A port cluster navigation channel guidance and assistance navigation fusion decision graph construction method

Abstract

The application discloses a kind of port group navigation channel guide navigation fusion decision graph construction methods, it is related to port channel intelligentization and knowledge graph technical field, comprising: obtaining multi-source heterogeneous environment data, including image and structured information;Image information is divided into time slot and element area segmentation, and directional class environmental elements are extracted;Multiple source data is uniformly reconstructed into structured data set organized according to time slot;Perform missing value processing and standardization, generate fusion environment data;Guide navigation knowledge graph mode, relationship template and state determination function are constructed;Finally, the port group navigation channel guide navigation fusion decision graph is generated by instantiation.The application can automatically extract image environment information, realize the unified time alignment and fusion expression of multi-source heterogeneous wind, wave, tide and other elements, construct computable, updateable, reasoning decision graph, provide efficient knowledge organization and decision support capability for port group navigation channel digital twin platform.

Description

Technical Field

[0001] This invention relates to the field of port and waterway intelligence and knowledge graph technology, and in particular to a method for constructing a navigation-aided decision graph for port clusters. Background Technology

[0002] In the application of intelligent and dynamic real-time digital twins for waterway facilities in large coastal port clusters, navigational aids for ships entering and leaving ports typically rely on a variety of environmental information, including wind speed, gusts, wind direction, wave height, wave period, wave direction, tide height, tide type, air pressure, and temperature. This information often originates from different types of data sources, such as marine meteorological forecasting platforms, on-site monitoring systems, historical statistical data, manual duty records, and external interface data within the digital twin scenario.

[0003] In existing technologies, the processing of this type of navigation aid environmental information mainly has the following methods and shortcomings: First, there's the manual browsing and transcription method. On-duty or dispatch personnel read wind, wave, and tide information from the marine weather forecast page and then manually compile it into tables or verbally analyze the results. This method is inefficient, prone to omissions and errors, and unsuitable for supporting high-frequency dynamic updates.

[0004] Second, there's the issue of structured processing methods oriented towards single data formats. Some systems can only handle tabular or database-type data, making it difficult to directly process the graphical environmental information contained in visual forecast pages, such as directional arrows, time-slotted layouts, and tide icons. This results in a large amount of environmental information that could be directly used for navigation assistance not being automatically incorporated into the digital twin system.

[0005] Third, the analysis approach is geared towards single elements. Most existing methods display or analyze elements such as wind, waves, and tides separately, lacking a mechanism for fusing multi-source heterogeneous environmental elements under a unified time reference, making it difficult to form a unified knowledge representation suitable for navigation-aided scenarios.

[0006] Fourth, the system is display-oriented rather than knowledge-organization-oriented. Existing systems often use the extracted results only for charts or lists, lacking the ability to organize "time nodes - environmental status - navigation conditions - navigation aids actions" into a computable, scalable, and updatable graph structure, which is not conducive to subsequent reasoning, retrieval, alarm, and decision reuse.

[0007] Fifth, there is insufficient handling of missing data, scale differences, and update mechanisms. Environmental data often suffers from issues such as missing measurements, inconsistent formats, inconsistent units, and different time granularities. Without a unified mechanism for data cleaning, missing data completion, standardization, and updating, even if data is acquired, it is difficult to stably support the construction of navigation-aided knowledge graphs.

[0008] In summary, existing technologies have at least the following drawbacks: weak ability to automatically extract image-based environmental information; difficulty in unifying multi-source heterogeneous wind, wave, and tide data into the same time slot and structure; lack of knowledge representation models for navigation aid scenarios; lack of continuously updated navigation aid decision map construction methods; and inability to directly provide highly reusable and highly interpretable underlying knowledge organization capabilities for port cluster waterway digital twin scenarios. Summary of the Invention

[0009] The purpose of this invention is to provide a method for constructing a fusion decision graph for navigation aids in port clusters, which solves the problems of weak automatic extraction capability of image-based environmental information, difficulty in fusion of multi-source heterogeneous data, and lack of surface-based navigation aid knowledge graph representation in the existing technology.

