A multi-source spatio-temporal data transparent fusion method based on urban information unit

Abstract

A kind of multi-source spatio-temporal data transparent fusion method based on urban information unit, according to the hierarchical management of city, the city is overall profiled, constructs spatial basic information entity, defines multi-level urban information unit and the multi-source heterogeneous spatio-temporal data contained therein;Extract the spatial, temporal and correlation attributes in geographic entity features, establish comprehensive geographic unit coding identification based on location and time attributes;Based on the unified geographic unit coding identification, integrate data from multiple departments and multiple types on a platform and perform multi-scale statistical display according to the profiled grid, realize fast request, generation and service based on profiled tower type hierarchical data.The present application solves the similarity, inconsistency problem of cross-domain data in geometric position, attribute semantics, logic, etc.Realize the transformation from static three-dimensional visualization to intelligent dynamic visualization, build multi-source heterogeneous spatio-temporal data resource pool, and realize the transparent fusion of multi-source, heterogeneous and closed system urban government affairs big data.

Description

Technical Field

[0001] This invention relates to the field of urban physical-digital spatial social sensing big data, and in particular to a method for transparent fusion of multi-source spatiotemporal data based on urban information units. This method can provide multi-level and rich information content to different users by using a multi-source spatiotemporal data transparent fusion model for spatiotemporal big data with multiple sources, scales, themes, and phases in dynamic environments, and provide core technical support for intelligent spatial information services. Background Technology

[0002] With the rise of technologies such as "Internet+", big data, 5G, cloud computing, and artificial intelligence, the application fields of spatiotemporal data are constantly expanding and the application levels are constantly deepening. Users have put forward higher requirements for the production, processing, and analysis of spatial data.

[0003] With the rapid development of sensor and internet technologies, the ability to acquire spatiotemporal data has greatly improved, and different industries have produced a large amount of spatial data according to their different needs. Urban physical-digital spatial social sensing big data is characterized by diverse sources, diverse data formats, decentralized storage, and large data volume, which brings great difficulties to the collaborative expression, information aggregation, information derivation and value-added, and data mining of data. The key to the fusion of multi-source spatiotemporal big data lies in the merging and processing of features such as the geometric location, semantic attributes, and topological relationships of geographic entities. That is, it is necessary to fuse entities with the same name in multi-source, multi-scale vector data to centralize different datasets and form a higher-quality spatiotemporal database that meets application requirements.

[0004] The fusion of multi-source heterogeneous data in urban physical-digital space is of great significance and has broad application prospects: 1) Spatiotemporal data fusion helps to broaden the scope and depth of vector data research and improve the relevant theoretical, technical and methodological system; 2) The integration and fusion of vector spatial data from different sources in the same region can make these data mutually corroborate, complement and correlate each other in terms of attributes and geometric location, which can accurately determine the relevant information of geographic entities and improve the accuracy of data; 3) Spatial matching of updated data with the original data can realize the updating of spatiotemporal data at multiple scales.

[0005] Data fusion refers to the coordinated optimization and comprehensive processing of data collected from multiple sensors using computer processing techniques and certain criteria to obtain higher-quality data that meets practical needs. The fused data combines the advantages of multiple data sources, leading to more accurate status and estimates. Spatial data fusion primarily involves the integration and consolidation of multi-source spatial data. It is the process of fusing different data sources to generate new, higher-quality data. Depending on the type of spatial data, it can be categorized into vector spatial data fusion, raster spatial data fusion, and vector and raster spatial data fusion, among others.

[0006] Currently, the fusion of multi-source heterogeneous data in urban physical-digital space faces the following problems: 1) Social sensing and government information data are complex in form, making it difficult to interconnect, communicate, and share dynamic information during data fusion. This hinders the comprehensive management of dynamic information and easily leads to fault tolerance and robustness issues in dynamic fusion; 2) Multi-source spatiotemporal data and urban entities are difficult to match accurately. Different types of data are relatively isolated and have poor correlation, making it difficult to analyze and construct spatiotemporal big data relationships, thus affecting the efficiency of dynamic fusion of urban entities and multi-source spatiotemporal data; 3) Most existing methods for dynamic fusion of data and spatial entities are based on geographic grids, dividing regular grids at different granularities as needed. This ignores the irregular area division in urban management and the implicit correlation between information units at multiple levels. However, the daily management area of ​​a city is divided into irregular areas, and the information at multiple levels in a city is also closely related.

