and link to the nearest RoadName through the coordinates of the street view image.

[0195] (4): Alignment of data in the vector layer (Vector Layer)

[0196] The vector data itself has RoadName information and can be directly used as the main index for alignment. Determine information such as sampling points (ID) and administrative regions (District) related to RoadName through spatial join.

[0197] (5): Alignment of data in the raster layer (Raster Layer)

[0198] In the raster layer, raster data is usually bound to sampling points (ID), and sampling points (ID) have been associated with RoadName through vector layer alignment.

[0199] (6): Unify the storage format and output, and form a unified data table based on RoadName. That is, using RoadName as the main index, store the attributes of the four data layers in the same table as shown in the following table:

[0200] Attribute table

[0201]

[0202] Convert to triple format (<entity, relationship, attribute> or <entity, relationship, entity>) based on PyCharm as follows:

[0203] <StoreName, located_in, RoadName>

[0204] <ID, located_in, RoadName>

[0205] <ID, near, RoadName>

[0206] <RoadName, Temperature, 25.6℃>

[0207] Furthermore, a multi-source geospatial big data knowledge graph is finally formed.

[0208] Step 4: Based on the obtained multi-source geospatial big data knowledge graph, use the Neo4j graph database and the Py2neo interface for data query, calculation, and visualization applications.

[0209] Specifically: Based on Neo4j + Py2neo, query, calculation, and visualization of multi-source geospatial data can be realized. The specific steps are as follows:

[0210] A. Data query

[0211] Execute data retrieval in the Neo4j graph database using the Cypher query language to extract information such as roads, administrative regions, shops, traffic, environment, and land use.

[0212] The query methods include:

[0213] Query based on geographical location (e.g., high-temperature roads and store distribution within a certain area);

[0214] Query based on time (e.g., the change trend of land use in the past 20 years);

[0215] Query based on the association of multi-source data (e.g., the relationship between temperature change and greening coverage rate).

[0216] B. Data calculation

[0217] Aggregation calculation (e.g., calculating the average temperature within a certain area and the density of shops in a certain block);

[0218] Time series analysis (e.g., analyzing land use changes based on three time nodes of 2000 / 2010 / 2020);

[0219] Spatial analysis (e.g., calculating the greening index and building visibility index of a certain street).

[0220] C. Data visualization

[0221] Combined with Python data visualization libraries (such as Matplotlib, Seaborn) and GIS visualization tools (such as Leaflet, ArcGIS), the query results are presented intuitively. Generate heat maps, time series change graphs, distribution maps, etc.

[0222] It should be noted that: Based on the Neo4j graph database and the Py2neo interface, data query, calculation, and visualization applications are divided into 5 application scenarios, namely urban heat island effect analysis, urban green view rate and building density analysis, urban travel recommendation, road network and traffic flow analysis, and urban land use change detection.

[0223] Such as Figure 2 shown, considering data availability, this embodiment selects Wuhan City to conduct experimental verification, and the spatial range of the example data is as Figure 2 shown. All of DataV, DataR, DataT, and DataI involved in this embodiment should be unified within the research area of Wuhan City.

[0224] In this embodiment, 5 applications under 5 smart city scenarios are preset, namely urban heat island effect analysis, urban green view rate and building density analysis, urban travel recommendation, road network and traffic flow analysis, and urban land use change detection.

[0225] Among them:

[0226] The main application direction of urban heat island effect analysis is to analyze the heat island area of the city based on the LST data of the Raster Layer. The query direction is to query temperature nodes in different years (2000, 2010, 2020) based on Python and the multi-source geospatial big data knowledge graph, find high-temperature areas, and provide cooling strategies for urban planning.