[0010] To achieve the above objectives, this invention provides a method for constructing a fusion decision map of navigation aids for port clusters, comprising the following steps: Step S1: Obtain the original data set of multi-source heterogeneous environment in the navigation aid scenario of port clusters. Each data item Represented as: ,in: Indicates the site identifier; Indicates time information; Indicates the type of environmental element; Indicates decision-making; Indicates the corresponding value or state; Indicate the data source; The data type can be represented as image data, structured data, or text data; the raw data from a multi-source heterogeneous environment includes at least image-based environmental information and structured environmental data. Step S2: Divide the imaged environmental information into time slots and feature regions, and divide the time axis layout region and environmental feature region. The environmental feature region includes directional environmental feature region, numerical environmental feature region and state environmental feature region. Identify and extract the directional environmental feature, numerical environmental feature region and state environmental feature region under each time slot. Step S3: Perform unified structural reconstruction on the structured environmental data and the directional environmental elements extracted in step S2 to form an environmental element dataset organized by date and time slot; Step S4: Perform missing value processing and data standardization on the environmental element dataset to generate fused environmental data; Step S5: Construct the navigation-aided knowledge graph schema, relation template, and state determination function: Define the entity type set. ,in, Represents a time entity. Represents environmental element entities, Represents an entity representing the state of the environment. Represents the navigation aid object entity. Represent the decision entity; define the set of relation types. The logical association between nodes and edges in a knowledge graph is explicitly defined as a knowledge graph relation triple structure. This relation triple structure is used to express the semantic links between entities in the schema layer. Relation templates are defined based on this relation triple structure. ,in: Indicates that something is being observed. Indicates impact, Indicates triggering, Indicates that it acts on, Indicates the data source. Indicates an update. Indicates indirect relationship, Indicate decision-making needs, Indicates the type of the head entity in the relation triple (such as an environment element entity or an environment state entity). Indicates the type of tail entity in the relation triple (such as time entity, environment state entity, or decision entity). This represents the specific instance relationship edge type between the header entity type and the tail entity type; the generated relationships include: This indicates that environmental elements were observed and recorded at a specific point in time. This indicates that current environmental factors have a direct impact on a certain early warning or safety environment status. This indicates that a certain environmental condition has reached the point where the corresponding navigation aid decision is triggered. This refers to the physical effects of environmental factors such as wind and waves on objects such as waterways or berths. This indicates the source relationship of data nodes from a specific information system or platform. This represents the incremental update relationship of the graph, where new attributes or nodes replace old data when new data is acquired. This refers to the indirect relationship between environmental elements and their state or other entities; Specific environmental conditions trigger a decision requiring navigation assistance procedures; define a state determination function. ,in: Indicates the first Class environment state; This represents the standardized environmental elements; This represents the state determination function; Step S6: Based on the graph pattern defined in step S5, instantiate the fusion environment data generated in step S4 into a node set. and relation edge set Generate a navigation-aided decision-making map for port clusters.

[0011] Preferably, step S2 specifically includes: dividing the entire image into multiple time slot units horizontally according to the forecast page layout template; within each time slot unit, the wind direction arrow area and wave direction arrow area are used as directional environmental element areas, the numerical area is used as a numerical environmental element area, and the icon area is used as a status environmental element area; the extracted directional environmental element area images are sequentially grayscaled and thresholded to extract the foreground pixel coordinate set; the principal component analysis method is used to calculate the first principal direction vector of the foreground pixels; the arctangent function is used to calculate the angle of the principal direction vector, and the angle is converted into a standard azimuth angle as the directional element value of the time slot.

[0012] Preferably, the specific method for forming the environmental element dataset in step S3 is as follows: the wide table structure of the original data is reconstructed into a long table structure containing fields such as "site identifier, date, time slot, environmental element name, environmental element value, data source identifier, and data type identifier"; and environmental elements from different data sources are aligned according to date and time slot to form a unified sample of multiple elements under each time slot.

[0013] Preferably, the missing value handling in step S4 is as follows: for numerical environmental elements, linear interpolation is performed on the time series according to the element category. For directional elements, nearest neighbor direction completion, adjacent time slot direction smoothing, and other methods can also be used to supplement them.

[0014] Preferably, the relation weight function in step S5 is defined as: ; in, Represents node entities and The strength of the relationship between them; This is the weight calculation function; These represent the environmental elements or state parameters corresponding to the head and tail entities associated with the relationship, respectively; t is the time variable.

[0015] Preferably, the navigation aid decision in step S5 is determined by the following function: ; Where D represents the navigation aid decision result; Represents environmental state variables; This represents the decision function.

