A query method for dynamic multi-objective space-time problems based on a grid space-time knowledge graph

Abstract

The query method for dynamic multi-target space-time problems based on the grid space-time knowledge graph of the disclosure encodes space-time data based on a global split grid to obtain a geographical grid set of the space-time data; encodes the time slices of the space-time data in binary based on a GeoSOT-T framework, constructs a grid space-time knowledge graph subgraph on each time slice, and arranges the grid space-time knowledge graph subgraphs in chronological order to obtain a grid space-time knowledge graph; takes the space-time data, the attributes and directions associated with the space-time data, and the geographical grids of the corresponding levels within the spatial range of the grid space-time knowledge graph subgraph as graph node elements; constructs graph edge elements according to the space-time relationships or semantic relationships between the graph node elements, queries the dynamic multi-target space-time problems based on the grid space-time knowledge graph, and outputs query answers; and can solve the problems of non-unique determination of space-time description, weak ability of space-time relationship completion, and low efficiency of space-time problem query of existing geographical knowledge graphs.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of knowledge graph, and particularly relates to a query method for dynamic multi-target space-time problems based on a grid space-time knowledge graph. BACKGROUND

[0002] With the rapid development of the Internet and mobile terminals, people have entered the era of big data. A large number of space-time big data with space-time location characteristics are complex and dynamically evolving, so that the expression and calculation of space-time big data and space-time knowledge have become a research hotspot. "Knowledge graph" is a common formalized expression of knowledge and information, and the essence of the knowledge graph is a semantic network representing entities in the real world and the relationships between the entities. The geographic space-time knowledge graph is an extension of the knowledge graph in time and space, and is a network knowledge structure formed by describing the concepts, entities, attributes and mutual relationships in the field of geographic space-time information. Through the construction of the geographic space-time knowledge graph, the complex space-time query and the related research of specific application scenarios, the current geographic knowledge graph mainly has the following deficiencies: the space-time description is not uniquely determined, the ability to calculate the space-time relationship is weak, and the space-time problem query efficiency is low.

[0003] The entity space-time description of the traditional geographic space-time knowledge graph has non-uniqueness and uncertainty. Geographic entities, such as buildings, mountains and cities, are highly heterogeneous, diverse and incomplete. Therefore, this rich data source is difficult to be directly used in real-world applications. The space-time description of the traditional geographic space-time knowledge graph is not unique, and the limited association is difficult to perform complex space-time reasoning between entities. The purpose of geographic spatial analysis is to find the spatial relationship between geographic entities and related entities. The existing knowledge graph describes the location of the same island, which may be directly described by latitude and longitude, or may select different reference objects for description. Such non-unique space-time description is difficult to use for complex spatial analysis in the knowledge graph.

[0004] The current geographic space-time knowledge graph has weak ability to calculate and complete space-time relationships. Knowledge graph question answering has the problems of data scarcity and cold start, which are difficult to solve. Although substantial progress has been made in open-domain question answering, it is still difficult for the knowledge graph to answer questions involving geographic entities or concepts and requiring space-time operations. Most geographic knowledge graph models that consider space mainly rely on geographic location labels, but these models lose information such as relative distance, direction and topological relationship between entities, which cannot be all stored in the graph database. The current knowledge graph query only answers the space-time relationships stored in the graph database. For multi-target and multi-condition space-time problems, it is difficult to answer and subsequently complete the corresponding relationships.

[0005] The current geospatial knowledge graph has low query and retrieval efficiency for complex spatio-temporal problems. When a large number of geospatial objects are queried, the efficiency of spatial and temporal indexes is crucial. However, spatio-temporal query efficiency is still a challenge for cross-database interoperability. Existing geospatial knowledge graphs use latitude and longitude methods for queries. However, as the size of spatio-temporal data increases, the number of relationships expands exponentially, and the efficiency of spatial queries using latitude and longitude methods will be greatly reduced. In recent years, some global discrete partition grid models have shown excellent performance in massive spatio-temporal data queries by taking advantage of the algebraic calculation of binary one-dimensional encoding and the spatio-temporal clustering of grid encoding. There is an urgent need for a geospatial knowledge graph that uses spatio-temporal discrete partition grids for spatial environment modeling and answer queries. SUMMARY

[0006] The present application overcomes one of the deficiencies of the prior art and provides a dynamic multi-target spatio-temporal problem query method based on a grid spatio-temporal knowledge graph, which can solve the problems of non-unique determination of spatio-temporal description, weak ability to complete spatio-temporal relationships, and low spatio-temporal problem query efficiency in existing geospatial (grid) spatio-temporal knowledge graphs.