[0227] Query example (retrieve the LST value of a certain road in 2020):

[0228] MATCH (road)-[r:LST2020]->(temp)

[0229] WHERE temp.value IS NOT NULL

[0230] RETURN road.name AS RoadName, temp.value AS Temperature

[0231] Visualization:

[0232] Road temperature distribution map (10 roads showing the highest and lowest temperatures); Heat island change trend map (analyzing the heat island expansion in different years), obtained through graph database queries using py2neo nodes Figures 6 - 8 。

[0233] The main application direction of the analysis of urban green view rate and building density is to analyze the relationship between the green view rate and building view rate in different areas of the city based on the GVI and BVI data of the Image Layer, facilitating the later optimization of the urban layout. The query direction is to query the GVI value and BVI value of the sampling points based on Python and the multi-source geospatial big data knowledge graph to obtain the visualization results.

[0234] Query example (obtaining GVI / BVI and geographical locations of each road):

[0235] MATCH (road)-[:GVI]->(gvi_node)

[0236] MATCH (road)-[:BVI]->(bvi_node)

[0237] MATCH (road)-[:Longitude]->(lon_node)

[0238] MATCH (road)-[:Latitude]->(lat_node)

[0239] RETURN road.name AS RoadName, gvi_node.value AS GVI, bvi_node.valueAS BVI, lon_node.value AS Longitude, lat_node.value AS Latitude

[0240] Visualization:

[0241] GVI / BVI spatial distribution map; Statistical chart of average GVI / BVI in different administrative regions. For reference examples, see Figure 9 。

[0242] The main application direction of the urban travel recommendation scenario is to analyze the distribution of positive reviews of stores in the city based on the store sentiment analysis data of the Text Layer, laying the foundation for downstream travel recommendations. The query direction is to query the RiewType relationship based on Python and the multi-source geospatial big data knowledge graph to draw the basic distribution map, and then conduct the analysis of the positive review rate at the road scale and the positive review rate of store categories on the returned data.

[0243] Query example (query the evaluation categories of nearby shops):

[0244] MATCH (store)-[:NearRoad]->(road)

[0245] MATCH (store)-[:Class]->(class)

[0246] MATCH (store)-[:ReviewType]->(review)

[0247] RETURN road.name AS RoadName, class.name AS Category, review.value ASReviewType

[0248] Visualization:

[0249] The heat map of urban shop evaluations, showing the proportion of positive evaluations counted by street and shop category, as shown Figure 10 below.

[0250] The main application direction of road network and traffic flow analysis is to analyze the traffic flow of urban streets based on the VVI data of the Image Layer, facilitating the optimization of traffic management strategies. The query direction is to draw a basic distribution map of VVI relationships based on Python and a multi-source geospatial big data knowledge graph query, and then perform traffic flow analysis at the road scale on the returned data.

[0251] Query example (obtain the VVI values and geographical information of each road):

[0252] MATCH (road)-[:VVI]->(vvi_node)

[0253] MATCH (road)-[:Longitude]->(lon_node)

[0254] MATCH (road)-[:Latitude]->(lat_node)

[0255] RETURN road.name AS RoadName, vvi_node.value AS VVI, lon_node.valueAS Longitude, lat_node.value AS Latitude

[0256] Visualization:

[0257] Urban traffic flow heat map (showing high - flow and low - flow areas); traffic flow trend chart (displaying the flow trends in different regions), the results are as Figure 11 shown.

[0258] The main application direction of urban land use change detection is to analyze the inflows and outflows of land use among different regions based on the time - series land use data of the Raster Layer, providing data support for land resource management. The query direction is to query the number of sampling points of various types of land in each District based on the Landuse relationship in Python and the multi - source geospatial big data knowledge graph, and then generate the land use inflow - outflow map between different years based on Python.

[0259] Query example: MATCH (road)-[:Landuse2000]->(landuse2000)

[0260] MATCH (road)-[:Landuse2010]->(landuse2010)

[0261] MATCH (road)-[:Landuse2020]->(landuse2020)

[0262] RETURN road.name AS RoadName, landuse2000.value AS LandUse2000,landuse2010.value AS LandUse2010, landuse2020.value AS LandUse2020

[0263] Visualization:

[0264] Land use change heat map (showing land use changes from 2000 to 2020); land use conversion statistics for different administrative regions, the final results are as Figure 12 shown.