[0016] Preferably, the instantiation and generation process of the map in step S6 includes: S61. Instantiation is defined as the process of mapping fused data to a specific graph structure, expressed as: ; in, The set of nodes that generate instances; This represents the set of relational edges that generate instances; S62. Define a mapping function with the following expression: ; in, For functional relation symbols, This represents the mapping relationship between nodes and relation edges; This represents the set of nodes generated after mapping. This represents the edge set generated after mapping; S63. Generate nodes and relation edges, specifically: For each data record Extract timestamp as Create the corresponding time node instance The extracted element values ​​are: Create corresponding environment element node instances According to the decision function Determine and generate state node instances ; Match the corresponding waterway or ship to generate navigation aid object node instances. ; Generate decision node instances based on specific business rules According to the relation template , build connections and observation relationship edge Influence on the relationship and triggering decision relationship edge Entity diagram structure; S64. Assign values ​​to the generated node attributes, including: value = standardized environmental element value; timestamp = time slot; source = data source; and output the instantiated result graph with time, environment, decision entity and related edges.

[0017] Preferably, the method also includes step S7: dynamic updating and version management of the map. When new raw data is obtained, the existing map is updated incrementally by matching the site, date and time slot. The incremental update includes: creating new time nodes for the new time slots, adding new attribute values ​​to existing environmental element entities, and recalculating and updating the weights of the relationship edges according to the relationship weight function.

[0018] Preferably, the fused decision graph is exported as one or more of the following: GraphML file, JSON file, graph database record, or graph interface data for the digital twin platform, to provide decision support for the port cluster waterway digital twin platform.

[0019] Therefore, the present invention employs the above-mentioned method for constructing a port cluster navigation aid fusion decision map, which has the following beneficial effects: (1) By dividing time slots, segmenting regions and using principal component analysis, it is possible to automatically identify and extract directional elements such as wind direction and wave direction from the images on the weather forecast page, overcoming the shortcomings of low efficiency and easy errors in manual browsing.

[0020] (2) The multi-source environmental information, including images, structures, and texts, is uniformly reconstructed into a long table structure with fields such as station, date, and time slot, so as to realize the fusion expression of elements such as wind, waves, and tides under a unified spatiotemporal benchmark.

[0021] (3) Time entity, environmental element entity, state entity, navigation aid object entity and decision entity and their relationship templates are defined, and a computable and reasonable knowledge expression system is established.

[0022] (4) Linear interpolation is used for numerical features and nearest neighbor completion is used for directional features to handle missing data, and a standardization mechanism is combined to ensure data quality and map stability.

[0023] (5) Based on the site, date and time slot matching mechanism, it can incrementally update new data, automatically add nodes, update attributes and relationship weights, and support graph version management to meet the continuous operation requirements of the digital twin platform.

[0024] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0025] Figure 1 This is an overall flowchart of the method for constructing a fusion decision map of navigation aids for port clusters according to the present invention; Figure 2 This is a schematic diagram of the page layout for graphical environmental information according to an embodiment of the present invention; Figure 3 This is a flowchart of image time slot segmentation and region segmentation according to an embodiment of the present invention; Figure 4 This is a flowchart illustrating the extraction process of directional environmental elements according to an embodiment of the present invention. Figure 5 This is a flowchart illustrating the data fusion reconstruction process according to an embodiment of the present invention. Figure 6 This is a schematic diagram of the navigation-aided navigation fusion decision graph structure according to an embodiment of the present invention; Figure 7 This is a flowchart illustrating the dynamic update process of the map according to an embodiment of the present invention; Figure 8 This is a schematic diagram of the map based on the Tianjin Xingang site, according to an embodiment of the present invention. Detailed Implementation

[0026] The following detailed description of embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0027] Please see Figures 1-8 A method for constructing a fusion decision map of navigation aids for port clusters includes the following steps: Step S1: Obtain the original data set of multi-source heterogeneous environment in the navigation aid scenario of port clusters. Each data item Represented as: ,in: Indicates the site identifier; Indicates time information (date and time slot); Indicates the type of environmental element; Indicates decision-making; Indicates the corresponding value or state; Indicate the data source; The data type is indicated (covering image data, structured data, or text data; for text-based record data, a rule engine is used for parsing and extraction); the raw data includes at least image-based environmental information and structured environmental data; The data types of the raw data in the multi-source heterogeneous environment include at least one or more of the following: Image-based environmental information from external marine meteorological forecasting sources; Structured environmental data from marine meteorological platforms, port operational systems, or on-site monitoring systems; Textualized environment information from duty logs, business systems, or interface systems.

[0028] Environmental factors include, but are not limited to: wind speed, gusts, wind direction, wave height, wave period, wave direction, tide height, tide type, air pressure, temperature, precipitation, and cloud cover.

[0029] The raw data is collected in order of site, date and time, and an index of the raw data is created.

