GIS Layer Analysis Method and System Based on Automatic Map Combination and Parsing

Abstract

This invention discloses a GIS layer analysis method and system based on automatic map merging and parsing, relating to the field of geographic information systems. The method includes: real-time access to multi-source dynamic layer data streams updated at a frequency greater than a preset frequency; constructing a spatiotemporal causal knowledge graph based on the multi-source dynamic layer data streams, using spatial features as nodes and association rules as directed edges; performing multi-layer joint simulation based on the spatiotemporal causal knowledge graph to generate a predictive thematic layer containing time series confidence intervals; calculating the dynamic time warping distance between the predictive thematic layer and the real-time monitoring layer, and generating a spatial coordinate binding difference warning signal when the dynamic time warping distance exceeds an adaptive distance threshold. This invention solves the technical problems of insufficient prediction accuracy and reliability in existing GIS layer analysis methods based on automatic map merging and parsing, achieving the technical effect of improving prediction accuracy and reliability.

Description

Technical Field

[0001] This application relates to the field of geographic information systems, and in particular to GIS layer analysis methods and systems based on automatic map merging and parsing. Background Technology

[0002] In fields such as urban planning, environmental management, and emergency response, real-time monitoring and prediction of geospatial data changes can effectively reduce potential risks and improve the efficiency and safety of urban operations. Currently, geospatial data is mainly processed and predicted using traditional Geographic Information System (GIS) technology combined with simple statistical analysis methods. These methods rely on static data and analysis at a single time scale, resulting in problems such as low data update frequency, lack of spatiotemporal correlation analysis, and insufficient early warning capabilities, leading to insufficient accuracy and reliability of predictions.

[0003] Currently, GIS layer analysis based on automatic map merging and parsing suffers from insufficient prediction accuracy and reliability. Summary of the Invention

[0004] This application provides a GIS layer analysis method and system based on automatic map merging and parsing. It employs several techniques, including first accessing high-frequency updated multi-source dynamic layer data streams in real time, then constructing a spatiotemporal causal knowledge graph with spatial features as nodes and association rules as directed edges, then conducting multi-layer joint simulation based on this graph to generate a predictive thematic layer containing time series confidence intervals, and finally calculating the dynamic time warping distance between the predictive thematic layer and the real-time monitoring layer. When the distance exceeds an adaptive threshold, a spatial coordinate binding difference warning signal is generated. These techniques achieve the technical effect of improving the accuracy and reliability of prediction.

[0005] This application provides a GIS layer analysis method based on automatic map merging and parsing, including: real-time access to multi-source dynamic layer data streams updated at a frequency greater than a preset frequency; constructing a spatiotemporal causal knowledge graph based on the multi-source dynamic layer data streams, using spatial features as nodes and association rules as directed edges; performing multi-layer joint simulation based on the spatiotemporal causal knowledge graph to generate a predictive thematic layer containing time series confidence intervals; calculating the dynamic time warping distance between the predictive thematic layer and the real-time monitoring layer, and generating a spatial coordinate binding difference warning signal when the dynamic time warping distance is greater than an adaptive distance threshold.

[0006] In a possible implementation, based on the multi-source dynamic layer data stream, a spatiotemporal causal knowledge graph is constructed using spatial features as nodes and association rules as directed edges. The following processes are performed: static spatial features and their corresponding attribute information are extracted from the multi-source dynamic layer data stream to construct static nodes of the knowledge graph; dynamic spatial features and their corresponding spatiotemporal change features are extracted from the multi-source dynamic layer data stream to construct dynamic nodes of the knowledge graph; association rules are constructed based on the interaction relationship between the static and dynamic spatial features, and directed edges of the knowledge graph are established based on the association rules; and a spatiotemporal causal knowledge graph is constructed based on the static nodes, the dynamic nodes, and the directed edges.

[0007] In a possible implementation, association rules are constructed based on the interaction between the static spatial elements and the dynamic spatial elements. Directed edges of the knowledge graph are then established based on these association rules, and the following processing is performed: if the spatial co-occurrence probability of any two static spatial elements is greater than a first preset threshold, a first dynamic association rule is generated, and a first directed edge of the knowledge graph is established based on the first dynamic association rule; if the temporal lag correlation of any two dynamic spatial elements is greater than a second preset threshold, a second dynamic association rule is generated, and a second directed edge of the knowledge graph is established based on the second dynamic association rule; if the dependency correlation between any static spatial element and any dynamic spatial element is greater than a third preset threshold, a third dynamic association rule is generated, and a third directed edge of the knowledge graph is established based on the third dynamic association rule.

[0008] In a possible implementation, multi-layer joint simulation is performed based on the spatiotemporal causal knowledge graph to generate a prediction topic layer containing time series confidence intervals. The following processes are then performed: the spatiotemporal causal knowledge graph is time-discreteized to generate a time propagation sequence; based on the time propagation sequence, starting from the state of each node at the current moment, state propagation simulation is performed along directed edges to obtain a first prediction topic layer; based on the first prediction topic layer, random noise is added to dynamic nodes to obtain a second prediction topic layer; confidence intervals are added to the second prediction topic layer to obtain a prediction topic layer containing time series confidence intervals.

[0009] In a possible implementation, confidence intervals are added to the second predictive thematic layer to obtain a predictive thematic layer containing time series confidence intervals. The following processing is then performed: The basic simulation iterations N are obtained, where N ≥ 1000; multi-layer joint simulation is performed according to the basic simulation iterations N to obtain N second predictive thematic layers; the spatiotemporal state values ​​of each dynamic spatial element in the N second predictive thematic layers are used to calculate the probability distribution function using Gaussian kernel density; for non-critical areas, the 5%–95th percentile of the probability distribution function is taken as the 90% confidence interval, and for critical areas, the 10%–90th percentile of the probability distribution function is taken as the 80% confidence interval; the mean of the spatiotemporal state values ​​of each dynamic spatial element in the N second predictive thematic layers is calculated to obtain a second mean predictive thematic layer; the confidence intervals are added to the second mean predictive thematic layer to obtain a predictive thematic layer containing time series confidence intervals.

