A knowledge graph-based map knowledge representation method

Abstract

The application belongs to the technical field of map knowledge, and particularly relates to a map knowledge representation method based on a knowledge graph. The method first acquires map data, extracts map knowledge therefrom, including concept class knowledge and rule class knowledge, the concept class knowledge being common sense knowledge in the field of cartography and including map element knowledge; the rule class knowledge being various models, rules and principles formed in the process of map making; then generates a map knowledge network graph based on the knowledge graph according to the map element knowledge; and finally performs knowledge reasoning according to the rule class knowledge, completes the map knowledge network graph, and generates new entities and / or new entity relationships. The application fully mines geographical knowledge contained in a map, provides a basis for all-round knowledge mining and fusion of the map, and has certain reference value for realizing the transformation of 'data-information-knowledge'.

Description

Technical Field

[0001] This invention belongs to the field of map knowledge technology, specifically relating to a map knowledge representation method based on knowledge graphs. Background Technology

[0002] The current phenomenon of "massive data, information explosion, and scarce knowledge" in the field of geography is becoming increasingly serious. To improve the application level of geographic information and promote the sharing and reuse of geospatial knowledge, a transformation from geographic information services to knowledge services is imperative. Maps, as one of the main carriers of geographic knowledge, are an important source of geographical knowledge due to their accuracy and scientific nature. However, for machine intelligence, map data models and representation models cannot be directly provided to machines. Therefore, it is necessary to formally model map knowledge, representing it in machine language that computers can understand, forming a "map knowledge base." This knowledge serves as prior knowledge for machines to understand map scenes, and then machine reasoning is used to achieve "intelligent cognition" of the scene, thereby providing map knowledge services.

[0003] Knowledge graphs possess powerful information integration and knowledge representation capabilities, and are currently one of the main technologies for explicit representation of geospatial information and decision-making reasoning. Many scholars have proposed construction processes for geographic knowledge graphs to address the massive and complex geographical knowledge, achieving significant results in theory, technology, and application.

[0004] In the classification of geographical knowledge, Wang Fuqiang's 2013 master's thesis, "Research on the Theory and Method of Spatial Knowledge Map Construction," categorizes spatial knowledge into factual knowledge, conceptual knowledge, and strategic knowledge based on the varying degrees to which they reflect the essential characteristics and inherent regularities of geographical entities. Lin Hui and You Lan's "A Preliminary Exploration of Virtual Geographical Environment Knowledge Engineering," published in the 17th issue of the *Journal of Geoinformation Science* in 2015, classifies geographical knowledge into declarative, deductive, and conclusive types based on its evolutionary stage. The triplet structure is a representative method in the study of geographical knowledge representation.<D,W,R> (Domain of discourse, world state, relation), quadruple structure<C,R,A,I> (Concepts, relationships, axioms, examples), quintuple structure<C,R,F,A,I> (Concepts, relations, functions, axioms, examples), six-tuple structure<C,AC,R,AR,H,X> (Concepts, attributes, relationships, attribute relationships, hierarchical relationships, axioms). Research findings on the classification and representation of geographical knowledge have significant implications for the classification and modeling of map knowledge. While some scholars have achieved geographical knowledge extraction and semantic modeling based on map data, they have not systematically studied map knowledge classification and modeling, nor have they fully explored the geographical knowledge contained within maps, thus failing to improve map knowledge services. Summary of the Invention

[0005] The purpose of this invention is to provide a map knowledge representation method based on knowledge graphs, in order to solve the problem that existing technologies do not fully explore the geographical knowledge contained in maps, thus failing to improve map knowledge services.

[0006] To address the aforementioned technical problems, this invention provides a map knowledge representation method based on knowledge graphs, comprising the following steps:

[0007] 1) Acquire map data and extract map knowledge from it; wherein, the map knowledge includes conceptual knowledge and rule-based knowledge, the conceptual knowledge is common knowledge in the field of cartography and includes map element knowledge, the map element knowledge includes entities, entity relationships and entity characteristics; the rule-based knowledge is various models, rules and principles formed in the process of map making;

[0008] 2) Based on the map element knowledge, generate a map knowledge network diagram using a knowledge graph;

[0009] 3) Based on rule-based knowledge, perform knowledge reasoning to complete the generated map knowledge network graph, thereby generating new entities and / or new entity relationships, and thus obtaining the completed map knowledge network graph.

