Lake wetland ecological restoration effect intelligent evaluation system based on multi-source data fusion

Abstract

The invention relates to the technical field of ecological environment information processing, in particular to a multi-source data fusion lake wetland ecological restoration effect intelligent evaluation system which comprises a data access module, a unified space-time fusion module, an intelligent evaluation module, a causal inference and combination generation module and a display and interaction module. The system accesses ecological state data and restoration measure data from multiple sources, and forms a data set under a unified space-time reference through time alignment, space alignment and quality processing. The system further constructs a spatial unit model and a graph structure, and performs spatio-temporal relationship modeling on the time series features and spatial topological relationships of the nodes to generate ecological state assessment data and prediction data. And constructing a causal relationship model based on the restoration measure variable and the ecological state variable, and generating attribution data representing the restoration measure influence degree and restoration measure combination data meeting the constraint condition according to the causal relationship model. The system also provides data display and interaction functions, and can be used for analysis and management of lake wetland ecological restoration.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ecological environment information processing, and particularly relates to a multi-source data fusion lake wetland ecological restoration effect intelligent evaluation system. BACKGROUND

[0002] Lake wetland ecosystems play an important role in regional water environment regulation, biodiversity maintenance and ecological function support. In order to improve the ecological condition of lake wetlands, various ecological restoration measures are widely used, including water body treatment, vegetation restoration, source control and reduction, hydrological regulation, etc. However, in the actual restoration management process, the ecological monitoring data is diverse, including manual investigation, monitoring station data, remote sensing image, video patrol data and management system records, etc. These data have obvious differences in time scale, spatial scale, data format and collection method, and are difficult to be directly utilized comprehensively.

[0003] Therefore, the prior art has deficiencies in multi-source data fusion, spatio-temporal structure modeling, causal relationship identification and restoration measure combination generation, and a lake wetland ecological restoration effect evaluation system capable of unified processing of multi-source data, construction of spatio-temporal structure expression, execution of causal analysis and generation of restoration measure combination is needed to support the analysis and decision-making needs in the ecological management process. SUMMARY

[0004] In order to make up for the above deficiencies, the present application provides a multi-source data fusion lake wetland ecological restoration effect intelligent evaluation system, aiming at improving the problems in the prior art that multi-source data is difficult to be unified processed, ecological state lacks spatio-temporal structured expression and the role relationship of restoration measures is not clear.

[0005] The present application provides the following technical scheme, a multi-source data fusion lake wetland ecological restoration effect intelligent evaluation system, comprising: A data access module is used to acquire multi-source, heterogeneous data related to the ecological state of lake wetlands and restoration measures, and to access and organize the data according to time and space identifiers; A unified spatio-temporal fusion module is used to perform time alignment, space alignment and data quality processing on the accessed data based on a preset spatial unit model, and to generate comprehensive feature data for subsequent analysis; An intelligent evaluation module is used to map the comprehensive feature data to a constructed spatio-temporal graph structure, and to perform spatio-temporal relationship modeling on the spatio-temporal graph structure to form evaluation data and prediction data for representing the ecological restoration state; A causal inference and combination generation module is used to construct a causal relationship structure according to the evaluation data and restoration measure related data, and to perform causal calculation in the structure to generate attribution data and restoration measure combination data; a display and interaction module for presenting the evaluation data, prediction data, attribution data and measure combination data, and receiving user instructions to trigger relevant data processing procedures.

[0006] Preferably, the data access step comprises: performing acquisition processing on data from different sources while preserving their time labels and spatial identities; performing initial parsing on the accessed data to form raw data sets conforming to preset structural requirements; organizing and storing the parsed data according to source types or uses to establish data input structures for subsequent processing.

[0007] Preferably, the basic processing step of spatio-temporal fusion comprises: performing time alignment processing on data from different time sources to form time series under a unified time reference; performing spatial alignment processing on data from different spatial sources to form spatial correspondence under a unified spatial reference; performing missing check, anomaly identification or consistency verification on the aligned data to form a data set that can be used for feature construction.