[0010] In a possible implementation, the dynamic time-warped distance between the predicted thematic layer and the real-time monitoring layer is calculated, and the following processing is performed: the predicted thematic layer and the real-time monitoring layer are preprocessed for temporal alignment to obtain an aligned layer pair; the aligned layer pair is dynamically warped at multiple scales to obtain a warped layer pair; the cosine similarity and root mean square error of the warped layer pair are calculated; and the cosine similarity and root mean square error are weighted and calculated to obtain the dynamic time-warped distance.

[0011] In a possible implementation, the aligned layer pairs are dynamically normalized at multiple scales to obtain normalized layer pairs, and the following processing is performed: the aligned layer pairs are coarse-grained aligned to obtain coarse-grained aligned layer pairs; the coarse-grained aligned layer pairs are fine-grained optimized to obtain normalized layer pairs.

[0012] In a possible implementation, when the dynamic time warp distance is greater than the adaptive distance threshold, a spatial coordinate-bound difference warning signal is generated, and the following processing is performed: obtain the historical dynamic time warp distance within a preset time range, and use the 90th percentile of the historical dynamic time warp distance as the baseline distance threshold; calculate the quality score of the current data, and adaptively adjust the baseline distance threshold according to the quality score to obtain the adaptive distance threshold; when the dynamic time warp distance is greater than the adaptive distance threshold, generate a multi-level difference warning signal with spatial coordinate bound according to the distance difference and mark the difference area on the map.

[0013] In a possible implementation, the following processing is performed: the multi-source dynamic layer data stream includes: infrastructure sensor data, human activity density heatmap, and environmental parameter time-series variation layer.

[0014] This application also provides a GIS layer analysis system based on automatic map merging and parsing, including: a multi-source dynamic layer data stream access module for real-time access to multi-source dynamic layer data streams updated at a frequency greater than a preset frequency; a spatiotemporal causal knowledge graph construction module for constructing a spatiotemporal causal knowledge graph based on the multi-source dynamic layer data streams, using spatial features as nodes and association rules as directed edges; a multi-layer joint simulation module for performing multi-layer joint simulation based on the spatiotemporal causal knowledge graph to generate a predictive thematic layer containing time series confidence intervals; and a difference warning signal generation module for calculating the dynamic time warping distance between the predictive thematic layer and the real-time monitoring layer, and generating a spatial coordinate-bound difference warning signal when the dynamic time warping distance is greater than an adaptive distance threshold.

[0015] The proposed GIS layer analysis method and system based on automatic map merging and parsing, as described in this application, firstly accesses a multi-source dynamic layer data stream updated at a frequency greater than a preset frequency in real time. Then, based on this multi-source dynamic layer data stream, a spatiotemporal causal knowledge graph is constructed, using spatial features as nodes and association rules as directed edges. Next, multi-layer joint simulation is performed based on the spatiotemporal causal knowledge graph to generate a predictive thematic layer containing time series confidence intervals. Finally, the dynamic time warping distance between the predictive thematic layer and the real-time monitoring layer is calculated. When the dynamic time warping distance exceeds an adaptive distance threshold, a spatial coordinate binding difference warning signal is generated. This achieves the technical effect of improving the accuracy and reliability of prediction. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a flowchart illustrating the GIS layer analysis method based on automatic map merging and parsing provided in an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of the structure of a GIS layer analysis system based on automatic map merging and parsing provided in an embodiment of this application.

[0019] Figure labeling: 10 Multi-source dynamic layer data stream access module, 20 Spatiotemporal causal knowledge graph construction module, 30 Multi-layer joint simulation module, 40 Differential warning signal generation module. Detailed Implementation

[0020] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below.

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0023] This application provides a GIS layer analysis method based on automatic map merging and parsing, such as... Figure 1 As shown, the method includes:

[0024] Step S100: Real-time access to multi-source dynamic layer data streams updated at a frequency greater than the preset frequency.

[0025] Specifically, message queues (such as Kafka and RabbitMQ) are used to receive dynamic layer data streams from various sensors, satellites, IoT devices, etc., in real time. Message queues can efficiently handle high-frequency data inflows and support the aggregation of multi-source data. The preset frequency refers to the minimum data update frequency that the system can handle. For example, if the preset frequency is 1Hz (i.e., once per second), the system needs to be able to receive data that updates once per second or more frequently in real time.

[0026] Develop adapters for different data sources to uniformly convert data of various formats into standard formats (such as GeoJSON and GML) for subsequent processing. Use stream processing engines (such as Apache Flink and Apache Storm) to clean, format, and initially filter real-time data to ensure data integrity and consistency. Stream processing engines can process high-frequency data in real time and support complex event processing logic. Filter out invalid or erroneous data using validation rules (such as data range checks and format validation). Temporarily store the processed data in an in-memory database (such as Redis) for fast access and processing later. In-memory databases provide low-latency data access, suitable for real-time analysis scenarios. For large-scale data, use a distributed file system (such as HDFS) for persistent storage to ensure data reliability and scalability.

[0027] For example, suppose we need to access real-time meteorological data from weather satellites and environmental monitoring data from ground sensors. By configuring a Kafka cluster, we can push the weather satellite data to a Kafka topic once per second, and simultaneously push the ground sensor data to the same topic once per second. The stream processing engine Apache Flink subscribes to this topic, formats the data, removes invalid data, and temporarily stores the processed data in Redis.

[0028] In one possible implementation, step S100 further includes step S110, wherein the multi-source dynamic layer data stream includes: infrastructure sensor data, human activity density heatmap, and environmental parameter time-series change layer.

[0029] Specifically, infrastructure sensor data refers to real-time monitoring data from urban infrastructure (such as traffic cameras, bridge sensors, water supply network sensors, etc.). Sensor data is transmitted in real-time to message queues (such as Kafka) via IoT protocols (such as MQTT and CoAP). Stream processing engines (such as Apache Flink) are then used to clean and format the data, extracting key information (such as sensor location and monitoring values).

[0030] Human activity density heatmaps refer to human activity data from mobile devices (such as mobile phone location data and social media check-in data). Data is obtained from data providers (such as mobile operators and social media platforms) via API interfaces (such as RESTful APIs) and pushed to a message queue. Geospatial processing tools (such as GDAL and GEOS) are used to convert the data into a geographic coordinate format (such as GeoJSON) and generate a heatmap.