[0007] According to an aspect of the present disclosure, a dynamic multi-target spatio-temporal problem query method based on a grid spatio-temporal knowledge graph is provided, the method comprising:

[0008] Grid encoding the spatio-temporal data based on a global partition grid to obtain a geospatial grid set of the spatio-temporal data;

[0009] Binary encoding the time slices corresponding to the spatio-temporal data based on a GeoSOT-T framework, and constructing a grid spatio-temporal knowledge graph subgraph on each time slice, arranging the grid spatio-temporal knowledge graph subgraphs in chronological order to obtain a basic framework of the grid spatio-temporal knowledge graph;

[0010] The grid spatio-temporal knowledge graph subgraph includes graph node elements and graph edge elements, the spatio-temporal data, its associated attributes and directions, and the geospatial grid corresponding to the level within the spatial range of the grid spatio-temporal knowledge graph subgraph are taken as the graph node elements of the grid spatio-temporal knowledge graph subgraph; the graph edge elements of the grid spatio-temporal knowledge graph subgraph are constructed according to the spatio-temporal relationships or semantic relationships between the graph node elements of the grid spatio-temporal knowledge graph subgraph;

[0011] Performing spatio-temporal queries on the dynamic multi-target spatio-temporal problem based on the grid spatio-temporal knowledge graph, and outputting the query answers of the dynamic multi-target spatio-temporal problem.

[0012] In one possible implementation, the grid encoding of the spatio-temporal data based on a global partition grid to obtain a geospatial grid set of the spatio-temporal data comprises:

[0013] Grid encoding the spatio-temporal data according to the location of the spatio-temporal data and the required representation accuracy based on a global dissected grid, to obtain a set of geographical grids of different scales abstracting the spatio-temporal data, each geographical network in the geographical grid set inherits the attribute of the spatio-temporal data at the location of the geographical grid.

[0014] In a possible implementation, when the graph node of the grid spatio-temporal knowledge graph is a geographical grid element, the geographical grid element is represented as an identified spatial range, and the geographical grid is associated with the spatio-temporal data, time, and attribute within the geographical grid;

[0015] When the graph node of the grid spatio-temporal knowledge graph is an entity element, the graph node is described as an independent geographical spatio-temporal data, and the entity element includes an identifier, a category, and a state;

[0016] When the graph node of the grid spatio-temporal knowledge graph is a heading element, the graph node is represented as a dynamic entity or a forward direction at the current time slice;

[0017] When the graph node of the grid spatio-temporal knowledge graph is an event element, the graph node is represented as a natural phenomenon or a human phenomenon affecting the current spatio-temporal scene;

[0018] When the graph node of the grid spatio-temporal knowledge graph is an attribute element, the graph node is represented as graph node feature information containing environmental information and semantic information.

[0019] In a possible implementation, the geographical grid is also used as a grid geographical implicit layer parameter to participate in spatio-temporal query and calculation between graph nodes of the grid spatio-temporal knowledge graph.

[0020] In a possible implementation, the graph edge element of the grid spatio-temporal knowledge graph includes two types of explicit relationship and implicit relationship;

[0021] In a possible implementation, the explicit relationship is used to represent the spatio-temporal relationship or semantic relationship between the graph nodes of the grid spatio-temporal knowledge graph, and is stored in the grid spatio-temporal knowledge graph;

[0022] The implicit relationship is hidden in the grid geographical implicit layer of the geographical grid, and is used to activate the geographical grid parameter to perform spatio-temporal relationship calculation and real-time completion of implicit spatio-temporal relationship when querying and calculating the graph node of the implicit relationship from the grid spatio-temporal knowledge graph.

[0023] In a possible implementation, the implicit relationship includes an entity-heading relationship, an entity-entity spatial relationship, and an entity-event influence relationship.

[0024] wherein, for the entity-direction relationship, if the entity has a direction in the same time slice, two graph nodes of the grid spatio-temporal knowledge graph are directly associated, otherwise the direction is calculated according to the geographical grid of the entity in the last or current time slice, and the entity-direction relationship is completed;

[0025] for the entity-entity spatial relationship, including topological spatial relationship, distance spatial relationship and direction spatial relationship, wherein the topological spatial relationship includes coincidence relationship, adjacent relationship and discontinuous relationship;

[0026] for the entity-event influence relationship, including threat relationship and incentive relationship, wherein the threat relationship is that the event influences the entity to escape the geographical grid where the event is located, and the incentive relationship is that the event influences the entity to go to the geographical grid where the event is located.

[0027] In a possible implementation, based on the grid spatio-temporal knowledge graph, the spatio-temporal problem of the dynamic multi-target is queried, and a query answer of the spatio-temporal problem of the dynamic multi-target is output, comprising:

[0028] P1: input the query sentence of the spatio-temporal problem of the dynamic multi-target based on the grid spatio-temporal knowledge graph, perform triple query of the subject, and the triple contains space-time elements;

[0029] P2: when the triple predicate of the query is an implicit relationship, perform spatio-temporal complex deduction query on the spatio-temporal problem of the dynamic multi-target, input the geographical grid associated with the entity or event of the triple of the implicit relationship into the grid geographical implicit layer, and activate the corresponding geographical grid parameters in the implicit layer; wherein the grid geographical implicit layer is a three-dimensional tensor arranged in a stereoscopic grid coding order;

[0030] P3: according to the implicit relationship and the stereoscopic grid spatio-temporal coding algebra, the grid geographical implicit layer performs topological calculation, distance calculation or direction calculation on the activated geographical grid, and outputs the spatio-temporal calculation result;

[0031] P4: according to the spatio-temporal calculation result, complete the spatio-temporal implicit relationship, and input the query result into the next sentence of the query sentence;

[0032] P5: if the query reaches the last query sentence, decode the geographical grid into latitude and longitude, and output the spatio-temporal complex deduction query result of the dynamic multi-target.