[0008] Preferably, the multi-modal feature construction step comprises: parsing and formatting raw feature information related to ecological status from multi-source data; mapping the raw feature information to corresponding spatial units according to a spatial unit model to form a spatial structure consistent feature representation; combining raw feature information from different sources into comprehensive feature data according to a preset fusion method.

[0009] Preferably, the spatial unit model generation step comprises: determining the unit division boundary based on the spatial range of the lake wetland area; dividing the area into units according to the boundary and spatial structure to generate a plurality of spatial units; setting a unique spatial identity for each spatial unit to support data mapping and graph structure construction.

[0010] Preferably, the graph structure construction step comprises: setting the spatial unit as a node in the graph structure as a basic component unit of the graph structure; establishing connection edges between nodes according to spatial adjacency relationships or ecological correlation relationships to form a graph topology reflecting the spatial structure; loading the comprehensive feature data to the corresponding node or connection edge to form a spatio-temporal graph data structure.

[0011] Preferably, the spatio-temporal relationship modeling step comprises: performing temporal relationship modeling on the time series data of the nodes to identify correlations in the time dimension; performing spatial relationship modeling on the connection relationships between the nodes to identify correlations in the spatial topology; combining the temporal relationship modeling results and the spatial relationship modeling results to generate the evaluation data and the state prediction data.

[0012] Preferably, the causal relationship structure construction step comprises: identifying information variables for describing the restoration measures and information variables for describing the ecological state; constructing a causal structure according to statistical correlations between the variables to represent directional relationships between the variables; performing parameter calculation on the variable relationships in the causal structure to form a causal relationship model.

[0013] Preferably, the attribution data generation step comprises: performing causal quantification calculation on the restoration measure variables in the causal relationship model; forming attribution data for representing the impact degree of the restoration measures according to the causal quantification calculation results.

[0014] Preferably, the restoration measure combination generation step comprises: setting constraint conditions required to be satisfied for generating the restoration measure combination; generating restoration measure combination data according to the causal relationship model and the constraint conditions; outputting the restoration measure combination data for display and calling by the interaction module.

[0015] The present application has the following beneficial effects: 1、In the present application, a cross-scale data processing mechanism is constructed through unified spatio-temporal fusion of multi-source data. Data from different time frequencies, spatial resolutions and data formats can be uniformly processed, so that various types of data have comparability and structural consistency under the same spatio-temporal reference. This method provides a complete data basis for subsequent feature construction, graph structure modeling and causal analysis, avoids the analysis chain breaking problem caused by non-uniform data formats in traditional methods, and improves the applicability of the system in complex ecological scenarios.

[0016] 2、In the present application, based on the space unit model, the spatio-temporal graph structure is constructed, and the structured expression of the ecological state is realized. By dividing the lake wetland into space units and establishing the node connection relationship, and by loading the comprehensive characteristics to the graph structure, the ecological state has expressiveness in space and traceability in time. The system can perform time relationship modeling and space relationship modeling on this basis, so that the change process and spatial correlation of the ecological state can be described in a unified model, enhancing the integrity and continuity of the analysis results.

[0017] 3、In the present application, the causal relationship structure and combination generation mechanism are adopted, and the structured inference ability for restoration measures is realized. By identifying the restoration measure variables and the ecological state variables, the causal relationship model is constructed and the relationship is quantified, which can form attribution data reflecting the influence degree of the measures. On this basis, the constraint condition is used to generate the measure combination scheme, so that the system can provide structured restoration measure reference, establish a complete logical chain from data, model to measure selection, and improve the applicability and operability of the system in the ecological restoration management scene. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 A system architecture diagram of a multi-source data fusion lake wetland ecological restoration effect intelligent evaluation system is provided. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0020] Embodiment one In the first embodiment of the present application, the present application provides a multi-source data fusion lake wetland ecological restoration effect intelligent evaluation system, as shown in Figure 1 The steps include: A data access module is used to acquire multi-source, heterogeneous data related to the ecological state and restoration measures of the lake wetland, and to organize the data according to time and space identifiers; Further, the data access step includes: Performing acquisition processing on data from different sources and retaining their time labels and space identifiers; Performing initial analysis on the accessed data to form a raw data set conforming to the preset structure requirements; Organize and store the analyzed data according to source type or purpose to establish a data input structure for subsequent processing.