[0031] The environmental parameter time-series variation layer refers to real-time environmental data (such as temperature, humidity, and air quality index) from meteorological stations and environmental monitoring stations. Data is obtained from monitoring stations via protocols such as FTP and HTTP and pushed to a message queue. Time series processing tools (such as Pandas) are used to perform time alignment and interpolation on the data to generate the time-series variation layer. By integrating infrastructure sensor data, human activity density heatmaps, and environmental parameter time-series variation layers, multi-source data fusion is achieved, providing a more comprehensive view of urban operational status and offering richer data support for urban planning and management.

[0032] Step S200: Based on the multi-source dynamic layer data stream, a spatiotemporal causal knowledge graph is constructed using spatial elements as nodes and association rules as directed edges.

[0033] Specifically, GIS software (such as ArcGIS and QGIS) or geospatial processing libraries (such as GDAL and GEOS) are used to extract spatial features (such as geographic coordinates and geographic region boundaries) from layer data. Spatial information is extracted from the data using parsing tools (such as Python's geopandas library) and converted to a unified coordinate system (such as WGS84). Machine learning algorithms (such as the association rule mining algorithms Apriori and FP-Growth) are used to mine association rules between spatial features from multi-source data. These algorithms can discover frequent patterns and causal relationships in the data. Time series analysis methods (such as autocorrelation analysis and cross-correlation analysis) are combined to mine spatiotemporal associations between spatial features. A spatiotemporal causal knowledge graph is constructed using graph databases (such as Neo4j and OrientDB), storing spatial features as nodes and association rules as directed edges in the graph database. The spatiotemporal causal knowledge graph is used to represent the spatiotemporal causal relationships between spatial features. A dynamic update mechanism for the graph is designed to ensure that new data can be updated to the knowledge graph in a timely manner, maintaining the graph's timeliness.

[0034] For example, spatial elements such as temperature and humidity are extracted from meteorological data, and spatial elements such as air quality index are extracted from environmental monitoring data. The Apriori algorithm is used to analyze the association rules between these elements, such as "when the temperature is above 30°C and the humidity is below 30%, the air quality index will decrease." These elements and rules are stored in the Neo4j graph database to form a spatiotemporal causal knowledge graph. When new data is received, the nodes and edges in the graph are updated in real time through a stream processing engine.

[0035] In one possible implementation, based on the multi-source dynamic layer data stream, a spatiotemporal causal knowledge graph is constructed using spatial features as nodes and association rules as directed edges. Step S200 further includes step S210, extracting static spatial features and their corresponding attribute information from the multi-source dynamic layer data stream to construct static nodes for the knowledge graph. Specifically, GIS tools (such as ArcGIS, QGIS) or geospatial processing libraries (such as GDAL, GEOS) are used to extract static spatial features, such as buildings, roads, and rivers, from the multi-source dynamic layer data stream. Attribute information of these features is extracted, such as building height, road type, and river flow. The extracted static spatial features and their attribute information are stored as static nodes of the knowledge graph in a graph database (such as Neo4j). Each static node contains spatial location information (such as latitude and longitude coordinates) and attribute information (such as type, height, and length). For example, building and road information is extracted from an urban planning layer. The geopandas library in Python is used to read the layer data in GeoJSON format, extracting the coordinates, height, and purpose of buildings, as well as the coordinates, type, and length of roads. This information is stored as static nodes in the Neo4j graph database.

[0036] Step S220: Extract dynamic spatial elements and their corresponding spatiotemporal variation features from the multi-source dynamic layer data stream to construct dynamic nodes for the knowledge graph. Specifically, use GIS tools or geospatial processing libraries to extract dynamic spatial elements from the multi-source dynamic layer data stream, such as traffic flow, air quality, and population movement. Extract the spatiotemporal variation features of these elements, such as time series data and rates of change. Store the extracted dynamic spatial elements and their spatiotemporal variation features as dynamic nodes in the knowledge graph database. Each dynamic node contains spatial location information, time series data, and variation features. For example, obtain real-time traffic flow data from traffic cameras and air quality data from environmental monitoring stations. Use Pandas to process the time series data and extract the rate of change of traffic flow and the time series variation of air quality. Store these dynamic elements and their features as dynamic nodes in the Neo4j graph database.

[0037] Step S230: Based on the interaction relationship between the static spatial elements and the dynamic spatial elements, association rules are constructed, and directed edges of the knowledge graph are established based on the association rules. Specifically, machine learning algorithms (such as Apriori and FP-Growth) are used to mine the association rules between static and dynamic spatial elements. The interaction relationships between elements are analyzed (such as the relationship between traffic flow and road type, and the relationship between air quality and building height). Based on the mined association rules, directed edges of the knowledge graph are established. Each directed edge represents a causal relationship or interaction relationship, including the relationship type and weight.

[0038] For example, analyzing the relationship between traffic flow and road type using the Apriori algorithm reveals that "traffic flow on main roads is usually higher than on secondary roads." This relationship is stored as directed edges in the Neo4j graph database, with the edge weights determined based on the support and confidence of the association rule.

[0039] Step S240: Based on the static nodes, dynamic nodes, and directed edges, a spatiotemporal causal knowledge graph is constructed. Specifically, a graph database (such as Neo4j) is used to integrate the static nodes, dynamic nodes, and directed edges into a unified spatiotemporal causal knowledge graph. Graph algorithms (such as pathfinding and community detection) are used to optimize the graph structure, improving query efficiency and analytical capabilities.

[0040] For example, static nodes (buildings, roads) and dynamic nodes (traffic flow, air quality), along with directed edges (association rules), are integrated into the Neo4j graph database. Graph queries and analyses are performed using Neo4j's Cypher query language, such as querying "which main roads have traffic flow exceeding 1000 vehicles / hour during peak hours." By extracting static and dynamic spatial elements, their attributes, and spatiotemporal variation characteristics, a spatiotemporal causal knowledge graph containing both static and dynamic information is constructed. This fusion provides a more comprehensive view of urban operational status, offering richer data support for urban planning and management.