[0010] Its beneficial effects are as follows: This invention formally models and represents map knowledge, dividing the extracted map knowledge into conceptual knowledge and rule-based knowledge. Conceptual knowledge includes map element knowledge, thus enabling the generation of a map knowledge network graph using map element knowledge. This network graph is then supplemented based on rule-based knowledge, generating new entities and / or new entity relationships and entity features, completing the initial representation of map knowledge. This process fully explores the geographical knowledge contained in maps, comprehensively showcasing the knowledge contained within them. It provides a foundation for comprehensive knowledge mining and integration of maps, offering valuable reference for the transformation from "data to information to knowledge," and providing an implementation method for better providing map knowledge services.

[0011] Furthermore, in step 1), the extracted map knowledge also includes decision-making knowledge, which is knowledge based on the analysis and application of geographical entities or geographical phenomena to complete relevant reasoning and decision-making; correspondingly, after step 3), step 4) is also included: Step 4) Based on the decision-making knowledge, the completed map knowledge network diagram is used to analyze the various elements in the map in order to conduct relevant reasoning or assist in decision-making.

[0012] Its beneficial effects are as follows: the extracted map knowledge also includes decision-making knowledge, which on the one hand presents the knowledge contained in the map more comprehensively, and on the other hand realizes the basic knowledge reasoning needs, which is of positive significance for realizing the transformation of map data services to knowledge services.

[0013] Furthermore, the extracted map knowledge is represented using a six-tuple model structure, which is as follows:

[0014] MK =<C,Ch,Re,Ru,F,I>

[0015] In the formula, MK represents the sum of map knowledge; C, Ch, Re, Ru, F, and I represent concepts, features, relations, rules, functions, and instances, respectively, as follows:

[0016] C = {Concepts, {C a}}

[0017] Ch = { <T,L,A)|T∈(T p T i ), L∈(L l L re )}

[0018] Re = {R} e (c i c j )|R e ∈(R el R et R es ), c i c j ∈(C∪I)}

[0019] Ru={R u :(s1, s2, s3, ..., s i , ..., s n-1 )→s n |s i s n ∈(C∪Re)}

[0020] F = {F: (c1, c2, ..., c...} n-1 →c n |c i ∈C}

[0021] I={I|Iisac i c i ∈C}

[0022] In the formula, Concepts represents geographical concept terms, {C a} represents the set of geometric concepts; T represents the temporal characteristics, including nodal time T.p and interval node T i L represents spatial features, including the entity's position coordinates L. l and spatial reference L re A represents the attribute feature; c i ,c j Representing concepts, instances, or attribute values, they exist as nodes in a knowledge graph, R e Represents a set of entity relations, defining the spatial relations R contained in the knowledge graph. el Time relationship R et and semantic relation R es It exists in the form of edges connecting nodes; s i ,…,s n R consists of concepts and relationships. u For a set of rules; c i ,…,c n F represents a concept node, I represents a function set, and F represents an instance.

[0023] Its beneficial effects are: it adopts a six-tuple model for map knowledge to realize a graphical representation of various types of map knowledge.

[0024] Furthermore, the rule-based knowledge is represented using Semantic Web rule language, which is represented in the form: Body(A1∧A2……A n →Head(B1∧B2……B n The above formula can be interpreted as: if each atom A in the preceding Body... i If all are true, then the inference result B of the consequent Head is... i It is also true, i = 1, 2, ..., n.

[0025] Its beneficial effects are: by using the semantic web rule language to represent rule-based knowledge, it is possible to realize the logical representation of various rule-based map knowledge.

[0026] Furthermore, the rule-based knowledge includes at least one of map projection models, cartographic generalization rules, spatial relationship reasoning rules, and symbol design principles. The spatial relationship reasoning rules include at least one of topological relationship reasoning, orientation relationship reasoning, distance relationship reasoning, and hybrid spatial relationship reasoning.

[0027] Furthermore, the decision-making knowledge is map function knowledge, used to represent the map function as a graph model. In the graph model, nodes represent variables in the map function, edges represent the weights of each variable, and the direction of the edges is determined by the interactions between the variables.

[0028] Furthermore, the relevant reasoning or decision support includes shortest path determination, buffer analysis, and visibility analysis.