[0033] In a possible implementation, based on the grid spatio-temporal knowledge graph, the spatio-temporal problem of the dynamic multi-target is queried, and a query answer of the spatio-temporal problem of the dynamic multi-target is output, further comprising:

[0034] When the query triple predicate is an explicit relationship, a simple query or a regular spatio-temporal query is performed on the dynamic multi-target spatio-temporal problem, and the query result is input to the next sentence of the query statement;

[0035] If the query reaches the last query statement, the geographic grid is decoded into latitude and longitude, and the spatio-temporal complex deduction query result of the dynamic multi-target is output.

[0036] In a possible implementation, the simple query is reasoning on the existing semantic relationship of the grid spatio-temporal knowledge graph;

[0037] The regular spatio-temporal query is to query all entities, time slices, geographic grids and graph nodes related to the geographic grids within a certain spatio-temporal range.

[0038] The query method of the dynamic multi-target spatio-temporal problem based on the grid spatio-temporal knowledge graph of the present disclosure encodes the spatio-temporal data based on the global split grid to obtain a geographic grid set of the spatio-temporal data; binary encodes the time slices corresponding to the spatio-temporal data based on the GeoSOT-T framework, and constructs a grid spatio-temporal knowledge graph subgraph on each time slice; arranges the grid spatio-temporal knowledge graph subgraphs in time sequence to obtain the grid spatio-temporal knowledge graph; the grid spatio-temporal knowledge graph subgraph includes graph node elements and graph edge elements, and the spatio-temporal data, its associated attributes and directions, and the corresponding level geographic grids within the grid spatio-temporal knowledge graph subgraph space are taken as the graph node elements of the grid spatio-temporal knowledge graph subgraph; the graph edge elements of the grid spatio-temporal knowledge graph subgraph are constructed according to the spatio-temporal relationship or semantic relationship between the graph node elements of the grid spatio-temporal knowledge graph subgraph; the grid spatio-temporal knowledge graph is used to query the dynamic multi-target spatio-temporal problem, and the query answer of the dynamic multi-target spatio-temporal problem is output. Through the grid geographic implicit layer, dynamic multi-target spatio-temporal problems can be answered, and various implicit relationships can be completed. By constructing the grid spatio-temporal knowledge graph, the algebraic calculation of the spatio-temporal grid encoding replaces the storage of complex massive spatio-temporal relationships, and the knowledge graph supports the efficient expression of multi-source heterogeneous data of geographic space nodes, provides spatio-temporal object storage, retrieval, reasoning, calculation and management capabilities, and can solve the problems of non-unique determination of existing geographic knowledge graph spatio-temporal description, weak ability of spatio-temporal relationship completion, and low efficiency of spatio-temporal problem query. BRIEF DESCRIPTION OF DRAWINGS

[0039] The accompanying drawings are used to provide a further understanding of the technical solutions of the present application or the prior art, and constitute a part of the specification. Among them, the drawings expressing the embodiments of the present application are used together with the embodiments of the present application to explain the technical solutions of the present application, but do not constitute a limitation on the technical solutions of the present application.

[0040] Figure 1 A query method flowchart of a dynamic multi-target spatio-temporal problem based on a grid spatio-temporal knowledge graph is shown according to an embodiment of the present disclosure.

[0041] Figure 2 A GeoSOT-T time partitioning encoding structure diagram is shown according to an embodiment of the present disclosure.

[0042] Figure 3 A grid spatio-temporal knowledge graph diagram is shown according to an embodiment of the present disclosure.

[0043] Figure 4 A spatio-temporal complex deduction query algorithm flowchart based on a grid spatio-temporal knowledge graph is shown according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0044] The embodiments of the present application will be described in detail below with reference to the accompanying drawings and embodiments, so that how the present application applies technical means to solve technical problems and achieves corresponding technical effects can be fully understood and implemented. The embodiments of the present application and each feature in the embodiments can be combined with each other without conflict, and the formed technical solutions are all within the protection scope of the present application.

[0045] In addition, the steps shown in the flowchart of the accompanying drawings can be executed in a computer such as a group of computer executable instructions. Moreover, although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown.

[0046] Figure 1 A query method flowchart of a dynamic multi-target spatio-temporal problem based on a grid spatio-temporal knowledge graph is shown according to an embodiment of the present disclosure. The method can be used in the construction of a knowledge graph in a complex geographical environment and the calculation and query of a spatio-temporal problem, etc. As shown in the figure, the method can include: Figure 1

[0047] Step S1: Grid encoding of spatio-temporal data based on a global partitioning grid, to obtain a geographical grid set of the spatio-temporal data, wherein the spatio-temporal data includes entities and events.