[0041] In one possible implementation, association rules are constructed based on the interaction between the static and dynamic spatial elements. Directed edges of the knowledge graph are then established based on these association rules. Step S230 further includes step S231: if the spatial co-occurrence probability of any two static spatial elements is greater than a first preset threshold, a first dynamic association rule is generated, and a first directed edge of the knowledge graph is established based on the first dynamic association rule. Specifically, spatial statistical methods (such as spatial autocorrelation analysis and spatial clustering analysis) are used to calculate the spatial co-occurrence probability of any two static spatial elements. For example, the probability of two buildings appearing in the same area, or the probability of two roads appearing in the same traffic network, is calculated. A first preset threshold is set (e.g., 0.8). If the spatial co-occurrence probability is greater than this threshold, a significant spatial association is considered to exist between the two elements, and a first dynamic association rule is generated, indicating a strong spatial association between the two static spatial elements. In the knowledge graph, a first directed edge is established based on this association rule, representing a spatial association relationship from one element to another.

[0042] For example, suppose in urban planning, two buildings (A and B) frequently appear in the same area. Spatial clustering analysis calculates their spatial co-occurrence probability to be 0.85, which is greater than a first preset threshold of 0.8. Therefore, a first dynamic association rule is generated, indicating a strong spatial association between building A and building B. In the knowledge graph, a first directed edge is established from building A to building B, representing this spatial association.

[0043] Step S232: If the time lag correlation between any two dynamic spatial elements is greater than a second preset threshold, a second dynamic association rule is generated, and a second directed edge of the knowledge graph is established based on the second dynamic association rule. Specifically, time series analysis methods (such as cross-correlation analysis and Granger causality test) are used to calculate the time lag correlation between any two dynamic spatial elements. For example, the time lag correlation between traffic flow and air quality index is calculated to analyze the delayed impact of traffic flow changes on air quality. A second preset threshold (e.g., 0.7) is set. If the time lag correlation is greater than this threshold, it is considered that there is a significant time correlation between the two elements, and a second dynamic association rule is generated, indicating that there is a strong time correlation between the two dynamic spatial elements. In the knowledge graph, a second directed edge is established based on this association rule, representing the time correlation relationship from one element to another.

[0044] For example, suppose time-series analysis of traffic flow and air quality data shows that an increase in traffic flow leads to a decrease in the air quality index after 2 hours, with a time lag correlation of 0.75, which is greater than the second preset threshold of 0.7. Therefore, a second dynamic association rule is generated to indicate a strong temporal correlation between traffic flow and air quality. In the knowledge graph, a second directed edge is established from traffic flow to air quality to represent this temporal correlation.

[0045] Step S233: If the dependency correlation between any static spatial element and any dynamic spatial element is greater than a third preset threshold, a third dynamic association rule is generated, and a third directed edge of the knowledge graph is established based on the third dynamic association rule. Specifically, statistical methods (such as Pearson correlation coefficient, mutual information) are used to calculate the dependency correlation between any static spatial element and any dynamic spatial element. For example, the dependency correlation between building height and surrounding air quality is calculated to analyze the impact of building height on air quality. A third preset threshold is set (e.g., 0.6). If the dependency correlation is greater than this threshold, it is considered that there is a significant dependency relationship between the two elements, and a third dynamic association rule is generated, indicating that there is a strong dependency relationship between the static spatial element and the dynamic spatial element. In the knowledge graph, a third directed edge is established based on this association rule, representing the dependency relationship from the static element to the dynamic element.

[0046] For example, assuming the analysis reveals a dependency between building height and surrounding air quality, with a dependency correlation of 0.65, which is greater than the third preset threshold of 0.6, a third dynamic association rule is generated, indicating that building height has a significant impact on air quality. In the knowledge graph, a third directed edge is established from building height to air quality to represent this dependency. By calculating spatial co-occurrence probability, temporal lag correlation, and dependency correlation separately, the association rules between static and dynamic spatial elements can be mined more precisely. This refined mining method can more accurately reflect the complex relationships between elements.

[0047] Step S300: Perform multi-layer joint simulation based on the spatiotemporal causal knowledge graph to generate a prediction topic layer containing time series confidence intervals.

[0048] Specifically, GIS simulation tools (such as ArcGIS Pro and QGIS simulation plugins) are used in conjunction with spatiotemporal causal knowledge graphs in map databases to jointly simulate the changing processes of spatial phenomena using multi-source layer data. Model fusion techniques (such as multi-model integration and model calibration) are employed to integrate model results from different sources, improving simulation accuracy. Time series prediction models (such as ARIMA and LSTM neural networks) are used to predict simulation results and calculate confidence intervals. These models can handle the dynamic changes in time series data and provide estimates of prediction uncertainty. Historical data is used to train the models, and their performance is evaluated through methods such as cross-validation to ensure the reliability of the prediction results. The prediction results and confidence intervals are visualized as thematic layers, and thematic maps are generated using GIS visualization tools (such as ArcMap and QGIS). Layer rendering techniques (such as symbolization and hierarchical rendering) are used to display the prediction results intuitively on the map, facilitating user understanding and use.

[0049] For example, by combining the spatiotemporal causal relationships between meteorological and environmental monitoring data, an LSTM neural network is used to perform time-series predictions of the air quality index. The prediction results show the trend of air quality index changes over the next 24 hours and calculate a 95% confidence interval. These prediction results and confidence intervals are then generated as thematic layers and displayed in QGIS, allowing users to visually see the prediction results and the range of uncertainty on a map.

[0050] In one possible implementation, multi-layer joint simulation is performed based on the spatiotemporal causal knowledge graph to generate a prediction topic layer containing time series confidence intervals. Step S300 further includes step S310, which performs time discretization processing on the spatiotemporal causal knowledge graph to generate a time propagation sequence. Specifically, continuous time series data is discretized into multiple time steps (e.g., per hour, per minute). For example, a day is discretized into 24 time steps. Time series data is divided into multiple time segments using a time window technique (e.g., a sliding window), with each segment corresponding to a time step. Based on the discretized time steps, a time propagation sequence is generated, representing the propagation path of node states within each time step. The time propagation sequence can be represented as a time series graph, where the states of each node at each time step are connected by directed edges.