[0265] It should be noted that the application directions of the multi - source geospatial big data knowledge graph in smart cities are not limited to the above five types, and the specific implementation can be determined according to the task requirements and the situation of the research area.

[0266] Based on the same technical concept of the above - mentioned embodiments, the present invention provides a multi - source geospatial big data knowledge graph construction system for smart city planning, including:

[0267] A data collection and processing module for collecting the original data of the vector layer, raster layer, text layer, and image layer, and performing pre - processing;

[0268] The ontology layer construction module is used to construct a vector entity set E based on the preprocessed vector layer v , a vector relationship set A v , and a vector attribute set R v . It constructs a raster entity set E based on the preprocessed raster layer r , a raster relationship set A r , and a raster attribute set R r . It constructs a text entity set E based on the preprocessed text layer t , a text relationship set A t , and a text attribute set R t . It constructs an image entity set E based on the preprocessed image layer i , an image relationship set A i , and an image attribute set R i . Using RoadName as the reference entity, it aligns the vector entity set E v , the raster entity set E r , the text entity set E t , and the image entity set E i through a unified reference entity to obtain the entity set E = {E v , E r , E t , E i}. According to the obtained entity set E, it obtains the relationship set A and the attribute set R to form the ontology layer of the knowledge graph;

[0269] The knowledge graph construction module is used to extract the entity set, relationship set, attribute set, and alignment method of each data layer in the target city based on the set and alignment method of the knowledge graph ontology layer, using the DeepLabV3+ semantic segmentation model, the NLP model, and the GIS platform based on the ArcGIS Pro, Google Earth Engine, and PyCharm platforms to form a multi-source geospatial big data knowledge graph.

[0270] It should be noted that this system also includes an application module, which is used to perform data query, calculation, and visualization applications based on the obtained multi-source geospatial big data knowledge graph using a graph database and an interface.

[0271] In actual application, taking Wuhan as the target city, the vector layer, raster layer, text layer, and image layer within Wuhan are imported into the data collection and processing module. After being processed by this module, the vector data, raster data, text data, and image data within Wuhan are obtained.

[0272] Subsequently, using RoadName as the reference entity, the vector data, raster data, text data, and image data within Wuhan City are integrated and aligned based on the DeepLabV3+ semantic segmentation model, NLP model, and GIS platform using ArcGIS Pro, Google Earth Engine, and PyCharm, thereby obtaining a multi-source geospatial big data knowledge graph with Wuhan City as the blueprint.

[0273] Finally, according to requirements, the application module is used to retrieve the required data from the multi-source geospatial big data knowledge graph through the Neo4j graph database and the Py2neo interface for data query, calculation, and visual presentation.

[0274] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a multi-source geospatial big data knowledge graph for smart city planning, characterized in that: include: Collect the original data of the vector layer, raster layer, text layer and image layer, and perform preprocessing to obtain vector data, raster data, text data and image data respectively; Constructing a knowledge graph ontology layer, wherein the constructing of the knowledge graph ontology layer includes: constructing a vector entity set, a vector attribute set, and a vector relationship set according to vector data, constructing a raster entity set, a raster attribute set, and a raster relationship set according to raster data, constructing a text entity set, a text attribute set, and a text relationship set according to text data, constructing an image entity set, an image attribute set, and an image relationship set according to image data, aligning the vector entity set, the raster entity set, the text entity set, and the image entity set with street names as reference entities to obtain an entity set, and determining an attribute set and a relationship set according to entity elements in the obtained entity set, positional relationships between the entity elements, and attributes of the entity elements; According to the collection and alignment constructed in the ontology layer of the knowledge graph, the DeepLabV3+ semantic segmentation model, NLP model and GIS platform are used to extract the entity collection, relationship collection, attribute collection and alignment in the vector layer, raster layer, text layer and image layer of the target city based on the software platform to form a multi-source geographic spatial big data knowledge graph.

2. The method for constructing a multi-source geospatial big data knowledge graph for smart city planning according to claim 1 is characterized by: The vector data at least includes road data and administrative district data in the study area.