[0030] Step S2: Divide the imaged environmental information into time slots and segment feature regions, such as... Figure 2 As shown, directional environmental elements are identified and extracted for each time slot; specifically, these include, for example... Figure 3As shown: Based on the forecast page layout template, the page layout is identified, and the entire image is divided horizontally into multiple time slot units. Within each time slot unit, the time axis layout area, directional environmental element areas such as wind direction arrows and wave direction arrows, numerical environmental element areas such as numerical values, and status environmental element areas such as icon areas are determined according to predefined coordinate ranges. In each time slot column, the vertical coordinate range corresponding to different environmental elements is determined according to the predefined layout template, thus forming: wind direction extraction area, wave direction extraction area, numerical value extraction area, and tide pattern extraction area. Through the above method, the original page image is converted into an image processing object that is "segmented by time slot and located by element area". The directional environmental elements of the located area image are sequentially grayscaled and thresholded to extract their azimuth angles; for example... Figure 4 As shown, specifically: Convert the corresponding region image to a grayscale image; Thresholding is applied to the grayscale image to obtain a binary image; Extract the set of foreground pixel coordinates from the binary image; The main direction of the arrow was extracted based on principal component analysis. Calculate the principal axis heading angle and convert it to the heading azimuth angle; Output the direction value corresponding to each time slot.

[0031] The calculation process for the principal axis direction angle is as follows: Principal component analysis is performed on the foreground pixels to obtain the first principal direction vector; The angle of the principal direction vector is calculated using the arctangent function; Convert image angles into a unified azimuth angle representation.

[0032] If enough valid arrow foreground pixels are not extracted within a certain time slot, the corresponding direction value of that time slot is marked as a missing value.

[0033] Alternatives to orientation extraction include, but are not limited to: orientation recognition based on contour detection; orientation recognition based on Hough transform; orientation recognition based on template matching; and orientation classification based on convolutional neural networks.

[0034] Step S3: Perform unified structural reconstruction of the structured environmental data and the directional environmental elements extracted in step S2 to form an environmental element dataset organized by date and time slot; such as Figure 5As shown, the specific method is as follows: the wide table structure of the original data is reconstructed into a long table structure containing fields such as "site identifier, date, time slot, environmental element name, environmental element value, data source identifier, and data type identifier"; and environmental elements from different data sources are aligned according to date and time slot to form a unified sample (fusion data table) of multiple elements under each time slot. In other words, information such as wind and wave data that were originally scattered in different sources, formats, and locations is uniformly mapped to the same spatiotemporal coordinate system.

[0035] Step S4: Perform missing value processing and data standardization on the environmental element dataset to generate fused environmental data. Missing value processing involves: for numerical environmental elements, performing linear interpolation over time series according to element category; for directional elements, methods such as nearest neighbor direction completion and adjacent time slot direction smoothing can also be used. Missing value processing is not limited to linear interpolation; spline interpolation and machine learning interpolation can also be used. Standardization is not limited to Z-score; min-max normalization can also be used.

[0036] Step S5: Construct the navigation-aided knowledge graph pattern, relation template, and state determination function: Define entity type collection ,in, Represents a time entity. Represents environmental element entities, Represents an entity representing the state of the environment. Represents the navigation aid object entity. Represents the decision-making entity; where: Time Entity : Indicates date, time slot, or time period; Environmental element entities : Indicates wind speed, gusts, wind direction, wave height, wave period, wave direction, tide height, tide type, air pressure, temperature, etc.; Environmental state entity This indicates conditions such as high winds, low tides, large waves, and abnormal temperatures. Navigation aid object entity : Indicates waterway sections, stations, and ship navigation scenarios; Decision-making entity This indicates operational actions such as suggesting entry into port, suggesting departure from port, delaying navigation, or strengthening monitoring.