[0021] Specifically, firstly, the system collects information representing ecological status or restoration measures from various sources, and records and retains the collected data according to its original time label and spatial identifier to ensure the traceability of data from different sources in both time and space dimensions. At the same time, the specific data source is not limited and can include monitoring platforms, remote sensing platforms, manual survey records, numerical model outputs, or management system records, etc. Data access methods can include interface calls, file imports, or database reads.

[0022] Then, the system performs initial parsing processing on the aforementioned access data. During the parsing process, the system uses corresponding parsing algorithms to extract structured content from different types of data according to preset data structure requirements: for tabular or database-type structured data, field mapping rules and an extraction-transformation-loading process can be used to map the original fields to unified field names and data types; for text records, natural language processing algorithms such as regular expression matching, template matching, word segmentation, and entity recognition can be used to extract time information, spatial location, indicator names, and their values ​​from the text; for image or remote sensing data containing metadata, attributes such as acquisition time, coverage area, and resolution can be obtained through metadata reading, spatial index parsing, and, when necessary, simple image segmentation or raster index calculation; for time series monitoring data, time series resampling and interpolation algorithms can be combined to standardize non-equidistant records. Through the above parsing and transformation, data from different sources and in different formats are unified into a structured raw dataset with clearly defined time fields, spatial fields, and attribute fields. Finally, the system organizes and stores the parsed data based on its source type, spatial attributes, or subsequent processing requirements. The data is categorized into time-series data, spatial unit data, or measure record data, and time, spatial, or type indexes are created as needed to ensure that the unified spatiotemporal fusion module can efficiently read and process this data. The data can be stored in a relational database, time-series database, or file system. Through the above data access steps, the system can transform data from multiple sources, heterogeneous structures, and different formats into raw datasets with unified structure, clearly defined fields, and time and space indexes, providing a standardized data input foundation for subsequent spatiotemporal fusion processing, spatiotemporal graph structure construction, and causal relationship analysis.

[0023] The unified spatiotemporal fusion module is used to perform time alignment, spatial alignment, and data quality processing on the accessed data based on a preset spatial unit model, and generate comprehensive feature data for subsequent analysis. Furthermore, the basic processing steps for spatiotemporal fusion include: Perform time alignment processing on data from different time sources to form a time series under a unified time base; Spatial alignment is performed on data from different spatial sources to form spatial correspondences under a unified spatial benchmark; Perform missing data checks, anomaly detection, or consistency checks on the aligned data to form a dataset that can be used for feature construction.

[0024] Specifically, for data from different sampling periods, different recording times, or irregular time intervals, the system can perform time alignment processing based on a preset unified time benchmark. Time alignment can be achieved through time resampling, time window aggregation, or time interpolation, so that each data point has a corresponding numerical or state expression at the same time node. Through this processing, data from different time sources can generate continuous or comparable time series under a unified time benchmark. For data from different spatial resolutions, coordinate systems, or geographic coverage areas, the system performs spatial alignment processing based on the spatial unit model, uniformly mapping the data to a preset spatial reference. Spatial alignment may include steps such as coordinate transformation, spatial resampling, or spatial rasterization. Through spatial alignment, data from different spatial sources can establish a unified spatial correspondence, facilitating subsequent unified analysis. After completing spatiotemporal alignment, the system performs necessary data quality processing, including missing data checking, outlier identification, and consistency verification between different data sources. Data quality processing can be achieved through missing data marker identification, threshold rule detection, statistical distribution analysis, or simple anomaly detection algorithms. Through this quality processing, the system can generate a dataset suitable for feature construction, providing a reliable data foundation for subsequent multimodal fusion, graph structure construction, and causal inference. Through the above spatiotemporal fusion basic processing steps, the system can unify data from different time and space sources under a consistent spatiotemporal benchmark, and form a data set with standardized structure and consistency through quality processing, providing directly callable basic data support for multi-source feature construction and spatiotemporal graph relationship modeling.