[0029] Furthermore, the spatial relationship reasoning rules include:

[0030] Rule 1: When point a is inside line b and line b is inside surface c, then point a is also inside surface c.

[0031] Rule 2: When point b is west of point a and point c is west of point b, then point c is west of point a.

[0032] Rule 3: When points a and b are very close, and points b and c are very far apart, then points a and c are very far apart.

[0033] Rule 4: When point a is adjacent to point b, then line a is very close to point b. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the map knowledge classification of the present invention;

[0035] Figure 2 This is a schematic diagram of the map knowledge concept representation model of the present invention;

[0036] Figure 3 This is a schematic diagram illustrating the hierarchical relationship of the road transportation network concept;

[0037] Figure 4 This is a diagram representing the conceptual declaration of a first-level railway entity;

[0038] Figure 5 This is a schematic diagram of the graphical representation of the buffer of the present invention;

[0039] Figure 6 This is an example diagram of map knowledge query in this invention;

[0040] Figure 7(a) is an example diagram of the 13th level spatial relationship of the present invention;

[0041] Figure 7(b) is an example diagram of the 11th level spatial relationship of the present invention;

[0042] Figure 7(c) is a diagram showing the spatial relationship reasoning results based on SWRL rules of the present invention;

[0043] Figure 8(a) is a diagram of the pollution diffusion range at t=1 hour according to the present invention;

[0044] Figure 8(b) is a diagram of the pollution diffusion range at t=2 hours according to the present invention;

[0045] Figure 8(c) is a diagram of the buffer analysis results at t=1 hour according to the present invention;

[0046] Figure 8(d) shows the buffer analysis results of the present invention at t=2 hours. Detailed Implementation

[0047] This invention addresses the problem of insufficient systematic research on the modeling and representation of map knowledge, which hinders the full extraction of geographical knowledge contained within maps. It proposes a classification of map knowledge and designs a conceptual representation model for map knowledge. A six-tuple formal modeling method for map knowledge is proposed, constructing logical representation models for conceptual knowledge, rule-based knowledge, and decision-based knowledge, thus realizing graphical representations of various types of map knowledge. Final experimental verification shows that the map knowledge graph's concepts, features, and relationships can satisfy the description of map knowledge, clearly representing the relevant rules and application analysis of associated entities. This provides a foundation for comprehensive knowledge mining and integration of maps and has certain reference value for realizing the transformation from "data to information to knowledge."

[0048] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0049] Examples of map knowledge representation methods based on knowledge graphs:

[0050] An embodiment of a knowledge graph-based map knowledge representation method of the present invention is as follows: Figure 2 As shown, the process is as follows:

[0051] Step 1: Acquire map data, extract map knowledge, including conceptual knowledge, rule-based knowledge, and decision-based knowledge, and represent them using corresponding models.

[0052] 1. Classification of map knowledge.

[0053] Maps possess three fundamental characteristics: mathematical rules, symbol systems, and comprehensive rules. In the process of map production, content representation, and service application, maps generate diverse map knowledge. Based on the essential characteristics, evolutionary processes, and interactions of geographical entities on a map, map knowledge can be categorized into conceptual knowledge, rule-based knowledge, and decision-based knowledge, such as… Figure 1 As shown. Conceptual knowledge refers to common-sense knowledge, which is helpful for understanding professional knowledge related to cartography. It mainly refers to cartographic definitions / concepts and the classification of map elements, including map elements, rule concepts, and decision concepts. Map elements are mainly categories of map elements artificially set by humans to perceive geographic entities and spatial relationships on the Earth's surface. They include not only concepts such as topography / landforms, settlements, transportation, water systems, and vegetation, as well as corresponding examples, but also the relationships and characteristics between entities. Rule knowledge mainly refers to various models, rules, and principles formed in the cartographic process, including map projection models, cartographic generalization rules, spatial relationship reasoning rules, and symbol design principles. Decision knowledge is based on the analysis and application of geographic entities or phenomena to complete related reasoning and decision-making, manifested in quantitative analysis, visibility analysis, and buffer zone analysis.

[0054] 2. Map knowledge six-tuple representation.