[0007] Therefore, how to overcome the shortcomings of existing technologies, study the dynamic fusion of multi-source heterogeneous data in physical and digital spaces, shield the complexity of data fusion, optimize data integration and management for the multiple semantic expressions of multi-source data, has become a technical problem that needs to be solved by existing technologies. Summary of the Invention

[0008] The purpose of this invention is to propose a multi-source spatiotemporal data fusion method based on urban information units, which can shield the complexity of the data fusion process, perform knowledge merging through unique data encoding, realize dynamic data linking, and thus achieve transparent fusion of multi-source data.

[0009] To achieve this objective, the present invention adopts the following technical solution:

[0010] A method for transparent fusion of multi-source spatiotemporal data based on urban information units includes:

[0011] Description steps S110 for the division of urban information units and integration of multi-source heterogeneous data:

[0012] Based on the city management level, the city is divided as a whole, spatial basic information entities are constructed, and multi-level city information units and multi-source heterogeneous spatiotemporal data contained in the city information units are defined.

[0013] Feature extraction step S120 for multi-source heterogeneous data:

[0014] Extract spatial, temporal, and relational attributes from the features of geographic entities, and establish comprehensive geographic unit coding identifiers based on location and temporal attributes;

[0015] Step S130: Multi-source heterogeneous data matching based on city units:

[0016] Based on a unified geographic unit coding identifier, data from multiple departments and types are integrated into a single platform and statistically displayed at multiple scales according to a grid, enabling rapid requesting, generation, and service of data based on a grid-like hierarchical structure.

[0017] Optionally, step S110 includes:

[0018] Urban information unit definition and division sub-step S111:

[0019] The city is divided into geographically independent urban units according to the city's management level. Basic government data and social sensing data accumulated in urban management are integrated into the urban units to obtain urban information units. Each unit contains the joint features of all units in the next lower level. The overall architecture of urban information is constructed based on the multi-level urban units.

[0020] City Information Unit Data Definition Sub-step S112:

[0021] The urban information unit contains multi-source heterogeneous spatiotemporal data, which is based on basic government data and social sensing data. Specifically, it includes: economic data, environmental data, construction data, and social data. The economic data includes: social security and economic development data; the environmental data includes: ecological protection, air quality, and water quality data; the construction data includes: urban and rural construction and transportation data; and the social data includes: public opinion, POI data, mobile phone signaling, and audio and video data. The basic government data and social sensing data are mainly divided into five types: text, images, videos, web pages, and tables.

[0022] Data integration step S113:

[0023] The process involves filtering and screening multi-source heterogeneous spatiotemporal data, removing unreasonable data, eliminating homonyms and synonyms, verifying consistency, deleting redundant data, and merging data.

[0024] Optionally, in step S120,

[0025] Attribute extraction is accomplished in the following way:

[0026] For text data, contextual features are automatically learned through deep learning models, and temporal information is extracted using random fields and maximum entropy models; attribute features and attribute values ​​related to geographic entities are obtained using rule matching and supervised learning methods.

[0027] For image data: Convolutional neural networks are used to identify objects and scenes and automatically generate content descriptions through image recognition technology;

[0028] The temporal feature extraction of the data is achieved by using the least squares criterion matching method and the interpolation and extrapolation time matching algorithm to synchronize asynchronous information about the same target from different sources to the same moment, thereby realizing the temporal registration of spatial data.

[0029] Spatial feature extraction of data involves semantic address matching of spatial information within the data. For a given corpus dataset D = {add1, add2, ..., add...} n The goal of semantic address matching is to find address pairs (add...). i ,add j ), satisfying add i =add j , where add i ∈D, add j ∈D and i≠j, to ensure the accuracy of the spatial location of the data.

[0030] Optionally, in step S120,

[0031] The integrated geographic unit coding identifier consists of a location code, a semantic code, a time code, and an association code.

[0032] Optionally, the coding rules for the integrated geographic unit coding identifier are as follows:

[0033] 1) Location code: The administrative region code has 9 digits, consisting of province, city, district and street. The coding conforms to the provisions of GB / T 2260 and GD / T10114. An additional 6 digits are added to represent the area and the smallest grid.