[0025] Furthermore, the steps for constructing multimodal features include: Extract raw feature information related to ecological status from multi-source data and format it accordingly; The original feature information is mapped to the corresponding spatial unit according to the spatial unit model to form a feature representation with consistent spatial structure; The raw feature information from different sources is combined into comprehensive feature data according to a preset fusion method.

[0026] Specifically, the system first parses raw feature information related to ecological status from structured, semi-structured, or unstructured data from different sources. Different processing methods can be used depending on the data type. The parsed raw feature information is then formatted to form feature items with unified field naming, unified data types, and unified time / space indicators. The system can map formatted raw feature information to corresponding spatial units based on a spatial unit model, achieving a unified expression of spatial structure. Mapping methods can include coordinate matching, spatial index lookup, region coverage determination, or rasterization mapping. For example, the spatial unit number of the data can be determined based on its spatial coordinates; for data with coverage regions, the overlap relationship with spatial units can be calculated to determine the corresponding unit affiliation or weighted allocation. Through these mapping steps, feature information from heterogeneous sources is unified into the same spatial unit structure. The system can combine original feature information from different sources belonging to the same spatial unit according to a preset feature fusion method to generate comprehensive feature data for subsequent spatiotemporal modeling. The fusion method can employ feature concatenation, weighted fusion, statistical aggregation, rule combination, or other methods that enable multi-feature merging. The fused comprehensive feature data retains representative features of various types of information and uses spatial units as the index structure, facilitating subsequent loading into the spatiotemporal graph structure. Through the above multimodal feature construction steps, the system can uniformly extract, map, and fuse original feature information from different sources, formats, and contents into comprehensive feature data. This enables subsequent spatiotemporal graph relationship modeling to be processed based on input data with consistent structure and complete information, thereby ensuring the operability and consistency of the system's data foundation.

[0027] The intelligent assessment module is used to map comprehensive feature data to the constructed spatiotemporal graph structure and perform spatiotemporal relationship modeling on the spatiotemporal graph structure to form assessment data and prediction data to represent the state of ecological restoration. Furthermore, the spatial unit model generation steps include: Determine the unit boundaries based on the spatial extent of the lake and wetland area; The region is divided into units based on its boundaries and spatial structure to generate multiple spatial units; Each spatial unit is assigned a unique spatial identifier to support data mapping and graph structure construction.

[0028] Specifically, the system first reads the spatial extent information of the lake and wetland area. This information can come from geographic boundary files, regional bounding boxes in remote sensing imagery, administrative division vector data, or a pre-built lake boundary database. Then, based on this, the system determines the boundary extent used for spatial unit division, which defines the spatial scope of subsequent unit divisions. Boundary reading and identification can be achieved through spatial vector analysis, raster edge extraction, and coordinate range calculation. After defining the boundaries, the system divides the lake wetlands into units based on the spatial structure of the region. Different strategies can be selected based on the specific application. The output of the division consists of multiple spatial units with clearly defined spatial boundaries. The system assigns a unique spatial identifier to each of these generated spatial units, such as a number, code, or other form of unique index. This identifier can be used to: map multi-source feature information to specific spatial units; serve as node identifiers when constructing the spatiotemporal graph structure; and manage indexes during data storage and retrieval. Through this step, the system forms a stable and referable set of spatial units, providing a consistent spatial structure foundation for subsequent data fusion, spatiotemporal graph modeling, and causal analysis. Through the spatial unit model generation step, the system can form discretized spatial units based on the spatial range and spatial structure of lakes and wetlands, and establish a unique identifier for each unit. This enables multi-source data and subsequent graph structures to be organized and processed under the same spatial benchmark, providing unified spatial structure support for building spatiotemporal models and conducting regional ecological analysis.

[0029] Furthermore, the steps for constructing a graph structure include: Spatial units are set as nodes in a graph structure to serve as the basic building blocks of the graph structure. Based on spatial adjacency or ecological association, connecting edges are established between nodes to form a graph topology that reflects the spatial structure; The comprehensive feature data is loaded into the corresponding nodes or connecting edges to form a spatiotemporal graph data structure.