[0037] Define a set of relation types The logical association between nodes and edges in a knowledge graph is explicitly defined as a knowledge graph relation triple structure. This relation triple structure is used to express the semantic links between entities in the schema layer. Relation templates are defined based on this relation triple structure. ,in: Indicates that something is being observed. Indicates impact, Indicates triggering, Indicates that it acts on, Indicates the data source. Indicates an update. Indicates indirect relationship, Indicate decision-making needs, Indicates the type of the head entity in the relation triple (such as an environment element entity or an environment state entity). Indicates the type of tail entity in the relation triple (such as time entity, environment state entity, or decision entity). This represents the specific instance relationship edge type between the header entity type and the tail entity type; the generated relationships include: This indicates that environmental elements were observed and recorded at a specific point in time. This indicates that current environmental factors have a direct impact on a certain early warning or safety environment status. This indicates that a certain environmental condition has reached the point where the corresponding navigation aid decision is triggered. This refers to the physical effects of environmental factors such as wind and waves on objects such as waterways or berths. This indicates the source relationship of data nodes from a specific information system or platform. This indicates the incremental update relationship of the graph, where new attributes or nodes replace old data when new data is acquired. This refers to the indirect relationship between environmental elements and their state or other entities; Specific environmental conditions trigger a decision requiring navigation assistance procedures; define a state determination function. ,in: Indicates the first Class environment state; This represents the standardized environmental elements; This represents the state determination function; It should be noted that the graph relationships are not limited to a fixed set of relationships; more relationship types can be expanded based on the navigation aid service template, such as... Figure 6 As shown.

[0038] Define state determination function ,in: Indicates the first Class environment state; This represents the standardized environmental elements; Represents the state determination function; where: The relation weight function is defined as follows: ; in, Represents node entities and The strength of the relationship between them (i.e., edge weight); This is the weight calculation function; These represent the environmental elements or state parameters corresponding to the head and tail entities associated with the relationship, respectively; t is the time variable.

[0039] General functions In practical port navigation aid scenarios, it is specifically instantiated as a linear combination of multiple environmental factors. That is, using wind speed U, wave height H, and tide height T extracted within the current time period as input variables, and then... The formula calculates the specific edge weights, thus realizing the transition from "abstract function" to "concrete multi-parameter calculation". This expression clarifies the calculation logic that the weights of the relationships between nodes are jointly determined by the values ​​of specific environmental factors at the corresponding time.

[0040] Navigation-aided decisions are determined by the following function: ; Where D represents the navigation aid decision result; Represents environmental state variables; This represents the decision function.

[0041] This embodiment uses wind speed and wave height as examples to illustrate the determination process: ; ; Where U represents wind speed; H represents wave height; , For the corresponding threshold; Indicates wind speed and environmental conditions; Indicates wave height and environmental conditions.

[0042] ; Furthermore, the edge weights can be calculated using the following function: ; Where U represents wind speed; H represents wave height; and T represents tide height; , , These are the corresponding weighting coefficients.

[0043] Step S6: Based on the graph pattern defined in step S5, instantiate the fusion environment data generated in step S4 into a node set. and relation edge set Generate a fusion decision map of navigation aids for port clusters; the instantiation and generation process of the map includes: S61. Instantiation is defined as the process of mapping fused data to a specific graph structure, expressed as: ; in, The set of nodes that generate instances; This represents the set of relational edges that generate instances; S62. Define a mapping function with the following expression: ; in, For functional relation symbols, This represents the mapping relationship between specific nodes and relational edges; This represents the specific set of node instances generated after mapping (including environmental element node instances; environmental state node instances; navigation aid object node instances; decision node instances, etc.). This represents the set of instance edges generated after mapping; S63. Generate nodes and relation edges, specifically: for each data record Extract timestamp as Create the corresponding time node instance The extracted element values ​​are: Create corresponding environment element node instances According to the decision function Determine and generate state node instances ; Match the corresponding waterway or ship to generate navigation aid object node instances. ; Generate decision node instances based on specific business rules (or decision candidate node); based on the relationship template , build connections and observation relationship edge Influence on the relationship and triggering decision relationship edge Entity graph structure.

[0044] S64. Assign values ​​to the generated node attributes, including: value = standardized environmental element value; timestamp = time slot; source = data source; and output the instantiated result graph with time, environment, decision entity and related edges.

[0045] When generating the map, at least the following map instance information should be generated for each time slot: A point in time; Multiple environmental element nodes; The relationship between each environmental element node and a time node; The influence relationship edges between various environmental element nodes; Environment status node; Decision node or decision candidate node.

[0046] The graph can be exported in one or more of the following formats: Graph Markup Language (GraphML) files; JavaScript Object Notation (JSON) files; Graph database storage format; Graph interface data format for digital twin platforms.

[0047] Step S7, Dynamic Map Update and Version Management: When new raw data is obtained, incremental updates are performed on existing maps by matching sites, dates, and time slots, such as... Figure 7 As shown, the update methods include, but are not limited to: Add new environmental element attributes to existing time points; Complete the missing nodes; Add new attribute values ​​to existing environmental element entities; Overwrite or retain historical version node attributes; Recalculate the weights or states of the relation edges; Number and archive the output versions of the atlas.