[0051] For example, suppose there is a spatiotemporal causal knowledge graph containing buildings (static nodes) and air quality (dynamic nodes). The time of day is discretized into 24 time steps (one step per hour). For each time step, the state changes of the air quality node are recorded, generating a time propagation sequence. For example, the time propagation sequence can be represented as: (t0, air quality 0) → (t1, air quality 1) → … → (t 23 air quality 23 ).

[0052] Step S320: Based on the time propagation sequence, starting from the state of each node at the current moment, a state transfer simulation is performed along the directed edges to obtain the first prediction thematic layer. Specifically, starting from the state of each node at the current moment, the transfer of node states is simulated according to the causal relationship of the time propagation sequence and directed edges. A state transition model (such as a Markov model) or a rule-based model (such as an association rule-based inference engine) is used to perform the state transfer simulation. The simulation results are visualized as the first prediction thematic layer, representing the predicted value of the node state within each time step.

[0053] For example, assuming the current air quality is "good," the simulation aims to model changes in air quality over the next 24 hours based on time-propagation sequences and causal relationships (such as increased traffic leading to decreased air quality). A Markov model is used for state-transition simulation, generating a first predictive thematic layer. This layer displays the predicted air quality values ​​at each time step; for example, it predicts that air quality will decrease during peak traffic hours.

[0054] Step S330: Based on the first prediction thematic layer, perform a simulation with random noise added to the dynamic nodes to obtain a second prediction thematic layer. Specifically, based on the first prediction thematic layer, random noise is added to the state of the dynamic nodes to simulate uncertainties in the real world. A random noise model (such as Gaussian noise or Poisson noise) is used to add noise to the state of the dynamic nodes. The simulation results with added noise are then updated to the second prediction thematic layer, representing the prediction results considering uncertainties.

[0055] For example, suppose the first forecast thematic layer shows the changes in air quality over the next 24 hours. To simulate uncertainty, Gaussian noise is added to the predicted air quality values. For instance, if the predicted value is "good," adding noise might change it to "good ± 5%." This generates a second forecast thematic layer that shows the air quality forecast after accounting for uncertainty.

[0056] Step S340: Add confidence intervals to the second prediction thematic layer to obtain a prediction thematic layer containing time series confidence intervals. Specifically, use statistical methods (such as bootstrapping or Bayesian methods) to calculate the confidence intervals for the dynamic node states within each time step. The confidence intervals represent the range of uncertainty of the predicted values. Add the confidence intervals to the second prediction thematic layer to generate the final prediction thematic layer, representing the predicted values ​​of the node states and their range of uncertainty within each time step.

[0057] For example, suppose the second forecast thematic layer displays the predicted air quality values ​​and their uncertainties for the next 24 hours. A bootstrap method is used to calculate the 95% confidence interval for the predicted air quality values ​​at each time step. For example, if the predicted value is "Good," the confidence interval is "Good ± 10%." The final forecast thematic layer is generated, displaying the predicted air quality values ​​and their confidence intervals at each time step. Through time discretization and state transfer simulation, the dynamic changes in the states of nodes in the spatiotemporal causal knowledge graph can be simulated with fine detail. This refined simulation method can more accurately reflect the complex dynamic processes in the real world. By adding random noise to dynamic nodes and calculating confidence intervals, the uncertainty of the modeled prediction results can be effectively quantified and described. This uncertainty modeling method can provide more reliable prediction results, helping decision-makers better assess risks.

[0058] In one possible implementation, confidence intervals are added to the second predictive thematic layer to obtain a predictive thematic layer containing time-series confidence intervals. Step S340 further includes step S341, obtaining the basic number of simulations N, where N ≥ 1000. Specifically, the basic number of simulations N is set according to actual needs and computing resources. N should be large enough to ensure the reliability of the statistical results. For example, in an environmental prediction system, N = 1000 simulations are set to ensure the stability and reliability of the statistical results.

[0059] Step S342: Perform multi-layer joint simulation according to the basic number of simulations N to obtain N second prediction thematic layers. Specifically, use GIS simulation tools (such as the simulation plugin of ArcGIS Pro or QGIS) combined with a spatiotemporal causal knowledge graph to perform multi-layer joint simulation. Each simulation generates one second prediction thematic layer, and a total of N simulations are performed to obtain N second prediction thematic layers.

[0060] For example, suppose there are two dynamic layers: traffic flow and air quality. Using a GIS simulation tool, 1000 simulations are performed, each simulation generating a thematic layer containing traffic flow and air quality predictions, ultimately resulting in 1000 second prediction thematic layers.

[0061] Step S343: For the spatiotemporal state values ​​of each dynamic spatial element in the N second prediction thematic layers, calculate the probability distribution function using Gaussian kernel density. Specifically, use the Gaussian kernel density estimation method to calculate the probability distribution function of the spatiotemporal state value of each dynamic spatial element. Gaussian kernel density estimation is a non-parametric estimation method that can smooth the probability distribution of the estimated data.

[0062] For example, for each dynamic spatial feature (such as the air quality index), its spatiotemporal state value is extracted from 1000 second-prediction thematic layers, and its probability distribution function is calculated using Python's scipy.stats.gaussian_kde function.

[0063] Step S344: For non-critical areas, the 5% to 95th percentiles of the probability distribution function are used as the 90% confidence interval; for critical areas, the 10% to 90th percentiles of the probability distribution function are used as the 80% confidence interval. Specifically, quantiles are calculated based on the probability distribution function. For non-critical areas, the 5% and 95th percentiles are calculated as the 90% confidence interval; for critical areas, the 10% and 90th percentiles are calculated as the 80% confidence interval. The division between critical and non-critical areas is defined according to the needs of the actual application scenario. For example, a critical area can be a densely populated area or an environmentally sensitive area.

[0064] For example, suppose the city center is a critical area and other areas are non-critical areas. For the air quality index, calculate the 90% confidence interval (5%–95th quantile) for non-critical areas and the 80% confidence interval (10%–90th quantile) for critical areas.