[0048] ​Global subdivision grid selects a subdivision and coding method for subdividing the earth surface space into a grid (GeoSOT), the full name of which is "2n and integer one-dimensional array global latitude and longitude subdivision grid". The subdivision space of the earth surface space grid subdivision adopts the coordinate space of latitude and longitude. In order to ensure the integer division of degrees, minutes and seconds, the earth space is expanded into a latitude and longitude space of 2 integer powers, defined as a 512°x512° grid. The 60' space of each degree is expanded to 64', and the 60" space of each minute is expanded to 64". The grid is subdivided to 32 levels, and the minimum expression precision can be accurately to 1.5 cm. The part in the range of 180°x360° in the expanded latitude and longitude coordinate space is consistent with the actual geographical space, and the part beyond the range of 180°x360° has no actual geographical significance.

[0049] In an example, grid coding the spatio-temporal data based on the global subdivision grid to obtain a geographical grid set of the spatio-temporal data can include:

[0050] Grid coding the spatio-temporal data based on the global subdivision grid according to the location and required representation precision of the spatio-temporal data to obtain a set of geographical grids of different scales abstracted from the spatio-temporal data. In addition, the geographical grid is also used as a grid geographical implicit layer parameter to participate in the spatio-temporal query and calculation between the graph nodes of the grid spatio-temporal knowledge graph.

[0051] Among them, the spatio-temporal data is mainly used to represent the entities (Entity) and events (Event) of the grid spatio-temporal knowledge graph. Based on the global subdivision grid, the spatio-temporal data is grid coded from the highest grid level to the lowest grid level to form a 32-level full-coverage grid space coding. At the same time, different spatio-temporal data can be abstracted and expressed as a set of geographical grids of different scales according to their location and required representation precision.

[0052] In the grid spatio-temporal knowledge graph, a grid level is usually selected to connect each spatio-temporal data to a unique GeoSOT grid (geographical grid) to avoid redundancy of spatio-temporal knowledge query. The spatio-temporal data is coded using the GeoSOT grid to represent the spatial information of each spatio-temporal data. The GeoSOT grid (geographical grid) is used as a node of the grid spatio-temporal knowledge graph, and the geographical grid is associated with entities, events and attributes within the geographical grid range.

[0053] Step S2: binary coding the time slice corresponding to the spatio-temporal data based on the GeoSOT-T framework, and constructing a grid spatio-temporal knowledge graph subgraph on each time slice, and arranging the grid spatio-temporal knowledge graph subgraph in time sequence to obtain the grid spatio-temporal knowledge graph.

[0054] Figure 2A diagram of a GeoSOT-T time partition coding structure is shown according to an embodiment of the present disclosure.

[0055] GeoSOT-T is a time discrete partition framework or a time partition coding model, and the core idea is to perform piecewise binary division and multi-scale coding on the time domain. As shown in Figure 2 , GeoSOT-T uniformly partitions the time domain from 131072 BC to 131071 AD into 64 levels in a time multi-scale manner, the range of time expression covers about 260,000 years before and after AD, the granularity of time expression is microsecond level, and the binary tree partition structure of year, month, day, hour, minute, and second is realized, thereby forming a multi-scale and unified discrete time coding system. Through GeoSOT-T, time is discretized into time slices (i.e., time intervals or time segments) of different lengths, and a unique binary code is assigned, so that time is discretized into time periods with lengths, and can be unified with the specification of the stereoscopic partition grid coding.

[0056] A grid spatio-temporal knowledge graph subgraph is constructed on each time slice to complete entity relationship analysis of the time sequence; the grid spatio-temporal knowledge graph subgraphs are arranged in time sequence on each time slice to obtain a basic framework of the grid spatio-temporal knowledge graph.

[0057] For example, let the time slice set T be:

[0058]

[0059] where t k is a unit time slice. The interval of the time slice is determined by the actual scene in the knowledge graph, for example, when the time sequence of the spatio-temporal data in the scene is closely arranged, the time interval is smaller.

[0060] Each time slice is connected to each node in the grid spatio-temporal knowledge graph (as a special node). Then the grid spatio-temporal knowledge graph can be defined as:

[0061]

[0062] where KG k is a sub-knowledge graph under the time slice t k , each KG k is composed of graph nodes and graph edges, and k is a positive integer. In this way, by arranging the knowledge graph KG k in time sequence on each time slice, a basic framework of the grid spatio-temporal knowledge graph is obtained.

[0063] Step S3: The grid spatiotemporal knowledge graph subgraph includes graph node elements and graph edge elements. The spatiotemporal data, its associated attributes and headings, and the corresponding level of geographic grids within the spatial range of the grid spatiotemporal knowledge graph subgraph are used as the graph node elements of the grid spatiotemporal knowledge graph subgraph. The graph edge elements of the grid spatiotemporal knowledge graph subgraph are constructed based on the spatiotemporal or semantic relationships between the graph node elements of the grid spatiotemporal knowledge graph subgraph.