[0048] Through the above mechanism, the navigation-aided decision map constructed by this invention can be applied to continuous updates and multi-round reuse in the dynamic real-time digital twin scenario of port cluster waterways.

[0049] The present invention also includes the following systems: Raw data acquisition module: used to collect external marine meteorological forecast information, tabular data, and text data; Image preprocessing and time slot segmentation module: used for layout recognition, time slot segmentation, and feature region segmentation of the image-based forecast page; Environmental Element Extraction Module: Used to extract directional, numerical, and state-related environmental elements. For numerical environmental elements such as wind speed, gusts, wave height, wave period, tide height, air pressure, and temperature, as well as state-related elements such as tide pattern and weather conditions, extraction can be performed using one or more of the following methods: (1) Directly read structured data files; (2) Perform character recognition on numerical regions in the image; (3) Perform template recognition or classification recognition on the icon areas in the image; (4) Perform rule parsing on textual records.

[0050] Numerical data can be read from table data synchronized with the image; numerical and status information can also be obtained directly from the image recognition module.

[0051] Fusion processing module: used to perform data reconstruction, time alignment, missing value handling, and standardization; Graph Modeling Module: Used to generate entity, attribute, and relation templates and complete graph instantiation; The graph update and storage module is used to output graph files, graph database records, and their version management results.

[0052] Example This embodiment further illustrates the invention based on the Tianjin Xingang site: 1) Acquisition of raw data: Obtain navigational aids information for Tianjin New Port channel stations over several consecutive dates.

[0053] The raw data includes: The visual forecast page includes wind direction, wave direction, weather conditions, etc. for multiple time slots; Structured numerical data synchronized with the images include wind speed, gusts, wave height, wave period, tidal height, air pressure, and temperature.

[0054] The above raw data is archived by date.

[0055] 2) Image segmentation and orientation extraction: The forecast image is used as input, and the entire image is divided into multiple time slot columns horizontally.

[0056] In this embodiment, the forecast image is divided horizontally in 3-hour increments, and an image containing a 2-day environmental forecast is divided into 16 time slot columns (e.g., ...). Figure 2 As shown), this segmentation logic and subsequent steps and Figure 8 ( Figure 8 (Only a portion of the image is shown) The layout structure remains consistent.

[0057] Predefine the longitudinal range of the wind direction region and the longitudinal range of the wave direction region in the image. For the direction region in each time slot, perform the following processing: Grayscale conversion; Binarization; Identify foreground pixels; When the number of valid pixels exceeds a preset threshold, principal component analysis is performed. Take the first principal direction vector; Convert the direction vector to an azimuth angle; Output the wind direction or wave direction value for this time slot.

[0058] If the number of valid pixels is lower than the threshold, the direction value of that time slot is recorded as missing.

[0059] In this embodiment, this step can process multiple date images in batches and output a direction extraction result table.

[0060] 3) Preparation for structured data fusion: Read the structured numerical environment data and perform the following processing: The original wide table was restructured into a long table structure containing fields such as "site identifier, date, time slot, environmental element name, environmental element value, data source identifier, and data type identifier". Extract the date from the time field; Establish time slot numbers for multiple time values ​​under the same date; Convert numeric fields to numeric format.

[0061] 4) Handling missing values: For numerical elements such as wind speed, gusts, air pressure, temperature, wave height, wave period, and tide height, linear interpolation is performed on the time series according to the element category to obtain continuous numerical series.

[0062] For directional elements, methods such as nearest neighbor direction completion and adjacent time slot direction smoothing can also be used for supplementation.

[0063] 5) Standardization: Z-score standardization is performed on numerical elements such as wind speed, gusts, air pressure, temperature, wave height, wave period, and tidal height. The standardized environmental elements are used for subsequent unified comparisons, state discrimination, and map attribute representation.

[0064] 6) Generation of merged data tables: Various environmental elements are aligned according to date and time slots to form a unified, integrated data table containing multiple types of environmental elements under each time slot, such as... Figure 8 As shown.

[0065] This table contains the following corresponding results: date, time slot, wind direction, wind speed, gusts, cloud cover, temperature, air pressure, wave direction, wave height, wave period, tide type, and tide height (tide level).