[0034] 2) Semantic code: Represents data attribute information;

[0035] 3) Time code: Represents the time elements "year", "month" and "day" in the time when the information unit was generated, in accordance with GB / T7408; add 4 digits to this to indicate whether it is a holiday and the time interval respectively;

[0036] 4) Association Code: The valid range is between 01 and 99, used to indicate the association strength between this city information unit and related information units.

[0037] Optionally, step S130 includes:

[0038] Data and geographic unit linking sub-step S131:

[0039] By integrating, extracting features from, and encoding multi-source spatiotemporal data, dynamic linking of data to geographic units is achieved based on location codes;

[0040] Multi-level connector step S132:

[0041] By matching city entities identified by geographic unit codes with data, multi-level linkage can be achieved.

[0042] Optionally, in step S110,

[0043] The city units are set up according to the city management level, including different levels such as province, city, administrative region, street, region and grid. Different management levels form an inclusion relationship. Based on the management level, the city is divided into multiple geographically independent city units.

[0044] Optionally, city units can be categorized from largest to smallest as province, city, administrative district, street, region, and geographic grid.

[0045] Urban units can address different urban management issues and apply different scales.

[0046] In summary, the present invention has the following advantages:

[0047] 1. By utilizing technologies such as knowledge graphs, visual knowledge, and deep learning, we can automatically perform 3D detection, segmentation, vector tracking, and attribute insertion for urban entities. This will organize multi-source heterogeneous and multimodal spatial big data in the physical world into a complex and massive data semantic network, solving the problems of similarity and inconsistency in geometric location, attribute semantics, and logic of cross-domain data.

[0048] 2. By combining space-air-ground integrated multi-source 3D data fusion and visualization technology, the transformation from static 3D visualization to intelligent dynamic visualization can be achieved.

[0049] 3. Construct a multi-source heterogeneous spatiotemporal data resource pool to achieve transparent integration of urban government big data from multi-source, heterogeneous, and closed systems. Attached Figure Description

[0050] Figure 1 This is a flowchart of a multi-source spatiotemporal data fusion method based on urban information units according to a specific embodiment of the present invention;

[0051] Figure 2 This is a data segmentation of a tower-type urban information unit according to a specific embodiment of the present invention;

[0052] Figure 3 This is the city information unit coding rule according to a specific embodiment of the present invention;

[0053] Figure 4 This is a high-precision urban environmental assessment system framework according to a specific embodiment of the present invention;

[0054] Figure 5 This is a schematic diagram of data multi-semantic representation and transparent fusion based on geocoding according to a specific embodiment of the present invention. Detailed Implementation

[0055] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0056] This invention utilizes technologies such as knowledge graphs, visual knowledge, and deep learning to automatically perform 3D detection, segmentation, vector tracking, and attribute attachment for urban entities, storing these data in a database. It organizes multi-source, heterogeneous, and multimodal spatial big data from the physical world into a complex and massive data semantic network, resolving issues of similarity and inconsistency in geometric location, attribute semantics, and logic across different domains. Furthermore, it combines space-air-ground integrated multi-source 3D data fusion and visualization technology to transform static 3D visualization into intelligent dynamic visualization. Finally, it constructs a multi-source, heterogeneous spatiotemporal data resource pool, enabling transparent integration of urban government big data from multi-source, heterogeneous, and closed systems.

[0057] Specifically, this invention first actively collects multi-source, multi-dimensional, and heterogeneous spatiotemporal big data, performs semantic parsing, and completes the spatiotemporal construction of geographic knowledge; establishes data matching and association models, and builds a transparent data fusion framework; combines multi-source heterogeneous data element matching technology to construct spatiotemporal data transparent fusion rules; and finally, with the support of numerous fusion methods, achieves transparent fusion of urban entities and spatiotemporal multi-source spatiotemporal data.

[0058] Specifically, see Figure 1 The method for transparent fusion of multi-source spatiotemporal data based on urban information units includes the following steps:

[0059] Description steps S110 for the division of urban information units and integration of multi-source heterogeneous data:

[0060] In practical applications of geospatial information, spatial data is diverse and complex, and different industries have different professional information. The production and maintenance of geospatial data are scattered among different units and use different data standards, resulting in inconsistent standards between departments or systems and between different historical stages.