[0064] Each graph unit or graph subgraph unit is in the form of a triple (subject, predicate, object), where the subject and predicate are two graph node elements of the grid spatiotemporal knowledge graph subgraph, and the object is the empty relation or semantic relation described between the two graph node elements, that is, the graph edge element of the grid spatiotemporal knowledge graph subgraph.

[0065] Figure 3 A schematic diagram of a grid-based spatiotemporal knowledge graph according to an embodiment of the present disclosure is shown.

[0066] Based on time slices, such as Figure 3 As shown, a grid-based spatiotemporal knowledge graph can include five different types of graph node elements: geographic grid, entity, heading, event, and attribute, as well as two different types of graph edge elements: explicit relation and implicit relation. Attributes can include environmental attributes and semantic attributes.

[0067] In one example, when the graph nodes of the spatiotemporal knowledge graph are geographic grid features, the geographic grid features represent the spatial range of identification, and the geographic grid is associated with the spatiotemporal data, time and attributes within the geographic grid, which can avoid repeated descriptions of the same basic type and reference location.

[0068] Geographic grid (Locgrid) elements can be represented as:

[0069]

[0070] Among them, Locgird code The coded name representing a geographic grid (spatial grid), Locgird level This represents the level of a geographic grid (spatial grid). Among them, Locgird... level The lower the level, the more locale code The larger the spatial range of the mapping, the better.

[0071] When the graph node of the grid spatio-temporal knowledge graph is an entity element, the graph node is described as an independent geospatial data, and the entity element includes an identifier, a category, and a state;

[0072] For example, when the graph node is an entity element, the graph node is described as an independent geographic object and can be distinguished from other entities. The entity element can be represented as:

[0073]

[0074] wherein, Entity name represents the unique identifier of the entity, Entity class represents the category of the entity, and Entity state represents the state of the entity. The state of the entity is divided into static and dynamic, and is used to indicate whether the entity can move. The dynamic entity will be associated with a heading element. Compared with the entity of the traditional knowledge graph, the grid spatio-temporal knowledge graph has an advantage in analyzing the spatial relationship of dynamic entities such as airplanes or ships.

[0075] When the graph node of the grid spatio-temporal knowledge graph is a heading element, the graph node represents a dynamic entity or a direction of movement at the current time slice.

[0076] wherein, the heading element can be represented as:

[0077]

[0078] Heading direction is the true heading of the dynamic entity, that is, the angle between the true north line and the heading line. The true heading is measured in the clockwise direction with the true north line (0°) as the basis, and the range is 0°-360°. The heading can be used to analyze the destination or intention of the related entity.

[0079] When the graph node of the grid spatio-temporal knowledge graph is an event element, the graph node represents a natural phenomenon or a man-made phenomenon that affects the current spatio-temporal scene.

[0080] wherein, the event element can be represented as:

[0081]

[0082] wherein, Event name is the unique identifier of the event element, Event threat represents the type of event based on the threat level, and Event threshold represents the farthest distance that the event element will affect.

[0083] When the graph node of the grid spatio-temporal knowledge graph is an attribute element, the graph node is represented as graph node feature information containing environmental information and semantic information. Each geographic grid (Locgrid), entity (Entity), and event (Event) can have attributes associated therewith.

[0084] An attribute (Attribute) element can be represented as:

[0085]

[0086] wherein, Attribute value represents an attribute value, and Attribute class represents an attribute category.

[0087] In an example, as shown in FIG. 1, the graph edge element of the grid spatio-temporal knowledge graph can include two different types of graph edge elements, namely explicit relation (Explicit Relation) and implicit relation (Implicit Relation). Figure 3

[0088] The explicit relation is used to represent the spatio-temporal relation or semantic relation between the graph nodes of the grid spatio-temporal knowledge graph and is stored in the grid spatio-temporal knowledge graph.

[0089] The implicit relation is not stored in the grid spatio-temporal knowledge graph and is used to activate the geographic grid parameter when the graph node is queried and calculated from the grid spatio-temporal knowledge graph.

[0090] For example, the explicit relation can directly represent a simple spatio-temporal relation or semantic relation between nodes, mainly including time relation, spatial location relation, and semantic relation between graph nodes. The time relation, spatial location relation, and semantic relation between graph nodes can be represented as:

[0091]

[0092] Entity-Locgrid relation = <Entity i ∈KG k , Located in, Locgird j ∈KG k >

[0093] Entity-Attribute relation = <Entity i ∈KG k , has, Attribute j ∈KG k >​

[0094] In the formula, time change, Located in, has respectively represent the predicate edge element in the three tuple of time relationship, spatial location relationship and semantic relationship, and the two graph node elements before and after are the subject and object of the three tuple.