3. The method for constructing a multi-source geospatial big data knowledge graph for smart city planning according to claim 2 is characterized by: Constructing the vector entity set, the vector attribute set and the vector relationship set includes: Extract road names, OSM logos and administrative divisions from vector data to form a vector entity set; Calculating the topological attributes and geometric attributes of each entity element in the vector entity set to form a vector attribute set, wherein the topological attributes at least include adjacent roads and road lengths, and the geometric attributes at least include longitude and latitude; The spatial and attribute relationships of the category, length, nearest road, Euclidean distance and administrative division of each entity element in the vector entity set are established to form a vector relationship set.

4. The method for constructing a multi-source geospatial big data knowledge graph for smart city planning according to claim 1 is characterized by: The raster data includes remote sensing data within the study area required by the actual application scenario and remote sensing data within the study area already available on the platform.

5. The method for constructing a multi-source geospatial big data knowledge graph for smart city planning according to claim 4 is characterized by: Constructing the grid entity set, grid attribute set and grid relationship set includes: Generate sampling point IDs along the roads in the study area and combine them with the time field to form a raster entity set; Extract the grid attribute value of each grid entity element in the grid entity set in the remote sensing data in the study area to form a grid attribute set; The geographical location of each raster entity element in the raster entity set in the remote sensing data in the study area is extracted, and the geographical location relationship of each raster entity element in the raster entity set is established to form a raster relationship set.

6. The method for constructing a multi-source geospatial big data knowledge graph for smart city planning according to claim 1 is characterized by: The text data includes at least store review data that can be segmented into longitude and latitude using natural language processing technology.

7. The method for constructing a multi-source geospatial big data knowledge graph for smart city planning according to claim 6 is characterized by: Constructing the text entity set, text relationship set and text attribute set includes: Extract the store names in the study area from the text data to form a text entity set; Extract the text attributes of each text entity element in the text entity set from the text data to form a text attribute set A t ; According to the names of each store in the text data, the geographical relationship between the store and the nearest road, the administrative division where it is located, and the longitude and latitude is constructed to form a text relationship set.

8. The method for constructing a multi-source geospatial big data knowledge graph for smart city planning according to claim 1 is characterized by: The image data includes picture data from which green visibility, building visibility and vehicle visibility can be extracted in the study area.

9. The method for constructing a multi-source geospatial big data knowledge graph for smart city planning according to claim 8 is characterized by: Constructing the image entity set, image attribute set and image relationship set includes: Based on the sampling point IDs along the roads in the study area, the geographic entities in the image data are marked to form an image entity set; The deep learning model DeepLabV3+ is used to extract the green view rate, building view rate and vehicle view rate in the image data to form an image attribute set; The geographical location relationship between the sampling points along the street and the roads, administrative divisions, and longitude and latitude is established to form an image relationship set.

10. A multi-source geospatial big data knowledge graph construction system for smart city planning, characterized in that: include: The data collection and processing module is used to collect the original data of the vector layer, raster layer, text layer and image layer, and perform preprocessing; An ontology layer construction module is used to construct a vector entity set, a vector relationship set and a vector attribute set according to the preprocessed vector layer, to construct a raster entity set, a raster relationship set and a raster attribute set according to the preprocessed raster layer, to construct a text entity set, a text relationship set and a text attribute set according to the preprocessed text layer, and to construct an image entity set, an image relationship set and an image attribute set according to the preprocessed image layer. The vector entity set, the raster entity set, the text entity set and the image entity set are aligned through a unified reference entity using RoadName as a reference entity to obtain an entity set, and a relationship set and an attribute set are obtained based on the obtained entity set to form a knowledge graph ontology layer. The knowledge graph construction module is used to extract the entity set, relationship set, attribute set and alignment of each data layer in the target city based on the collection and alignment of the knowledge graph ontology layer, using the DeepLabV3+ semantic segmentation model, NLP model and GIS platform based on ArcGIS Pro, Google Earth Engine and PyCharm platforms to form a multi-source geospatial big data knowledge graph.

Images