[0061] Therefore, this step is as follows: based on the city management level, the city is divided into a whole, spatial basic information entities are constructed, and multi-level city information units and multi-source heterogeneous spatiotemporal data contained in the city information units are defined.

[0062] Specifically,

[0063] Urban information unit definition and division sub-step S111:

[0064] The city is divided into geographically independent urban units based on management levels. Basic government data and social sensing data accumulated in urban management are integrated into urban units to obtain urban information units. Each unit contains the joint features of all units in the next lower level. The overall architecture of urban information is constructed based on multi-level urban units.

[0065] See Figure 3 This is a schematic diagram of a tower-style urban information unit data segmentation provided by an embodiment of the present invention. It sets different levels according to urban management hierarchy, including provinces, cities, administrative districts, streets, regions, and grids. Different management levels form an inclusion relationship, dividing the city into multiple geographically independent urban units according to management hierarchy. The urban units, from largest to smallest, are provinces, cities, administrative districts, streets, regions, and geographical grids, forming a multi-level unit with hierarchical relationships. For example, Beijing has 16 administrative districts; Haidian District consists of 22 streets; each street is composed of areas such as schools, shopping malls, sports fields, and residential communities. Each area is further divided into grids, which is the smallest urban unit.

[0066] Urban information units are multi-granular, therefore, in this sub-step, different scales of urban units can be applied to address different urban management problems. For example, when studying the city's water, electricity, and gas consumption, the city or district level can meet the needs, while when studying the crowd gathering situation at a concert, the region or grid level should be selected.

[0067] City Information Unit Data Definition Sub-step S112:

[0068] The urban information unit contains multi-source heterogeneous spatiotemporal data, which is based on basic government data and social sensing data. Specifically, it includes: economic data, environmental data, construction data, and social data. The economic data includes: social security and economic development data; the environmental data includes: ecological and environmental protection, air quality, and water quality data; the construction data includes: urban and rural construction and transportation data; and the social data includes: public opinion, POI data, mobile phone signaling, and audio and video data. The basic government data and social sensing data are mainly divided into five types: text, images, videos, web pages, and tables.

[0069] Data integration step S113:

[0070] The process involves adjusting steps such as filtering and screening of multi-source heterogeneous spatiotemporal data, eliminating unreasonable data, removing homonyms and synonyms, verifying consistency, deleting redundant data, and merging data.

[0071] Feature extraction step S120 for multi-source heterogeneous data:

[0072] Extract spatial, temporal, and relational attributes from the features of geographic entities, and establish a comprehensive geographic unit coding identifier based on location and temporal attributes.

[0073] Specifically, attribute extraction is accomplished in the following way:

[0074] For text data, contextual features are automatically learned through deep learning models, and temporal information is extracted using random fields and maximum entropy models. Attribute features and values ​​related to geographic entities are obtained using rule matching and supervised learning methods.

[0075] Because deep learning models have lower reliance on corpora, they are better able to identify place names, addresses, and other information in the data, and then use place name dictionaries and contextual features to disambiguate geographic entities. Simultaneously, deep learning models can also distinguish relationships between two geographic entities from the text, such as temporal relationships like "same," "before," and "after," and spatial relationships like "intersecting" and "mutually exclusive." Temporal information reflects changes in geographic entities; this invention extracts temporal information using models such as random fields and maximum entropy. Methods for extracting attribute information include rule matching and supervised learning, obtaining attribute features and values ​​related to geographic entities.

[0076] For image data: Convolutional neural networks are used to identify objects and scenes and automatically generate content descriptions through image recognition technology.

[0077] Specifically, the mathematical expression for convolution in the convolutional neural network in this invention is defined by equation (1):

[0078] z(t)=f(t)*g(t)=∑f(τ)g(t-τ) (1)

[0079] Here, F and G are called mathematical operators, calculated from the area of ​​the overlapping region of the function g, and obtained by flipping and then translating the characterizing function f. Its integral form is:

[0080] z(t)=f(t)*g(t)=∫f(τ)g(t-τ)dτ=∫f(t-τ)g(τ)dτ (2)

[0081] Representing the image using functions in a two-dimensional coordinate system, denoted as f(x,y) and g(x,y), yields z(x,y):

[0082] z(x,y)=f(x,y)*g(x,y) (3)

[0083] Image feature extraction is achieved through convolution, and the expression for calculating two-dimensional coordinates is:

[0084] z(x,y)=f(x,y)*g(x,y)=∑∑f(t,h)g(xt,yh) (4)

[0085] Its integral form is:

[0086] z(x,y)=(fg)(x,y)=∫∫f(t,h)g(xt,yh)dtdh (5)

[0087] The definition of an m*n convolution kernel is:

[0088] Z(x,y)=f(x,y)*g(x,y)=∑∑f(t,h)g(xt,yh) (6)

[0089] Where f is the input image, g is the convolution kernel, m and n are the kernel sizes, and t and h are the number of rows and columns of matrix f.