[0065] Step S345: The spatiotemporal state values ​​of each dynamic spatial element in the N second prediction thematic layers are averaged to obtain a second mean prediction thematic layer. Specifically, the average spatiotemporal state values ​​of each dynamic spatial element in the N second prediction thematic layers are averaged to obtain the average predicted value for each element. These average predicted values ​​are then visualized as a second mean prediction thematic layer.

[0066] For example, for the Air Quality Index, the average value of each time step across 1000 second-means prediction thematic layers is calculated to generate a second-means prediction thematic layer. This layer displays the average predicted air quality value within each time step.

[0067] Step S346: Add the confidence interval to the second mean prediction thematic layer to obtain a prediction thematic layer containing time series confidence intervals. Specifically, the confidence interval calculated in step S344 is added to the second mean prediction thematic layer to generate the final prediction thematic layer. This layer contains not only the mean of the predicted values ​​but also the confidence interval of the predicted values ​​within each time step.

[0068] For example, based on the second mean prediction thematic layer, confidence intervals for the predicted air quality values ​​within each time step are added. For non-critical areas, 90% confidence intervals are added; for critical areas, 80% confidence intervals are added. The final generated thematic prediction layer displays the average predicted air quality value and its confidence interval for each time step. Through multiple simulations (N≥1000) and statistical analysis, the generated thematic prediction layer provides more reliable prediction results. This method effectively reduces the random errors introduced by a single simulation.

[0069] Step S400: Calculate the dynamic time warping distance between the predicted thematic layer and the real-time monitoring layer. When the dynamic time warping distance is greater than the adaptive distance threshold, generate a spatial coordinate binding difference warning signal.

[0070] Specifically, the Dynamic Time Warping (DTW) algorithm is used to calculate the similarity between the time series of the predicted thematic layer and the real-time monitoring layer. The DTW algorithm can handle the scaling and deformation of time series and is suitable for comparing non-linear time series. The difference between the predicted results and the actual monitoring data is quantified by calculating the DTW distance. An adaptive distance threshold is dynamically calculated through statistical analysis (such as moving average and standard deviation) to determine whether the difference between the time series is significant. The adaptive threshold can be automatically adjusted according to the statistical characteristics of the data, improving the robustness of the system. A dynamic update mechanism for the threshold is designed to ensure that the threshold can be adjusted according to the latest data and adapt to data changes. When the DTW distance is greater than the adaptive threshold, a spatial coordinate-bound difference warning signal is generated using GIS spatial analysis tools (such as ArcGIS's spatial analysis module). The specific spatial coordinate location is marked on the GIS map to remind the user that there is a significant difference between the prediction and the actual monitoring in that area.

[0071] For example, the DTW algorithm is used to calculate the similarity between the predicted air quality index time series and the real-time monitored air quality index time series. If the DTW distance exceeds an adaptive threshold (e.g., 1.5 times the standard deviation) calculated by moving average and standard deviation, a difference warning signal is generated on the GIS map, marking the specific spatial coordinates to alert the user that there is a significant difference between the prediction and the actual monitoring in that area.

[0072] This application's embodiments employ techniques such as first accessing high-frequency updated multi-source dynamic layer data streams in real time, then constructing a spatiotemporal causal knowledge graph using spatial elements as nodes and association rules as directed edges, then conducting multi-layer joint simulation based on this graph to generate a prediction thematic layer containing time series confidence intervals, and finally calculating the dynamic time warping distance between the prediction thematic layer and the real-time monitoring layer. When the distance exceeds an adaptive threshold, a spatial coordinate binding difference warning signal is generated. These techniques achieve the technical effect of improving the accuracy and reliability of prediction.

[0073] In one possible implementation, the dynamic time-warped distance between the predicted thematic layer and the real-time monitoring layer is calculated. Step S400 further includes step S410, which performs time-series alignment preprocessing on the predicted thematic layer and the real-time monitoring layer to obtain an aligned layer pair. Specifically, a time alignment algorithm (such as timestamp alignment or interpolation alignment) is used to align the time series of the predicted thematic layer and the real-time monitoring layer. This ensures that the time steps of the two layers are consistent for subsequent dynamic warping calculations.

[0074] For example, suppose the prediction thematic layer has a time step of once per hour, while the real-time monitoring layer has a time step of once every 30 minutes. Through interpolation alignment, the time step of the real-time monitoring layer is adjusted to once per hour, ensuring that the time steps of the two layers are consistent. The final result is an aligned layer pair, with each layer having a time step of once per hour.

[0075] Step S420: Perform multi-scale dynamic warping on the aligned layer pairs to obtain warped layer pairs. Specifically, use a dynamic time warping algorithm to perform multi-scale dynamic warping on the aligned layer pairs. The DTW algorithm can handle the scaling and deformation of time series and is suitable for comparing nonlinear time series. Multi-scale dynamic warping can be performed at different time scales (such as hours, days, weeks) to capture changes at different time scales.

[0076] For example, the DTW algorithm is used for multi-scale dynamic normalization of the aligned prediction thematic layer and real-time monitoring layer. At the hourly scale, the DTW distance between the two layers is calculated; at the daily scale, the time series are aggregated by day, and the DTW distance is calculated again. The final result is a pair of normalized layers, each of which has undergone multi-scale dynamic normalization.

[0077] Step S430: Calculate the cosine similarity and root mean square error of the regularized layer pairs. Specifically, the similarity of the regularized layer pairs is calculated using the cosine similarity formula. Cosine similarity measures the angle between two vectors; the closer the value is to 1, the more similar they are. The cosine similarity formula is: Cosine... in, This represents the dot product (inner product) of vectors x and y. The dot product is the sum of the product of corresponding elements of the two vectors. i and y i Let x and y represent the elements at the i-th position of vector x and vector y, respectively, and n represent the dimension of the vectors (i.e., the number of elements). Let x represent the modulus (length) of vector x. The modulus is the square root of the sum of the squares of all the elements of the vector. This represents the magnitude (length) of vector y.

[0078] The root mean square error (RMSE) formula is used to calculate the error between normalized layer pairs. RMSE measures the difference between predicted and actual values; a smaller value indicates a smaller error. The RMSE formula is as follows: Where m represents the length of the time series data, i.e., the number of data points, and p j Let q represent the j-th data point in the predicted value sequence. j This represents the j-th data point in the actual value sequence.