[0030] Specifically, the system reads all defined spatial units in the spatial unit model and sets each spatial unit as a node in the graph structure. Each node uses the unique spatial identifier of the spatial unit as its index to accurately correspond to its spatial location in subsequent feature loading and model calculation. The number of nodes is consistent with the number of spatial units. Then, the system establishes connecting edges for the nodes based on the relationships between spatial units to form a graph topology describing the spatial structure. Spatial adjacency relationships can be obtained through spatial unit boundary contact judgment, spatial distance threshold judgment, nearest neighbor search, or calculation based on regional coverage. Ecological association relationships can be determined based on information such as water connectivity, ecological function relationships, physical connections, or habitat continuity. Edges can be undirected, directed, or weighted, and the weights can be distance values, association strength, or values ​​based on preset rules. After constructing the graph topology, the system loads the comprehensive feature data generated in the aforementioned multimodal feature construction steps onto the corresponding nodes or connecting edges. Node features may contain the feature vector sequence of the spatial unit at different times; edge features may contain information such as the distance between nodes, the degree of mutual influence, or ecological relevance. The feature data is organized into a time series according to time labels, enabling the graph structure to carry spatiotemporal data and realize subsequent spatiotemporal relationship modeling. Through this step, the system forms a spatiotemporal graph data structure containing spatial structure information and temporal feature information for use by the intelligent evaluation module. Through the graph structure construction steps, the system can express the adjacency or association between spatial units and them in a graph structure, and load comprehensive feature data onto nodes or connecting edges, thereby forming a spatiotemporal graph data structure for spatiotemporal analysis, providing a data foundation that can be directly called for the intelligent evaluation module to perform spatial and spatiotemporal modeling.

[0031] Furthermore, the spatiotemporal relationship modeling steps include: Perform time relationship modeling on the time series data of the nodes to identify correlations in the time dimension; Perform spatial relationship modeling on the connections between nodes to identify relevant relationships in the spatial topology; The results of temporal relationship modeling and spatial relationship modeling are combined to generate evaluation data and state prediction data.

[0032] Specifically, the system first analyzes the feature sequences corresponding to each spatial unit node at different times to capture the changing patterns of the nodes in the time dimension. Temporal relationship modeling can be implemented using recursive network structures, convolutional temporal modeling structures, time window statistical rules, or other methods suitable for time series analysis. The goal is simply to extract temporal dependencies from the historical feature sequences of the nodes. Through this processing, the system obtains the temporal modeling representation of the nodes at each time step, which is then used for subsequent spatial modeling. After obtaining the temporal modeling representation of the nodes, the system performs spatial modeling processing on the node representations at the same time segment based on the adjacency or ecological relationships between nodes in the graph structure. Spatial relationship modeling can be achieved through graph structure computation methods, such as feature update methods based on graph convolution, message passing mechanisms, or topological constraints. Through this step, the system obtains the spatial modeling representation of each node, which reflects the dependencies between nodes in the spatial dimension. The system then combines the node representations obtained from the temporal and spatial modeling and inputs them into a predefined assessment or prediction model to generate assessment data of the ecological restoration status and prediction data of the future status. The combination method can be splicing, weighted combination, or other methods that enable feature fusion. The generated data may include the current ecological status assessment value, trend prediction value, or hierarchical data for management analysis of spatial units. Through the aforementioned spatiotemporal relationship modeling steps, the system can identify the temporal dependencies and spatial topological relationships of nodes on the spatiotemporal graph data structure, and combine the two to generate assessment and prediction data for ecological restoration status analysis, providing the necessary data foundation for subsequent causal analysis and the generation of action combinations. The causal inference and combination generation module is used to construct a causal relationship structure based on assessment data and remediation action-related data, and to perform causal calculations in this structure to generate attribution data and remediation action combination data. Furthermore, the steps for constructing a causal relationship structure include: Identify information variables used to describe restoration measures and information variables used to describe ecological state; Construct causal structures based on statistical relationships between variables to represent directional relationships between variables; Perform parameter calculations on the variable relationships in the causal structure to form a causal relationship model.