[0066] To clearly demonstrate Figure 8 The corresponding instantiation results and steps are described in Table 1 below, which illustrates the actual data fusion of the Tianjin Xingang (Tianjin Port Main Channel) station at 20:00 on March 23, 2026 (corresponding to the 17:00-20:00 time period): Table 1: Results of Tianjin Port Site Data Integration Example

[0067] 7) Graph modeling and instantiation results: Based on the navigation aid service requirements, the following entities are constructed: Time Node; Wind Speed ​​Node; Wave Height Node; Tide Height Node; Environmental Status Node; Navigation Aid Action Node; and the following relationships are established: Establish an "observation-time" relationship between environmental element nodes and time nodes; The wind speed node and wave height node establish an "influence" relationship with the tide height node or channel status node; The tidal height node or environmental status node establishes a "need to make a decision" relationship with the navigation aid action node.

[0068] In this embodiment, the graph is represented as a directed graph and exported as a GraphML file and a JSON file. Specifically: Combining the fusion values ​​in Table 1, in generating... Figure 8 When the navigation aid map is displayed, the system accurately creates the corresponding time entity node in the map library. ("2026-03-23 ​​17:00-20:00"), and corresponding environmental element nodes were created respectively. The following are entity nodes: wind speed with an attribute value of "11kts", wave height with an attribute value of "0.3m", and tide level with an attribute value of "0.7m". Each of these element nodes generates a "observed at ( The system defines directed edges based on the element entities. Determine the environmental status And combined with navigation aid object nodes The Tianjin Port main channel triggers a stoppage exceeding limits alarm according to the state determination function, and then generates a corresponding "delayed departure" decision node. (by relation edge) (Triggering) fully ensures the correct transformation results from multi-source parameters to fused knowledge graph entities.

[0069] 8) Map Update: When the next round of forecast images or measured data arrives, the system automatically performs the following: construction of new time slot nodes; updating of existing node attributes; construction of new edge relationships; and output of a new version of the map. This enables dynamic updating of the navigation-aided decision map.

[0070] Therefore, this invention employs the aforementioned method for constructing a fusion decision map of port cluster navigation aids, effectively solving problems such as difficulty in extracting image-based information, difficulty in fusing multi-source heterogeneous data, and inconsistent knowledge representation. First, image-based forecast pages and structured environmental data are collected. Second, the images are horizontally divided into time slots, and wind and wave direction arrow areas are located based on layout templates. The main direction of the arrows is automatically extracted and converted into standard azimuth angles through grayscale conversion, threshold segmentation, and principal component analysis, significantly improving the efficiency and accuracy of directional element recognition. Then, the extracted directional elements and structured data are uniformly reconstructed into an aligned long table containing station, date, time slot, element name, and value. Missing values ​​are interpolated and standardized using time series, generating high-quality fused environmental data and achieving unified spatiotemporal alignment of elements such as wind, waves, and tides. Based on this, a map pattern containing time, environmental elements, environmental state, navigation aids, and decision entities is constructed, defining relationship templates such as "observed at," "affected," and "triggering decision," as well as state judgment functions. Finally, the fused data is instantiated into nodes and relation edge sets to form a navigation-aid decision graph that supports dynamic incremental updates and version management. It can be exported as GraphML, JSON, or graph database format, directly providing computable, reasonable, and reusable intelligent decision support for the port cluster digital twin platform.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for constructing a fusion decision map of navigation aids for port clusters, characterized in that, Includes the following steps: Step S1: Obtain the original data set of multi-source heterogeneous environment in the navigation aid scenario of port clusters. Each data item Represented as: ,in: Indicates the site identifier; Indicates time information; Indicates the type of environmental element; Indicates decision-making; Indicates the corresponding value or state; Indicate the data source; The data type can be represented as image data, structured data, or text data; the raw data from a multi-source heterogeneous environment includes at least image-based environmental information and structured environmental data. Step S2: Divide the imaged environmental information into time slots and feature regions, and divide the time axis layout region and environmental feature region. The environmental feature region includes directional environmental feature region, numerical environmental feature region and state environmental feature region. Identify and extract the directional environmental feature, numerical environmental feature region and state environmental feature region under each time slot. Step S3: Perform unified structural reconstruction on the structured environmental data and the directional environmental elements extracted in step S2 to form an environmental element dataset organized by date and time slot; Step S4: Perform missing value processing and data standardization on the environmental element dataset to generate fused environmental data; Step S5: Construct the navigation-aided knowledge graph schema, relation template, and state determination function: Define the entity type set. ,in, Represents a time entity. Represents environmental element entities, Represents an entity representing the state of the environment. Represents the navigation aid object entity. Represent the decision entity; define the set of relation types. The logical association between nodes and edges in a knowledge graph is explicitly defined as a knowledge graph relation triple structure. This relation triple structure is used to express the semantic links between entities in the schema layer. Relation templates are defined based on this relation triple structure. ,in: Indicates that something is being observed. Indicates impact, Indicates triggering, Indicates that it acts on, Indicates the data source. Indicates an update. Indicates indirect relationship. Indicate decision-making needs, Indicates the type of the head entity in a relation triple. Indicates the tail entity type in the relation triple; Represents the specific instance relationship edge type between the header entity type and the tail entity type; defines the state determination function. ,in: Indicates the first Class environment state; This represents the standardized environmental elements; This represents the state determination function; Step S6: Based on the graph pattern defined in step S5, instantiate the fusion environment data generated in step S4 into a node set. and relation edge set Generate a fusion decision map of navigation aids for port clusters.