[0079] For example, for the normalized predicted thematic layer and the real-time monitoring layer, cosine similarity and root mean square error are calculated. Assume the time series of the predicted layer is [10, 12, 15, 18], and the time series of the actual monitoring layer is [11, 13, 16, 19]. The calculated cosine similarity is 0.98, and the root mean square error is 1.

[0080] Step S440: The cosine similarity and root mean square error are weighted and calculated to obtain the dynamic time warped distance. Specifically, based on actual needs, weights are assigned to the cosine similarity and root mean square error, and the dynamic time warped distance is calculated using a weighted method. The formula for the dynamic time warped distance is: DTW Distance=α·(1-Cosine Similarity)+β·RMSE, where α and β are the weights of the cosine similarity and root mean square error, respectively, and α+β=1.

[0081] For example, assuming weights α = 0.6 and β = 0.4, based on the calculation results of step S430, the cosine similarity is 0.98 and the root mean square error is 1. The dynamic time-normalized distance (DTW Distance) is calculated as: DTW Distance = 0.6 × (1 - 0.98) + 0.4 × 1 = 0.412. By calculating cosine similarity and root mean square error, the similarity and error between the predicted thematic layer and the real-time monitoring layer can be evaluated in a more refined manner. Cosine similarity measures directional similarity, while root mean square error measures numerical differences; the combination of the two provides a more comprehensive evaluation.

[0082] In one possible implementation, the aligned layer pairs are dynamically warped at multiple scales to obtain warped layer pairs. Step S420 further includes step S421, performing coarse-grained alignment on the aligned layer pairs to obtain coarse-grained aligned layer pairs. Specifically, a dynamic time warping algorithm is used to perform coarse-grained alignment on the aligned layer pairs. In the coarse-grained alignment stage, the focus is mainly on the overall shape and trend of the time series, ignoring local details. This can be achieved through a large time window or a low time resolution.

[0083] For example, suppose there are two time series layers: a prediction layer c = [10, 12, 15, 18] and an actual monitoring layer d = [11, 13, 16, 19]. Using the DTW algorithm for coarse-grained alignment, the time series are aggregated by larger time windows (e.g., every two time steps), resulting in coarse-grained aligned layer pairs. For example, the aggregated time series would be: c ′ = [11, 16.5] and d ′ = [12, 17.5].

[0084] Step S422 involves fine-grained optimization of the coarse-grained aligned layer pairs to obtain regularized layer pairs. Specifically, based on the coarse-grained alignment, the alignment accuracy of the time series is further optimized. This can be achieved through a smaller time window or a higher time resolution. Local optimization algorithms (such as local adjustment and smoothing) are used to refine the coarse-grained alignment results, ensuring that the time series are aligned as much as possible in local details.

[0085] For example, based on coarse-grained alignment, the aggregated time series c ′ = [11, 16.5] and d ′ Fine-grained optimization is performed on the time series at [12, 17.5]. A local adjustment algorithm is used to fine-tune each time step, ensuring that the time series are aligned as closely as possible even in local details. For example, the optimized results are c″=[11, 16.5] and d″=[11.5, 17]. Coarse-grained alignment can quickly capture the overall similarity of the time series, while fine-grained optimization can further improve the alignment accuracy. This staged alignment method can effectively improve the accuracy and reliability of time series alignment.

[0086] In one possible implementation, when the dynamic time warp distance exceeds an adaptive distance threshold, a spatial coordinate binding difference warning signal is generated. Step S400 further includes step S450, obtaining historical dynamic time warp distances within a preset time range, and using the 90th quantile of the historical dynamic time warp distances as a baseline distance threshold. Specifically, dynamic time warp distances within a preset time range are obtained from a historical database. The preset time range can be the past month, three months, or one year, depending on the application scenario requirements. The 90th quantile of the historical dynamic time warp distances is calculated and used as the baseline distance threshold. The 90th quantile indicates that in the historical data, the dynamic time warp distance of 90% of the data points is less than or equal to this value.

[0087] For example, suppose we obtain dynamic time-normalized distance data from the past three months, and the 90th percentile of this data is calculated to be 1.5. Therefore, the baseline distance threshold is set to 1.5.

[0088] Step S460: Calculate the quality score of the current data, and adaptively adjust the baseline distance threshold based on the quality score to obtain an adaptive distance threshold. Specifically, calculate the quality score of the current data. The quality score is evaluated based on indicators such as data completeness, accuracy, and consistency. Adaptively adjust the baseline distance threshold based on the quality score of the current data. If the data quality is high, appropriately lower the threshold; if the data quality is low, appropriately increase the threshold to ensure the reasonableness and reliability of the threshold.

[0089] For example, suppose the current data quality score is 0.8 (out of 1), indicating high data quality. Based on the quality score, the baseline distance threshold is adaptively adjusted from 1.5 to 1.2. If the quality score is low, such as 0.4, the baseline distance threshold is adjusted to 1.8.

[0090] Step S470: When the dynamic time warp distance is greater than the adaptive distance threshold, a multi-level difference warning signal bound to spatial coordinates is generated based on the distance difference, and the difference area is marked on the map. Specifically, when the dynamic time warp distance is greater than the adaptive distance threshold, a multi-level difference warning signal is generated based on the distance difference. The multi-level warning signal is divided into different levels according to the magnitude of the distance difference, such as low, medium, and high. The difference warning signal is bound to specific spatial coordinates to ensure accurate location of the difference area. The difference area is marked on the map, using different colors or symbols to represent different levels of warning signals, making it convenient for users to intuitively view the difference area.

[0091] For example, suppose the dynamic time-warped distance is 1.6, which is greater than the adaptive distance threshold of 1.2. Based on the distance difference (1.6 - 1.2 = 0.4), a medium-level difference warning signal is generated. This signal is then associated with specific spatial coordinates (such as latitude and longitude), and the medium-level warning area is marked in yellow on the map. If the distance difference is larger, such as 0.8, a high-level warning signal is generated and marked in red.