[0055] The ontology approach is used to logically model map knowledge. By analyzing the classification, grading and expression of map knowledge, the five-tuple of the traditional ontology modeling metawords is expanded into a six-tuple. The ontology six-tuple model structure is shown in Equation (1):

[0056] MK=〈C, Ch, Re, Ru, F, I> (1)

[0057] Where MK represents the sum of map knowledge, and C, Ch, Re, Ru, F, and I represent concepts, features, relations, rules, functions, and instances, respectively. The explanations of each metaword are as follows:

[0058] 1) Concept. A concept is an abstract expression of an instance and belongs to the upper-level node of the instance. Based on the geometric characteristics and properties of geographic entities, map elements are classified according to the national standard "Classification and Code of Topographic Map Elements", such as settlements, transportation, boundary divisions, etc. Since geographic entities are often artificially divided into geometric objects, it is necessary to start from their geometric characteristics and express them through points, lines, surfaces and aggregated objects. Thus, the concept can be expressed as equation (2):

[0059] C = {Concepts, {C a}} (2)

[0060] Where Concepts represents geographical concepts, {C a} represents a set of geometric concepts, such as points, lines, and surfaces.

[0061] 2) Features. Features are the special symbols or marks that make an object identifiable. The features of map entities can be divided into three categories: temporal features (T), spatial features (L), and attribute features (A). Temporal and spatial features are prerequisites for the existence of an entity. Temporal features are further divided into point-time (T). p and interval time T i Spatial feature L mainly refers to the position coordinates L of the entity. l and spatial reference L re Attribute feature A is used to characterize the properties of the object itself, such as length and width. Therefore, the entity feature is represented by equation (3):

[0062] Ch = {<T,L,A> |T∈(T p T i ), L∈(L l L re )} (3)

[0063] 3) Relationships. Map entity relationships enable better association of map entities, thereby further expressing map knowledge. Temporal relationships are mainly used to describe various geographical phenomena with obvious time-varying characteristics. Expressed by time features, the same geographical entity node can be extended into sub-nodes to express temporal relationships. Spatial relationships mainly include topological relationships, directional relationships, and distance relationships. Topological relationships are expressed using a 9-intersection model based on dimension expansion, using the dimension of the intersection between the interior, boundary, and exterior of a geographical entity as the framework for describing topological relationships. Distance relationships are divided into qualitative and quantitative expressions. Qualitative expressions are mainly divided according to the needs of various progressive spatial organization, generally including vague descriptions such as very close, relatively close, near, medium, far, relatively far, and very far. Quantitative expressions accurately calculate the corresponding distance based on the target's position in two-dimensional space. Directional relationships are the relative positions of geographical entities in spatial latitude, mainly expressed qualitatively using directional models, including eight directions: East (E), South (S), West (W), and North (N). Semantic relationships are divided into three types: "concept-concept," "concept-instance," and "instance-instance," including parent-child relationships, similarity relationships, and equivalence relationships. Through the classification and analysis of entity relationships, the relationship can be expressed as equation (4):

[0064] R e ={R e (c i c j )|R e ∈(R el R et R es ), c i c j ∈(C∪I)} (4)

[0065] Among them, c i c j Representing concepts, instances, or attribute values, they exist as nodes in a knowledge graph; R e Represents a set of entity relations, defining the spatial relations R contained in the knowledge graph. el Time relationship R et and semantic relation R es It exists in the form of edges connecting nodes.

[0066] 4) Rules. Rule knowledge mainly refers to the dynamic knowledge required in the production, design and conversion of maps. It aims to generate new entities or relationships through rule-based reasoning of entities and their relationships. It is an explicit expression and extension of the implicit relationships of entities, and can be represented as Equation (5):

[0067] R u ={Ru :(s1, s2, s3, ..., s i , ..., s n-1 )→s n |s i s n ∈(C∪Re)} (5)

[0068] Among them, s i , ..., s n R consists of concepts and relationships. u It is a set of rules.

[0069] 5) Functions. A map itself is a data model. Functions are mainly used to constrain and expand entity relationships and attributes. For example, in the map production process, it is necessary to determine attribute characteristics such as scale and coordinate reference. The algorithms and models required for map analysis and application also belong to the category of functions. A function represents a situation where a certain concept is determined by multiple concepts. The purpose is to use the graph model to realize the representation of map functions, which can be expressed as equation (6):

[0070] F={F:(c1,c2,...,c n-1 →c n |c i ∈C} (6)

[0071] Among them, c i c n Let F represent the concept node and F represent the set of functions.