[0033] Specifically, the system first identifies information variables representing restoration measures and information variables representing ecological status from a uniformly managed feature dataset. Restoration measure variables include recorded information related to human intervention, management actions, or engineering measures; ecological status variables include water environment indicators, vegetation indicators, or habitat indicators, etc. The system treats these variables as potential causal and outcome variables, respectively, and establishes basic structures such as variable names, variable types, and variable value ranges for them, which are then used to participate in the subsequent construction of causal structures. Based on the joint data of identified restoration measure variables and ecological state variables, the system analyzes the statistical correlations between the variables to form a preliminary causal relationship structure. Statistical correlations can be obtained through correlation analysis, conditional independence tests, structure learning algorithms, or rule inference, and are used to determine whether there are potential causal directions or dependencies between the variables. The system then constructs a causal structure graph based on the correlation results, where nodes represent variables and connections represent directional dependencies between variables. This causal structure can be a directed graph structure or other data structures that can represent directional relationships. After the causal structure diagram is determined, the system performs parameter calculations on the variable relationships represented by the connecting edges in the diagram to form a causal relationship model that can be used to quantify these relationships. Parameter calculations can be based on regression estimation, probabilistic inference, conditional distribution learning, or other computational methods capable of quantifying variable relationships, and are used to characterize the intensity or direction of the influence of restoration measures on ecological state variables. Through the above processing, the system forms a causal relationship model describing the directional relationship between restoration measures and ecological state, providing a structured data foundation for attribution analysis and the generation of measure combinations. Through the causal relationship structure construction step, the system can identify potential causal dependencies from restoration measure variables and ecological state variables, and organize them into a causal relationship model that can be used for calculation, thus establishing the necessary structural foundation for subsequent causal effect calculation and measure combination generation.

[0034] Furthermore, the attribution data generation steps include: Perform causal quantification calculations on the remedial measures variable in the causal relationship model; Attribution data is generated based on the causal quantification results to represent the degree of impact of remediation measures.

[0035] Specifically, based on the established causal relationship model, the system performs causal quantification calculations for each restoration measure variable to assess the intensity of its impact on the ecological state variable. Causal quantification can be achieved through various methods, including but not limited to inference based on conditional probability distributions, effect estimation based on comparative scenarios, parameter solving based on structural equation models, or causal effect assessment based on perturbation simulations. Through this calculation step, the system obtains quantitative data reflecting the direction and extent of the impact of each restoration measure variable on the ecological state variable. Then, based on the results of causal quantification calculations, the system generates attribution data to represent the contribution of each restoration measure variable to the ecological state. Attribution data can be in scalar, vector, or structured record format to reflect the influence relationships between different restoration measures on different ecological state indicators.

[0036] In the actual generation process, the system can organize, collect or structure the quantitative results so that the attribution data can be used in the subsequent measure combination generation steps or for analysis in the system display module. Through the attribution data generation step, the system can quantify the causal relationship between restoration measures and ecological status based on the causal relationship model, and organize the quantification results into attribution data, providing a directly usable causal basis for the generation of combined restoration measures.

[0037] Furthermore, the steps for generating the remedial measures combination include: Define the combination of repair measures to generate the constraints that need to be satisfied; Data on combinations of remedial measures are generated based on causal relationship models and constraints. Output the combination of repair measures data for use by the display and interactive modules.

[0038] Specifically, the system first sets constraints for the generation of combined measures based on actual lake and wetland restoration needs. These constraints may include the spatial and temporal scope of implementation, resource availability, technical constraints, ecological protection requirements, or management strategies. Constraints can be obtained through manual configuration or system presets, or automatically generated based on attribution and evaluation data. This invention does not limit the type or number of constraints, as long as they can constrain or screen the combination generation process. After the constraints are set, the system uses the directional relationships between variables in the causal model and the quantified degree of influence in the attribution data to generate restoration measure combination data that meets the constraints. Combination generation can be accomplished through rule selection, search strategies, combination optimization, constraint reasoning, or other feasible methods for constructing combinations. Therefore, the system can select combination schemes that meet the constraints from multiple candidate measures based on the influence relationships of different restoration measures on ecological state indicators in the causal model, and form structured combination data. The generated remediation measure combination data is then processed and output to the system's display and interaction module. The output data may include the remediation measure items included in the combination, their corresponding spatial locations, timelines, collaborative relationships between measures, or optional alternatives. The system can use this combination data for display, analysis, or subsequent decision support as needed. Through the steps of generating restoration measures combinations, the system can combine causal relationship models and set constraints to generate data on combinations of measures that can be used for ecological restoration management. This data is then provided to the display module for organization and retrieval, enabling the system to generate restoration plans and form a closed loop with the overall data analysis process.