[0095] The implicit relationship can include entity-heading relationship, entity-entity spatial relationship and event-entity influence relationship. The implicit relationship refers to the spatio-temporal relationship not stored in the graph, which can be obtained by spatio-temporal grid algebra calculation. The implicit relationship is initially hidden in the implicit layer of the geographic grid (Locgrid). When the corresponding calculation result is inferred from the grid spatio-temporal knowledge graph query, the relevant geographic grid (Locgrid) parameters are activated to realize the real-time completion of the implicit relationship.

[0096] For the entity-heading relationship, if the entity has the same time slice heading, the two graph nodes of the grid spatio-temporal knowledge graph are directly associated, otherwise the heading is calculated according to the geographic grid of the entity in the last or current time slice, and the entity-heading relationship is completed.

[0097] The entity-heading relationship can be represented as:

[0098]

[0099] In the formula, is Led by represents the predicate edge element of the entity-heading relationship. Through the relationship between the entity (Entity) and the heading (Heading) in the current time slice, the future entity intention can be predicted, and it can also be used as an auxiliary condition to check the entity trajectory.

[0100] For the entity-entity spatial relationship, it includes topological spatial relationship, distance spatial relationship and direction spatial relationship, wherein the topological spatial relationship includes coincidence relationship, adjacent relationship and discontinuous relationship.

[0101] The entity-entity spatial relationship can be represented as:

[0102] Entity-Entity relation =<Entity i ∈KG k ,Spatiallocation,Entity j ∈KG k >。

[0103] In the formula, Spatial location represents the predicate edge element of the entity-entity spatial relationship, which includes topological spatial relationship, distance spatial relationship and direction spatial relationship.

[0104] For entity-event influence relations, including threaten (Threaten) relations and motivate (Motivate) relations, wherein the threaten relation is that the event influences the entity to escape the geographic grid where the event is located, and the motivate relation is that the event influences the entity to go to the geographic grid where the event is located.

[0105] Event-Entity

[0106] Event-Entity relation =<Event i ∈KG k ,Influence,Entity j ∈KG k >

[0107] In the formula, Influence represents the predicate edge element of the entity-event influence relation, that is, threaten (Threaten) or motivate (Motivate). Through the entity-event influence relation, it can be judged that the event influences the relationship between the entity and the geographic grid where the event is located.

[0108] Step S4: performing a spatiotemporal query on the spatiotemporal problem of the dynamic multi-target based on the grid spatiotemporal knowledge graph, and outputting a query answer of the spatiotemporal problem of the dynamic multi-target.

[0109] Figure 4 A flow chart of a spatiotemporal complex deduction query algorithm based on a grid spatiotemporal knowledge graph according to an embodiment of the present disclosure is shown. As shown in Figure 4 , this step can include:

[0110] P1: inputting a query sentence of the spatiotemporal problem of the dynamic multi-target based on the grid spatiotemporal knowledge graph, performing a triple query of a main sentence, and the triple containing a space-time element;

[0111] P2: when the query triple predicate is an implicit relation, performing a spatiotemporal complex deduction query on the spatiotemporal problem of the dynamic multi-target, inputting a geographic grid (Locgrid) associated with an entity (Entity) or an event (Event) of the triple of the implicit relation into a grid geographic implicit layer, and activating a corresponding geographic grid parameter in the implicit layer; wherein the grid geographic implicit layer is a three-dimensional tensor arranged in a stereoscopic grid coding order;

[0112] P3: according to an implicit relation and a stereoscopic grid spatiotemporal coding algebra, the grid geographic implicit layer performs topological calculation, distance calculation or direction calculation on the activated geographic grid, and outputs a spatiotemporal calculation result.

[0113] For example, in the grid geographic implicit layer, the spatio-temporal implicit calculation mainly includes distance calculation, topological relation calculation and direction calculation. In order to improve the calculation efficiency, the height dimension encoding of the geographic grid (Locgrid) can be ignored in the case of small height variation.

[0114] Suppose that two positioning grids Locgrid A and Locgrid B are given, the spatio-temporal calculation in the grid geographic implicit layer is expressed by encoding algebraic operation as follows:

[0115] The distance calculation in the grid geographic implicit layer is defined as:

[0116]

[0117]

[0118]

[0119] In the formula, level scale is the grid level of the grid geographic current level.f span represents the grid distance of movement in the corresponding dimension, g1 and g2 are the encoding first digits of the corresponding levels of the positioning grids Locgrid A and Locgrid B , g1 = g2 indicates that the graph nodes of the grid spatio-temporal knowledge graph are located in the same hemisphere, and g1 ≠ g2 indicates that the positions of the graph nodes of the grid spatio-temporal knowledge graph are different.

[0120] The topological calculation in the grid geographic implicit layer is defined as:

[0121]

[0122] In the formula, when d grid = 1, the two graph nodes of the grid spatio-temporal knowledge graph are edge adjacent; and when d grid = 2, the two graph nodes of the grid spatio-temporal knowledge graph are corner adjacent.

[0123] The direction calculation in the grid geographic implicit layer is defined as:

[0124]

[0125] In the formula, f span represents the grid distance of movement in the latitude or longitude, and the calculation method is consistent with f span in the distance calculation of the grid geographic implicit layer.