[0072] 6) Examples. Examples are concrete expressions of concepts, such as Science Avenue and Zhengkai Avenue in the transportation network system. Specifically, they can be represented as:

[0073] I={I|Iisac i c i ∈C} (7)

[0074] In the formula, I represents an instance, and c i Represents a concept node.

[0075] 3. Map element knowledge representation.

[0076] Map element knowledge includes entity concepts, instances, features, and relationships. It employs a semantic network representation, using the triple <node, relationship, node> as the smallest semantic unit. Through the induction and classification of map knowledge, combined with expert knowledge, domain concepts are extracted, defining classes, attributes, and their interrelationships. The definition of classes requires clarifying the hierarchical relationships of concepts based on the classification and grading of map knowledge. Some hierarchical relationships of concepts include... Figure 3As shown. Using the W3C-proposed OWL (Web Ontology Language) ontology language as the standard, and employing the ontology modeling tool Protégé, taking a first-level railway entity as an example, its declaration is as follows: Figure 4 As shown.

[0077] 4. Representation of map rule knowledge.

[0078] Map knowledge representation utilizes logical reasoning language to logically represent map rules, and is expressed using Semantic Web Rule Language (SWRL). The syntax of its rules is shown in equation (8):

[0079] Body(A1∧A2……A n →Head(B1∧B2……B n (8)

[0080] Specifically, if all atoms in the antecedent (Body) are true, then the thrust of the consequent (Head) will also be true.

[0081] The map knowledge system is vast, with numerous rules, but not all rules can be formally expressed, such as cartographic generalization rules and symbol design rules. Cartographic generalization consists of four operations: selection, simplification, generalization, and displacement. Each operation may generate different rules for different elements, and the combination and order of these rules often require expert experience to determine based on cartographic needs. Similarly, the design concepts of map symbol types and visual variables are mostly "tacit knowledge" of cartographic experts, and the formal representation of tacit knowledge is currently a challenge in map knowledge representation. This embodiment only models spatial relationship reasoning rules, which are relatively easy to formally express.

[0082] Spatial relation reasoning rules aim to deduce unknown spatial relations from known spatial relations when the map scale changes, and to merge the spatial relations of map entities at all scale levels. Spatial relation reasoning includes topological relation reasoning, orientation relation reasoning, distance relation reasoning, and mixed spatial relation reasoning. Some examples of SWRL rule representations are shown in Table 1.

[0083] Table 1 Example of SWRL rule base

[0084]

[0085] 5. Representation of map function knowledge.

[0086] The purpose of function knowledge representation is to represent map functions as graph models, and then use graph models to assist decision analysis. In a graph model, nodes represent variables in the function, edges represent the weights of each variable, and the direction of the edges is determined by the interactions between the variables. Taking buffer analysis in map mathematical model analysis as an example, the mathematical expression of the buffer of its spatial target set O is shown in (9):

[0087]

[0088] like Figure 5 As shown, in the graphical model of buffer analysis, the input spatial target O i coordinate set p i The minimum Euclidean distance d(p) is obtained. i O i Then, determine its relationship with the predetermined buffer distance R, and output the buffer analysis results; B i Indicates a space target; O i The buffer is a buffer of R; (x1,y1), (x2,y2),......(x i ,y i ) represents a series of coordinate values.

[0089] Step two, based on the extracted map feature knowledge, including the relationships between geographic entities (such as... Figure 2 Region B in the text) and entity features (such as Figure 2 Region C in the graph), generating a map knowledge network graph based on the knowledge graph (e.g., region C). Figure 2 (Area A in the text).

[0090] Step 3, utilize rule-based knowledge (such as...) Figure 2 Knowledge reasoning can be performed on the D region of the map to generate new map knowledge, such as new entities, relationships between entities, and features, thereby completing the map knowledge network graph.

[0091] Step four: Based on the relevant mathematical models in decision-making knowledge, conduct various qualitative and quantitative analyses of the map elements in the map knowledge network diagram (e.g., ...). Figure 2 In region E of the map, knowledge graph computing is used to analyze the spatiotemporal distribution characteristics and potential patterns of map objects, thereby achieving the purpose of related reasoning and decision support.