[0090] Let M*N be the size of an image and n*n be the size of the convolution kernel. Multiplying n*n by each M*M is equivalent to taking out all n*n and forming a vector. After two operations, we get (M-n+1)*(M-n+1). After all n*n are successfully represented by the small image, we get the final representation.

[0091] The temporal features of the data are extracted by using the least squares criterion matching method, interpolation and extrapolation and other time matching algorithms to synchronize asynchronous information about the same target from different sources to the same time to achieve temporal registration of spatial data. For example, the MODIS remote sensing image data of Beijing from January 1, 2022 and February 1, 2022 can both be retrieved by the MODIS remote sensing image of Beijing in 2022.

[0092] Spatial feature extraction of data involves semantic address matching of spatial information within the data. For a given corpus dataset D = {add1, add2, ..., add...} n The goal of semantic address matching is to find address pairs (add...). i ,add j ), satisfying add i =add j , among which ad□ i ∈D, add j ∈D and i≠j, to ensure the accuracy of the spatial location of the data.

[0093] For integrated geographic unit coding and identification, in the application of multi-source spatiotemporal big data, a reasonable data storage method can improve data retrieval efficiency, save data application costs, and improve data service quality.

[0094] See Figure 5 This paper illustrates an encoding rule. The method employs a data cube based on urban information units and assigns a comprehensive geographic unit encoding identifier. Building upon traditional two-dimensional geospatial encoding, it adds encoding for temporal attributes. The comprehensive geographic unit encoding identifier consists of a location code, a semantic code, a time code, and a correlation code. Based on urban information units, a spatiotemporal data transparent fusion service model is designed and researched. This data service follows a transparent computing paradigm, supporting the provision of data services through transparent integration and fusion of data from different sources. The data is encoded based on the data features extracted in the previous step.

[0095] Specifically, the coding rules for integrated geographic unit identifiers are as follows:

[0096] 1) Location code: The administrative region code has 9 digits, consisting of province, city, district and street. The coding conforms to the provisions of GB / T 2260 and GD / T10114. An additional 6 digits are added to represent the area and the smallest grid.

[0097] 2) Semantic code: Represents data attribute information;

[0098] 3) Time code: Represents the time elements "year", "month" and "day" in the time when the information unit was generated, in accordance with GB / T7408; add 4 digits to this to indicate whether it is a holiday and the time interval respectively;

[0099] 4) Association Code: The valid range is between 01 and 99. It is used to indicate the association strength between this city information unit and related information units. The closer the number is to 01, the more special the record is, and it is used to monitor spatiotemporal behavioral anomalies. For example, if a building experiences a power outage at a certain time (time code) (location code), while surrounding information units do not experience such an event, the association code of this record should be closer to 01 to indicate the special nature of the data.

[0100] City units can also be coded using location codes. The larger the granularity, the shorter the location code, and vice versa. For example, Beijing is coded as 110, and the Sanlitun Subdistrict of Chaoyang District, Beijing is coded as 110102004.

[0101] Step S130: Multi-source heterogeneous data matching based on city units:

[0102] Based on a unique geographic unit code identifier, it supports the rapid entry and centralized management of various environmental information, realizing the fusion of address semantics and spatial information. Geocoding can convert spatial queries into integer code numerical matching queries, greatly improving the efficiency of vector and area of ​​interest queries.

[0103] Therefore, this step is as follows: Based on a unified geographic unit coding identifier, data from multiple departments and types are integrated into a single platform and statistically displayed at multiple scales according to a grid, enabling rapid requesting, generation, and service of data based on a grid-like hierarchical structure.