[0039] The display and interaction module is used to present evaluation data, prediction data, attribution data, and combination of measures data, and to receive user instructions to trigger relevant data processing flows. Specifically, through the display and interaction module, the system can display evaluation data, prediction data, attribution data, and combination of measures data, and support interactive operation of the data analysis process by users. This allows the system's calculation results to be presented in a visual form, and the data processing flow can be dynamically triggered through interactive commands, thereby building a complete data analysis and decision support chain.

[0040] Example 2: In a lake wetland restoration project, due to the diverse sources of monitoring data, including manual survey records, monitoring station data, remote sensing image data, drone patrol data, and restoration measure records in the management system, there are problems such as inconsistencies in time scales, spatial mismatches, and inconsistent data formats among the different data. To solve these problems, this invention provides an intelligent evaluation system for the ecological restoration effect of lake wetlands that integrates multi-source data. Its structure is as follows: Figure 1 As shown. The specific implementation process of this system is as follows: Data access involves the system accessing ecological status data and restoration measure data from monitoring platforms, remote sensing platforms, inspection records, and management systems, and performing data parsing and format standardization to form a processable raw dataset.

[0041] The unified spatiotemporal fusion system performs time and spatial alignment processing on data from different sources, making all data comparable under a unified time scale and a unified spatial benchmark. The system then performs missing data checks, anomaly identification, and consistency verification to form a dataset for feature construction.

[0042] Multimodal feature construction involves extracting ecological state-related features from various input data, mapping the features to corresponding spatial units based on the spatial unit model, and then fusing features from different sources using a preset method to form comprehensive feature data.

[0043] The spatial unit model is generated by determining the spatial extent of this embodiment based on the lake and wetland boundaries and dividing it into several spatial units according to the regional structure. Each spatial unit is assigned a unique number to support subsequent feature loading and graph structure construction.

[0044] In graph structure construction, the system uses spatial units as nodes and constructs connecting edges based on spatial adjacency relationships to form a graph topology corresponding to the actual structure of lakes and wetlands. Then, the aforementioned comprehensive feature data is loaded onto the nodes or connecting edges to form a spatiotemporal graph data structure for spatiotemporal modeling.

[0045] Spatiotemporal relationship modeling involves performing time modeling on the time series characteristics of nodes and combining graph structure to model the spatial relationships between nodes, generating assessment data to describe the ecological state and prediction data for the future state.

[0046] The causal relationship structure is constructed by identifying restoration measure variables and ecological state variables, and constructing a causal structure based on the statistical relationships between the variables; parameter calculations are performed on the relationships in the causal structure to form a causal relationship model.

[0047] Attribution data generation utilizes a causal relationship model to perform causal quantification calculations on the remediation measure variables and generates attribution data to represent the degree of impact of each remediation measure.

[0048] The remediation measure combination generation process generates remediation measure combination data that meets the constraints after setting spatial, temporal, or resource constraints, based on the causal relationship model and attribution data.

[0049] The system provides visualization and interaction of assessment data, forecast data, attribution data, and combination of measures data, and receives user interaction commands to trigger relevant data processing flows.

[0050] Finally, it should be noted that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart evaluation system for the ecological restoration effect of lakes and wetlands based on multi-source data fusion, characterized in that, include: The data access module is used to acquire multi-source, heterogeneous data related to the ecological status and restoration measures of lakes and wetlands, and to access and organize the data according to time and spatial identifiers. The unified spatiotemporal fusion module is used to perform time alignment, spatial alignment, and data quality processing on the accessed data based on a preset spatial unit model, and generate comprehensive feature data for subsequent analysis. The intelligent assessment module is used to map the comprehensive feature data to the constructed spatiotemporal graph structure and perform spatiotemporal relationship modeling on the spatiotemporal graph structure to form assessment data and prediction data for representing the state of ecological restoration. The causal inference and combination generation module is used to construct a causal relationship structure based on the assessment data and remediation measure-related data, and to perform causal calculations in the structure to generate attribution data and remediation measure combination data; The display and interaction module is used to present the evaluation data, prediction data, attribution data, and combination of measures data, and to receive user instructions to trigger relevant data processing flows.