[0092] The following is a specific experimental case analysis.

[0093] 1. Experimental Data and Processing. The experimental data in this paper includes: 1) Linear roads, areal administrative regions, and point service areas in parts of Zhengzhou City from OSM level 13 and 11 tiles. ArcMap software was used to vectorize, fill, and trim the map data, and the uniqueness of entity names was checked; 2) The encyclopedia knowledge base CN-DBpedia, containing 80 million entities and 120 million relations. The map knowledge obtained through knowledge modeling and extraction was stored and visualized using the Neo4j graph database.

[0094] 2. Experimental evaluation and analysis.

[0095] 1) Map Feature Knowledge Query. Using D2R tools, based on relational table mapping of map data, map feature entities and their attributes are extracted and stored in the Neo4j database. Entity features are then queried and retrieved using the Cypher language. For example... Figure 6 As shown, the concept node "City" extends from top to bottom, with the "Zhengzhou" node as an example. It is further divided into two sub-nodes based on time characteristics, representing the attribute characteristics affected by time changes, such as Zhengzhou's per capita GDP in 2015 and 2019. Entity spatial characteristics and attribute characteristics are represented by nodes such as coordinates and reference frames. For example, the coordinates of "Zhengzhou" are "(116.4, 39.9)" and the area of ​​"High-tech Zone" is "99 km²". 2 The relationships and attribute types are represented by edges, such as the adjacency relationship between "Zhengzhou University" and "Science Avenue" and the inclusion relationship between "Zhengzhou" and "High-tech Zone".

[0096] 2) Rule-based knowledge reasoning. Knowledge reasoning based on SWRL rules is mainly implemented through the Jess inference engine. The Jess rule base built by SWRL rules is combined with the Jess fact base transformed from the ontology model, and finally the map rule reasoning is completed by the Jess inference engine. Taking the reasoning of spatial relationship rules as an example, the spatial relationship is as follows: First, the spatial relationship between the three points "Henan University of Technology", "Zhengzhou University of Light Industry" and "Zhengzhou University" in the 13th and 11th level maps are extracted respectively, as shown in Figure 7(a) and Figure 7(b), resulting in (Zhengzhou University of Light Industry, Westof, Henan University of Technology), (Henan University of Technology, Westof, Zhengzhou University), and (Zhongyuan University of Technology, Eastof, Zhengzhou University); then, based on the SWRL rule [Point(?a)^Point(?b)^Point(?c)^westof(?a,?b)^westof(?b,?c)->westof(?a,?c)], the spatial relationship between the three points is reasoned and merged, as shown in Figure 7(c), where the thickest line segment represents the merged relationship and the second thickest line segment represents the reasoned relationship.

[0097] 3) Functional Knowledge Graph Model. Functional knowledge exists as nodes in the graph database. Entities extend their attributes and relationships by associating functional knowledge nodes, thereby achieving the purpose of assisting decision analysis. Taking the air pollution from Xinwang Chemical Plant as an example, the diffusion range of air pollution in two time periods, 1 hour and 2 hours after the pollution outbreak, is analyzed, and the spatial characteristics and relationships of the entities affected within the range are extracted, as shown in Figures 8(a) and 8(b). The nodes covered by the gray area represent the polluted villages in the two time periods. Then, the pollution areas at different time periods centered on the chemical plant node are constructed by associating the coordinates, area, and other regional characteristics of the entities (the areas in Figures 8(c) and 8(d)). The results show that the graph model can more intuitively and clearly understand the regional impact caused by the changes in the temporal characteristics of entities, and can store the analysis results in the form of knowledge.

[0098] In summary, this invention, starting from the needs of knowledge services and machine intelligence, elucidates the importance of map knowledge modeling. First, it proposes a classification of map knowledge based on the essential characteristics and inherent laws of geographic entities and designs a conceptual model for spatial cognitive reasoning. Second, it uses an ontology approach to design map knowledge matrix modeling, proposing a six-tuple model containing concepts, features, relationships, functions, rules, and instances, and formally expressing it using a modeling language. Finally, it conducts experimental verification based on OSM map data, completing the extraction of element knowledge and using SWRL rules for dynamic knowledge reasoning and analysis. The results show that the knowledge graph-based map knowledge representation method proposed in this invention can meet the basic description of map knowledge in terms of concepts, features, and their semantic relationships, comprehensively displaying the knowledge contained in maps. Furthermore, the rule base built through this model can meet basic knowledge reasoning needs, demonstrating feasibility and reference value.