2. The method for constructing a port cluster navigation aid-based decision map according to claim 1, characterized in that, Step S2 specifically includes: dividing the entire image into multiple time slot units horizontally according to the forecast page layout template; within each time slot unit, the wind direction arrow area and wave direction arrow area are used as directional environmental element areas, the numerical area is used as a numerical environmental element area, and the icon area is used as a status environmental element area; the extracted directional environmental element area images are sequentially grayscaled and thresholded to extract the foreground pixel coordinate set; the principal component analysis method is used to calculate the first principal direction vector of the foreground pixels; the arctangent function is used to calculate the angle of the principal direction vector, and the angle is converted into a standard azimuth angle as the directional element value of the time slot.

3. The method for constructing a port cluster navigation aid-based decision map according to claim 2, characterized in that: The specific method for forming the environmental element dataset in step S3 is as follows: the wide table structure of the original data is reconstructed into a long table structure containing fields such as "site identifier, date, time slot, environmental element name, environmental element value, data source identifier, and data type identifier"; and environmental elements from different data sources are aligned according to date and time slot to form a unified sample of multiple elements under each time slot.

4. The method for constructing a port cluster channel navigation aid fusion decision map according to claim 3, characterized in that, The missing value handling in step S4 is as follows: for numerical environmental features, linear interpolation is performed on the time series according to the feature category; for directional features, nearest neighbor direction completion and adjacent time slot direction smoothing are used to fill in the missing values.

5. The method for constructing a port cluster channel navigation aid fusion decision map according to claim 4, characterized in that: The relation weight function in step S5 is defined as follows: ； in, Represents node entities and The strength of the relationship between them; This is the weight calculation function; These represent the environmental elements or state parameters corresponding to the head and tail entities associated with the relationship, respectively; t is the time variable.

6. The method for constructing a port cluster channel navigation aid fusion decision map according to claim 5, characterized in that: In step S5, the navigation aid decision is determined by the following function: ； Where D represents the navigation aid decision result; Represents environmental state variables; This represents the decision function.

7. The method for constructing a port cluster navigation aid-based decision map according to claim 6, characterized in that, The instantiation and generation process of the map in step S6 includes: S61. Instantiation is defined as the process of mapping fused data to a specific graph structure, expressed as: ； in, The set of nodes that generate instances; This represents the set of relational edges that generate instances; S62. Define a mapping function with the following expression: ； in, For functional relation symbols, This represents the mapping relationship between nodes and relation edges; This represents the set of nodes generated after mapping. This represents the edge set generated after mapping; S63. Generate nodes and relation edges, specifically: For each data record Extract timestamp as Create the corresponding time node instance The extracted element values ​​are: Create corresponding environment element node instances According to the decision function Determine and generate state node instances ; Match the corresponding waterway or ship to generate navigation aid object node instances. ; Generate decision node instances based on specific business rules According to the relation template Build connections and observation relationship edge Influence on the relationship and triggering decision relationship edge Entity diagram structure; S64. Assign values ​​to the generated node attributes, including: value = standardized environmental element value; timestamp = time slot; source = data source; and output the instantiated result graph with time, environment, decision entity and related edges.

8. The method for constructing a port cluster channel navigation aid fusion decision map according to claim 7, characterized in that, It also includes step S7: dynamic update and version management of the map. When new raw data is obtained, the existing map is updated incrementally by matching the site, date and time slot. The incremental update includes: creating new time nodes for the new time slots, adding new attribute values ​​to existing environmental element entities, and recalculating and updating the weights of the relationship edges according to the relationship weight function.

9. The method for constructing a port cluster navigation aid-based decision map according to claim 8, characterized in that: The fused decision graph is exported as one or more of the following: GraphML file, JSON file, graph database record, or graph interface data for digital twin platforms, to provide decision support for port cluster waterway digital twin platforms.

Images