2. The intelligent evaluation system for lake and wetland ecological restoration effects based on multi-source data fusion as described in claim 1, characterized in that, Data access steps include: Perform acquisition processing on data from different sources and retain their time stamps and spatial identifiers; Perform initial parsing on the incoming data to form an original dataset that meets the preset structure requirements; The parsed data is organized and stored according to its source type or purpose to establish a data input structure for subsequent processing.

3. The intelligent evaluation system for lake and wetland ecological restoration effects based on multi-source data fusion as described in claim 1, characterized in that, The basic processing steps for spatiotemporal fusion include: Perform time alignment processing on data from different time sources to form a time series under a unified time base; Spatial alignment is performed on data from different spatial sources to form spatial correspondences under a unified spatial benchmark; Perform missing data checks, anomaly detection, or consistency checks on the aligned data to form a dataset that can be used for feature construction.

4. The intelligent evaluation system for lake and wetland ecological restoration effects based on multi-source data fusion according to claim 1, characterized in that, The steps for constructing multimodal features include: Extract raw feature information related to ecological status from multi-source data and format it accordingly; The original feature information is mapped to the corresponding spatial unit according to the spatial unit model to form a feature representation with consistent spatial structure; The original feature information from different sources is combined into comprehensive feature data according to a preset fusion method.

5. The intelligent evaluation system for lake and wetland ecological restoration effects based on multi-source data fusion according to claim 1, characterized in that, The steps for generating a spatial unit model include: Determine the unit boundaries based on the spatial extent of the lake and wetland area; The region is divided into units based on its boundaries and spatial structure to generate multiple spatial units; Each spatial unit is assigned a unique spatial identifier to support data mapping and graph structure construction.

6. The intelligent evaluation system for lake and wetland ecological restoration effects based on multi-source data fusion according to claim 1, characterized in that, The steps for constructing a graph structure include: Spatial units are set as nodes in a graph structure to serve as the basic building blocks of the graph structure. Based on spatial adjacency or ecological association, connecting edges are established between nodes to form a graph topology that reflects the spatial structure; The comprehensive feature data is loaded into the corresponding nodes or connecting edges to form a spatiotemporal graph data structure.

7. The intelligent evaluation system for lake and wetland ecological restoration effects based on multi-source data fusion according to claim 1, characterized in that, The steps for spatiotemporal relationship modeling include: Perform time relationship modeling on the time series data of the nodes to identify correlations in the time dimension; Perform spatial relationship modeling on the connections between nodes to identify relevant relationships in the spatial topology; The results of temporal relationship modeling and spatial relationship modeling are combined to generate evaluation data and state prediction data.

8. The intelligent evaluation system for lake and wetland ecological restoration effects based on multi-source data fusion according to claim 1, characterized in that, The steps for constructing a causal relationship structure include: Identify information variables used to describe restoration measures and information variables used to describe ecological state; Construct causal structures based on statistical relationships between variables to represent directional relationships between variables; Perform parameter calculations on the variable relationships in the causal structure to form a causal relationship model.

9. The intelligent evaluation system for lake and wetland ecological restoration effects based on multi-source data fusion according to claim 1, characterized in that, The attribution data generation steps include: Perform causal quantification calculations on the remedial measures variable in the causal relationship model; Attribution data is generated based on the causal quantification results to represent the degree of impact of remediation measures.

10. The intelligent evaluation system for lake and wetland ecological restoration effects based on multi-source data fusion according to claim 1, characterized in that, The steps for generating a combination of remedial measures include: Define the combination of repair measures to generate the constraints that need to be satisfied; Generate remedial measure combination data based on the causal relationship model and the aforementioned constraints; The combined data of the repair measures is output for use by the display and interaction modules.

Images