Claims

1. A map knowledge representation method based on knowledge graphs, characterized in that, Includes the following steps: 1) Acquire map data and extract map knowledge from it; wherein, the map knowledge includes conceptual knowledge, rule-based knowledge, and decision-based knowledge; conceptual knowledge is common knowledge in the field of cartography and includes map element knowledge, which includes entities, entity relationships, and entity characteristics; rule-based knowledge is various models, rules, and principles formed in the process of mapmaking, including map projection models, cartographic generalization rules, spatial relationship reasoning rules, and symbol design principles; decision-based knowledge is knowledge based on the analysis and application of geographical entities or geographical phenomena to complete relevant reasoning and decision-making, including shortest path determination, buffer analysis, and visibility analysis; wherein, the extracted map knowledge is represented by a six-tuple model structure containing concepts, characteristics, relationships, rules, functions, and instances; 2) Based on the map element knowledge, rule-based knowledge, and decision-based knowledge, a map knowledge network graph is generated using a knowledge graph. During the generation of the map knowledge network graph, map functions in the decision-based knowledge are represented as graph models to aid decision analysis. Function knowledge exists as nodes in the graph database. Entities extend their attributes and relationships by associating with function knowledge nodes. Nodes in the graph model represent variables in the function, edges represent the weights of each variable, and the direction of the edges is determined by the interactions between variables. In the graph model corresponding to buffer analysis, nodes include spatial target nodes O. i , coordinate set node p i The minimum Euclidean distance node d(p) i O i ), buffer distance node R, buffer result node B i The edges include those from O i and p i Pointing to d(p) i O i The edge of p i An edge pointing to R, and an edge from R pointing to d(p) i O i The edge of d(p) i O i ) points to B i The edge; 3) Based on rule-based knowledge, perform knowledge reasoning to complete the generated map knowledge network graph, thereby generating new entities and / or new entity relationships, and thus obtaining the completed map knowledge network graph.

2. The map knowledge representation method based on knowledge graphs according to claim 1, characterized in that, The method also includes the following steps: 4) Based on decision-making knowledge, analyze the elements in the map using the completed map knowledge network diagram to make relevant inferences or assist in decision-making.

3. The map knowledge representation method based on knowledge graphs according to claim 1, characterized in that, Concepts, features, relationships, rules, functions, and instances are respectively represented as: In the formula, These respectively represent concepts, characteristics, relationships, rules, functions, and instances; Geographical terms, Represents a set of geometric concepts; Representing time characteristics, including node time and interval nodes , Representing spatial features, including the location coordinates of entities. and spatial reference , Represents attribute characteristics; Representing concepts, instances, or attribute values, they exist as nodes in a knowledge graph. Represents a set of entity relations, defining the spatial relations contained in the knowledge graph. Due to time constraints and semantic relations It exists in the form of edges connecting nodes; It consists of concepts and relationships; Represents a concept node.

4. The map knowledge representation method based on knowledge graphs according to claim 1, characterized in that, The rule-based knowledge is represented using the Semantic Web rule language, which is expressed in the following form: The above formula can be interpreted as: if the antecedent each atom If all are true, then the consequent The result of reasoning It is also true. .

5. The knowledge graph-based map knowledge representation method according to claim 1 or 4, characterized in that, The spatial relationship reasoning rules include at least one of the following: topological relationship reasoning, orientation relationship reasoning, distance relationship reasoning, and mixed spatial relationship reasoning.

6. The knowledge graph-based map knowledge representation method according to claim 3, characterized in that, The set of geometric concepts includes points, lines, and surfaces.

7. The knowledge graph-based map knowledge representation method according to claim 5, characterized in that, The spatial relationship reasoning rules include: Rule 1: When point a is inside line b and line b is inside surface c, then point a is also inside surface c. Rule 2: When point b is west of point a and point c is west of point b, then point c is west of point a. Rule 3: When points a and b are very close, and points b and c are very far apart, then points a and c are very far apart. Rule 4: When point a is adjacent to point b, then line a is very close to point b.

Images