[0104] Specifically, step S130 includes:

[0105] Data and geographic unit linking sub-step S131:

[0106] By integrating, extracting features from, and encoding multi-source spatiotemporal data, dynamic linking of data to geographic units is achieved based on location codes.

[0107] See Figure 2 It demonstrates a city unit tower-style partitioning and data matching rules.

[0108] Multi-level connector step S132:

[0109] By matching city entities identified by geographic unit codes with data, multi-level linkage can be achieved.

[0110] For example, when the city unit code matches the first 2, 4, 6, 9, 12, and 15 digits of the data code, it can be linked. If a piece of data comes from Yangfangdian Subdistrict, Haidian District, Beijing, then the information units that can be linked to this data are Beijing, Haidian District, Beijing, and Yangfangdian Subdistrict, Haidian District, Beijing.

[0111] By using a unified urban information unit geocoding system, it is possible to establish multi-source data association and matching based on basic urban spatiotemporal geographic data, thereby achieving efficient indexing of multi-dimensional spatial information.

[0112] In this invention, the data transparency fusion framework is used for topic-oriented fusion of multi-source heterogeneous data in urban management. This framework can dynamically connect massive amounts of multi-source heterogeneous data to urban information units, achieving topic-oriented decision-level fusion. Users can obtain the required fused results without needing to understand the original data source or the full picture of the data, making urban management more convenient and efficient. See also... Figure 5 This demonstrates the multi-semantic representation and transparent fusion of geocoding-based data.

[0113] Example:

[0114] See Figure 4This paper presents a high-precision urban environmental quality assessment framework. This type of framework is part of the data fusion rule base in the urban big data resource pool. The data fusion rule base contains numerous algorithms and frameworks, and is improved with the continuous access to data and the continuous expansion of fusion objectives.

[0115] The framework shown here consists of three main parts:

[0116] 1) Homogeneous data aggregation mainly involves matching different source data through difference analysis, data analysis, and relationship analysis to find a set of homogeneous data with high similarity. Through semantic relationships and knowledge association, factors related to urban environmental assessment are found and features are extracted, such as meteorological data, land cover, pollutants, and remote sensing images.

[0117] 2) Construct sub-classifiers. In homogeneous multi-source data, select appropriate modeling methods for data types to reason about the task objectives. For urban environmental assessment, three sub-classifiers are mainly considered: time classifier, spatial classifier and image classifier, which respectively identify the temporal change pattern of urban environment, the current spatial distribution status and conduct real-time assessment of the current environmental status based on image information.

[0118] 3) Model ensemble: Based on the inference results of multiple sub-classifiers, the models are ensembled to obtain the final inference model. This chapter uses a multi-layer neural network based on Extreme Learning Machine (ELM) to aggregate the multiple sub-classifiers.

[0119] The present invention has the following advantages:

[0120] 1. By utilizing technologies such as knowledge graphs, visual knowledge, and deep learning, we can automatically perform 3D detection, segmentation, vector tracking, and attribute insertion for urban entities. This will organize multi-source heterogeneous and multimodal spatial big data in the physical world into a complex and massive data semantic network, solving the problems of similarity and inconsistency in geometric location, attribute semantics, and logic of cross-domain data.

[0121] 2. By combining space-air-ground integrated multi-source 3D data fusion and visualization technology, the transformation from static 3D visualization to intelligent dynamic visualization can be achieved.

[0122] 3. Construct a multi-source heterogeneous spatiotemporal data resource pool to achieve transparent integration of urban government big data from multi-source, heterogeneous, and closed systems.

[0123] Obviously, those skilled in the art will understand that the various units or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device, or alternatively, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by the computing device. Alternatively, they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0124] The above description is a further detailed explanation of the present invention in conjunction with specific preferred embodiments. It should not be considered that the specific embodiments of the present invention are limited to this. For those skilled in the art, several simple deductions or substitutions can be made without departing from the concept of the present invention, and all such deductions or substitutions should be considered to fall within the scope of protection of the present invention as defined by the submitted claims.