[0126] P4: According to the spatio-temporal calculation result, the spatio-temporal implicit relationship is completed, and the query result is input to the next sentence of the query statement;

[0127] P5: If the query is to the last query statement, the geographic grid is decoded into longitude and latitude, and the spatio-temporal complex deduction query result of the dynamic multi-target is output.

[0128] In a possible implementation, as shown in Figure 4 based on the grid spatio-temporal knowledge graph, the spatio-temporal problem of the dynamic multi-target is queried, and the result of the spatio-temporal problem of the dynamic multi-target is obtained, and the method further comprises:

[0129] When the query triple predicate is an explicit relationship, the spatio-temporal problem of the dynamic multi-target is simply queried or conventionally queried, the query result is input to the next sentence of the query statement, and if the query is to the last query statement, the geographic grid is decoded into longitude and latitude, and the spatio-temporal complex deduction query result of the dynamic multi-target is output.

[0130] The simple query can refer to reasoning of an existing semantic relationship in the knowledge graph, such as an entity attribute and an associated relationship. The simple query belongs to a basic knowledge semantic question and answer, and the simple query method of the grid spatio-temporal knowledge graph is similar to other knowledge graphs, which is not described in detail here.

[0131] The conventional spatio-temporal query can refer to that a triple of a query statement contains a space-time element (such as a geographic grid, a time slice, and a time change). When the conventional spatio-temporal query is a spatio-temporal range query, all entities (Entity) in a certain spatio-temporal range are queried. The time slice to be queried in the spatio-temporal range query is determined, all geographic grids (Locgrid) in the spatio-temporal range are calculated, and all nodes related to all geographic grids (Locgrid) are queried.

[0132] The query method of the dynamic multi-target spatio-temporal problem based on the grid spatio-temporal knowledge graph of the present disclosure obtains a geographical grid set of the spatio-temporal data by grid encoding the spatio-temporal data based on a global split grid; performs binary encoding on a time slice corresponding to the spatio-temporal data based on a GeoSOT-T framework, and constructs a grid spatio-temporal knowledge graph subgraph on each time slice; arranges the grid spatio-temporal knowledge graph subgraphs in chronological order to obtain the grid spatio-temporal knowledge graph; the grid spatio-temporal knowledge graph subgraph includes graph node elements and graph edge elements, the spatio-temporal data and its associated attributes and directions, and the geographical grid of the corresponding level within the spatial range of the grid spatio-temporal knowledge graph subgraph are taken as the graph node elements of the grid spatio-temporal knowledge graph subgraph; the graph edge elements of the grid spatio-temporal knowledge graph subgraph are constructed according to the spatio-temporal relationship or semantic relationship between the graph node elements of the grid spatio-temporal knowledge graph subgraph; the spatio-temporal problem of the dynamic multi-target is queried based on the grid spatio-temporal knowledge graph, and the query answer of the spatio-temporal problem of the dynamic multi-target is output. Through the grid geographical implicit layer, the dynamic multi-target spatio-temporal problem can be answered, and various implicit relationships can be completed. The problem that the existing geographical knowledge graph has non-unique certainty in spatio-temporal description, the ability to complete spatio-temporal relationship is weak, and the spatio-temporal problem query efficiency is low can be solved.

[0133] The above examples are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for querying dynamic multi-objective spatio-temporal problems based on a grid spatio-temporal knowledge graph, characterized in that, The method comprises: spatial-temporal data is grid encoded based on a global split grid to obtain a geographical grid set of the spatial-temporal data; time slices corresponding to the spatial-temporal data are binary encoded based on a GeoSOT-T framework, and a grid spatial-temporal knowledge graph subgraph is constructed on each time slice, and the grid spatial-temporal knowledge graph subgraphs are arranged in time sequence to obtain the grid spatial-temporal knowledge graph; the grid spatial-temporal knowledge graph subgraph comprises graph node elements and graph edge elements, the spatial-temporal data, attributes and directions associated therewith, and geographical grids in the grid spatial-temporal knowledge graph subgraph space range are taken as graph node elements of the grid spatial-temporal knowledge graph subgraph, graph edge elements of the grid spatial-temporal knowledge graph subgraph are constructed according to spatial-temporal relationships or semantic relationships between the graph node elements of the grid spatial-temporal knowledge graph subgraph, the graph edge elements of the grid spatial-temporal knowledge graph comprise two types of explicit relationships and implicit relationships, the explicit relationships are used to represent spatial-temporal relationships or semantic relationships between graph nodes of the grid spatial-temporal knowledge graph and are stored in the grid spatial-temporal knowledge graph, the implicit relationships are hidden in a grid geographical implicit layer of the geographical grid and are used to activate the geographical grid parameters to perform spatial-temporal relationship calculation and real-time completion of implicit spatial-temporal relationships when the implicit relationships are queried and calculated from the graph nodes of the grid spatial-temporal knowledge graph, and the geographical grid is also used as a grid geographical implicit layer parameter to participate in the query and calculation between graph nodes of the grid spatial-temporal knowledge graph; spatial-temporal problems of the dynamic multi-target are queried based on the grid spatial-temporal knowledge graph, and query answers of the spatial-temporal problems of the dynamic multi-target are output.