Claims

1. A method for transparent fusion of multi-source spatiotemporal data based on urban information units, characterized in that: Description steps S110 for the division of urban information units and integration of multi-source heterogeneous data: Based on the city management level, the city is divided as a whole, spatial basic information entities are constructed, and multi-level city information units and multi-source heterogeneous spatiotemporal data contained in the city information units are defined. Feature extraction step S120 for multi-source heterogeneous data: Extract spatial, temporal, and relational attributes from the features of geographic entities, and establish comprehensive geographic unit coding identifiers based on location and temporal attributes; Step S130: Multi-source heterogeneous data matching based on city units: Based on a unified geographic unit coding identifier, data from multiple departments and types are integrated into a single platform and statistically displayed at multiple scales according to a grid, enabling rapid request, generation, and service of data based on a grid-like hierarchical structure. In step S120, Attribute extraction is accomplished in the following way: For text data, contextual features are automatically learned through deep learning models, and temporal information is extracted using random fields and maximum entropy models; attribute features and attribute values ​​related to geographic entities are obtained using rule matching and supervised learning methods. For image data: Convolutional neural networks are used to identify objects and scenes and automatically generate content descriptions through image recognition technology; The temporal feature extraction of the data is achieved by using the least squares criterion matching method and the interpolation and extrapolation time matching algorithm to synchronize asynchronous information about the same target from different sources to the same moment, thereby realizing the temporal registration of spatial data. Spatial feature extraction from data involves semantic address matching of spatial information within the data, given a corpus dataset. The goal of semantic address matching is to find address pairs. ,satisfy ,in , and To ensure the accuracy of the spatial location of the data, The integrated geographic unit coding identifier consists of a location code, a semantic code, a time code, and an association code.

2. The method for transparent fusion of multi-source spatiotemporal data based on urban information units according to claim 1, characterized in that: Step S110 includes: Urban information unit definition and division sub-step S111: The city is divided into geographically independent urban units according to the city's management level. Basic government data and social sensing data accumulated in urban management are integrated into the urban units to obtain urban information units. Each unit contains the joint features of all units in the next lower level. The overall architecture of urban information is constructed based on the multi-level urban units. City Information Unit Data Definition Sub-step S112: The urban information unit contains multi-source heterogeneous spatiotemporal data, which is based on basic government data and social sensing data. Specifically, it includes: economic data, environmental data, construction data, and social data. The economic data includes: social security and economic development data; the environmental data includes: ecological and environmental protection, air quality, and water quality data; the construction data includes: urban and rural construction and transportation data; and the social data includes: public opinion, POI data, mobile phone signaling, and audio and video data. The basic government data and social sensing data are divided into five major types: text, images, videos, web pages, and tables. Data integration step S113: The process involves filtering and screening multi-source heterogeneous spatiotemporal data, removing unreasonable data, eliminating homonyms and synonyms, verifying consistency, deleting redundant data, and merging data.

3. The method for transparent fusion of multi-source spatiotemporal data based on urban information units according to claim 1, characterized in that: The coding rules for integrated geographic unit identifiers are as follows: 1) Location code: The administrative region code has 9 digits, consisting of province, city, district and street. The coding conforms to the provisions of GB / T 2260 and GD / T10114. An additional 6 digits are added to represent the area and the smallest grid. 2) Semantic code: Represents data attribute information; 3) Time code: Represents the time elements "year", "month" and "day" in the time when the information unit was generated, in accordance with GB / T 7408; add 4 digits to this to indicate whether it is a holiday or a time interval. 4) Association code: The valid range is between 01 and 99, which is used to mark the association strength between this city information unit and related information units.

4. The method for transparent fusion of multi-source spatiotemporal data based on urban information units according to claim 1, characterized in that: Step S130 includes: Data and geographic unit linking sub-step S131: By integrating, extracting features from, and encoding multi-source spatiotemporal data, dynamic linking of data to geographic units is achieved based on location codes; Multi-level connector step S132: By matching city entities identified by geographic unit codes with data, multi-level linkage can be achieved.

5. The method for transparent fusion of multi-source spatiotemporal data based on urban information units according to claim 2, characterized in that: In step S110, The city units are set up according to the city management level, including different levels such as province, city, administrative region, street, region and grid. Different management levels form an inclusion relationship. Based on the management level, the city is divided into multiple geographically independent city units.

6. The method for transparent fusion of multi-source spatiotemporal data based on urban information units according to claim 5, characterized in that: The urban units, from largest to smallest, are province, city, administrative district, street, region, and geographical grid.

7. The method for transparent fusion of multi-source spatiotemporal data based on urban information units according to claim 5, characterized in that: Urban units can address different urban management issues and apply different scales.

Images