2. The query method of dynamic multi-objective spatio-temporal problems according to claim 1, characterized in that, The grid encoding of the spatial-temporal data based on the global split grid to obtain the geographical grid set of the spatial-temporal data comprises: the spatial-temporal data is grid encoded based on the global split grid according to positions of the spatial-temporal data and required representation accuracy to obtain a set of geographical grids of different scales into which the spatial-temporal data is abstracted, and each geographical network in the geographical grid set inherits attributes of the spatial-temporal data at positions of the geographical grids.

3. The query method of the dynamic multi-target spatial-temporal problem according to claim 1, characterized in that: when the graph node of the grid spatial-temporal knowledge graph is a geographical grid element, the geographical grid element is represented as an identified spatial range, and the geographical grid is associated with spatial-temporal data, time and attributes in the geographical grid; when the graph node of the grid spatial-temporal knowledge graph is an entity element, the graph node is described as an independent geographical spatial-temporal data, and the entity element comprises an identifier, a category and a state; when the graph node of the grid spatial-temporal knowledge graph is a direction element, the graph node is represented as a dynamic entity or a forward direction of time on the current time slice; when the graph node of the grid spatial-temporal knowledge graph is an event element, the graph node is represented as a natural phenomenon or a human phenomenon affecting a current spatial-temporal scene. When the graph node of the grid spatio-temporal knowledge graph is an attribute element, the graph node is represented as graph node feature information containing environmental information and semantic information.

4. The method of claim 1, wherein, The implicit relationship includes an entity-direction relationship, an entity-entity spatial relationship, and an entity-event influence relationship. For the entity-direction relationship, if the entity has a direction in the same time slice, two graph nodes of the grid spatio-temporal knowledge graph are directly associated, otherwise, the direction is calculated according to the geographical grid of the entity in the last or current time slice, and the entity-direction relationship is completed. For the entity-entity spatial relationship, it includes a topological spatial relationship, a distance spatial relationship, and an orientation spatial relationship, wherein the topological spatial relationship includes a coincidence relationship, an adjacent relationship, and a discontinuous relationship. For the entity-event influence relationship, it includes a threat relationship and an incentive relationship, wherein the threat relationship is that the event influences the entity to escape from the geographical grid where the event occurs, and the incentive relationship is that the event influences the entity to go to the geographical grid where the event occurs.

5. The method of claim 1, wherein, Based on the grid spatio-temporal knowledge graph, the spatio-temporal problem of the dynamic multi-target is queried, and the query answer of the spatio-temporal problem of the dynamic multi-target is output, including: P1: input the query sentence of the spatio-temporal problem of the dynamic multi-target based on the grid spatio-temporal knowledge graph, perform triple query of the subject, and the triple contains a space-time element; P2: when the query triple predicate is an implicit relationship, perform spatio-temporal complex deduction query on the spatio-temporal problem of the dynamic multi-target, input the geographical grid associated with the entity or event of the triple of the implicit relationship into the grid geographical implicit layer, and activate the corresponding geographical grid parameters in the implicit layer; wherein the grid geographical implicit layer is a three-dimensional tensor arranged in a stereoscopic grid coding order; P3: according to the implicit relationship and the stereoscopic grid spatio-temporal coding algebra, the grid geographical implicit layer performs topological calculation, distance calculation, or orientation calculation on the activated geographical grid, and outputs the spatio-temporal calculation result; P4: according to the spatio-temporal calculation result, complete the spatio-temporal implicit relationship, and input the query result into the next sentence of the query sentence; P5: if the query reaches the last query sentence, decode the geographical grid into latitude and longitude, and output the spatio-temporal complex deduction query result of the dynamic multi-target.

6. The query method of dynamic multi-objective spatio-temporal problems according to claim 1, characterized in that, The query based on the grid spatio-temporal knowledge graph on the spatio-temporal problem of the dynamic multi-target, and the query answer of the spatio-temporal problem of the dynamic multi-target also includes: When the query triple predicate is an explicit relationship, perform simple query or regular spatio-temporal query on the spatio-temporal problem of the dynamic multi-target, and input the query result into the next sentence of the query sentence; If the query reaches the last query sentence, decode the geographical grid into latitude and longitude, and output the spatio-temporal complex deduction query result of the dynamic multi-target.

7. The query method of the dynamic multi-target spatio-temporal problem according to claim 6, wherein the simple query is reasoning on the existing semantic relationship of the grid spatio-temporal knowledge graph; The conventional spatio-temporal query is to query all entities, time slices, geographic grids and graph nodes related to the geographic grids in a certain spatio-temporal range. The conventional spatio-temporal query is to query all entities, time slices, geographic grids and graph nodes related to the geographic grids in a certain spatio-temporal range.

Images