[0092] By calculating the 90th percentile of historical data as a baseline distance threshold and adaptively adjusting it based on the quality score of the current data, the rationality and reliability of the threshold can be ensured. This method can adapt to different data quality conditions, improving the robustness of the system. Generating multi-level difference warning signals and binding them with spatial coordinates can more precisely reflect the difference between prediction and actual monitoring. Multi-level warning signals can provide different response measures according to the severity of the difference, enhancing the system's early warning capability.

[0093] In the above text, refer to Figure 1 This paper describes in detail a GIS layer analysis method based on automatic map merging and parsing according to embodiments of the present invention. Next, reference will be made to... Figure 2 This invention describes a GIS layer analysis system based on automatic map merging and parsing according to an embodiment of the present invention.

[0094] The GIS layer analysis system based on automatic map merging and parsing according to embodiments of the present invention addresses the technical problems of insufficient prediction accuracy and reliability in existing technologies, thereby improving the accuracy and reliability of predictions. The GIS layer analysis system based on automatic map merging and parsing includes: a multi-source dynamic layer data stream access module 10, a spatiotemporal causal knowledge graph construction module 20, a multi-layer joint simulation module 30, and a difference warning signal generation module 40.

[0095] The multi-source dynamic layer data stream access module 10 is used to access multi-source dynamic layer data streams updated at a frequency greater than a preset frequency in real time; the spatiotemporal causal knowledge graph construction module 20 is used to construct a spatiotemporal causal knowledge graph based on the multi-source dynamic layer data stream, using spatial elements as nodes and association rules as directed edges; the multi-layer joint simulation module 30 is used to perform multi-layer joint simulation based on the spatiotemporal causal knowledge graph to generate a prediction thematic layer containing time series confidence intervals; the difference warning signal generation module 40 is used to calculate the dynamic time warping distance between the prediction thematic layer and the real-time monitoring layer, and generate a spatial coordinate-bound difference warning signal when the dynamic time warping distance is greater than the adaptive distance threshold.

[0096] The specific configuration of the spatiotemporal causal knowledge graph construction module 20 will be described in detail below. As mentioned above, based on the multi-source dynamic layer data stream, using spatial elements as nodes and association rules as directed edges, a spatiotemporal causal knowledge graph is constructed. The spatiotemporal causal knowledge graph construction module 20 may further include: a static node construction unit for extracting static spatial elements and corresponding attribute information from the multi-source dynamic layer data stream to construct static nodes of the knowledge graph; a dynamic node construction unit for extracting dynamic spatial elements and corresponding spatiotemporal change features from the multi-source dynamic layer data stream to construct dynamic nodes of the knowledge graph; a directed edge establishment unit for constructing association rules based on the interaction relationship between the static spatial elements and the dynamic spatial elements, and establishing directed edges of the knowledge graph based on the association rules; and a spatiotemporal causal knowledge graph construction unit for constructing the spatiotemporal causal knowledge graph based on the static nodes, the dynamic nodes, and the directed edges.

[0097] Specifically, based on the interaction relationship between the static spatial elements and the dynamic spatial elements, association rules are constructed, and directed edges of the knowledge graph are established based on the association rules. The directed edge establishment unit may further include: a first directed edge establishment subunit used to generate a first dynamic association rule if the spatial co-occurrence probability of any two static spatial elements is greater than a first preset threshold, and to establish a first directed edge of the knowledge graph based on the first dynamic association rule; a second directed edge establishment subunit used to generate a second dynamic association rule if the time lag correlation of any two dynamic spatial elements is greater than a second preset threshold, and to establish a second directed edge of the knowledge graph based on the second dynamic association rule; and a third directed edge establishment subunit used to generate a third dynamic association rule if the dependency correlation between any static spatial element and any dynamic spatial element is greater than a third preset threshold, and to establish a third directed edge of the knowledge graph based on the third dynamic association rule.

[0098] The specific configuration of the multi-layer co-simulation module 30 will be described in detail below. As mentioned above, multi-layer co-simulation is performed based on the spatiotemporal causal knowledge graph to generate a prediction topic layer containing time series confidence intervals. The multi-layer co-simulation module 30 may further include: a time discretization processing unit for performing time discretization processing on the spatiotemporal causal knowledge graph to generate a time propagation sequence; a state transfer simulation unit for performing state transfer simulation based on the time propagation sequence, starting from the state of each node at the current time, and performing state transfer simulation along directed edges to obtain a first prediction topic layer; a random noise addition simulation unit for performing random noise addition simulation on dynamic nodes based on the first prediction topic layer to obtain a second prediction topic layer; and a confidence interval addition unit for adding confidence intervals to the second prediction topic layer to obtain a prediction topic layer containing time series confidence intervals.

[0099] Specifically, a confidence interval is added to the second predicted thematic layer to obtain a predicted thematic layer containing time series confidence intervals. The confidence interval adding unit may further include: a simulation base number acquisition subunit for acquiring the simulation base number N, where N≥1000; a multi-layer joint simulation subunit for performing multi-layer joint simulation according to the simulation base number N to obtain N second predicted thematic layers; and a probability distribution function calculation subunit for calculating the probability distribution of the spatiotemporal state values ​​of each dynamic spatial element in the N second predicted thematic layers using Gaussian kernel density. The function; the confidence interval determination subunit is used to take the 5% to 95th percentile of the probability distribution function as the 90% confidence interval for non-critical areas, and the 10% to 90th percentile of the probability distribution function as the 80% confidence interval for critical areas; the mean calculation subunit is used to calculate the mean of the spatiotemporal state values ​​of each dynamic spatial element in the N second prediction thematic layers to obtain the second mean prediction thematic layer; the confidence interval addition subunit is used to add the confidence interval to the second mean prediction thematic layer to obtain a prediction thematic layer containing the time series confidence interval.

[0100] The specific configuration of the difference warning signal generation module 40 will be described in detail below. As mentioned above, to calculate the dynamic time-normalized distance between the predicted thematic layer and the real-time monitoring layer, the difference warning signal generation module 40 may further include: a temporal alignment preprocessing unit for performing temporal alignment preprocessing on the predicted thematic layer and the real-time monitoring layer to obtain an aligned layer pair; a multi-scale dynamic normalization unit for performing multi-scale dynamic normalization on the aligned layer pair to obtain a normalized layer pair; a calculation unit for calculating the cosine similarity and root mean square error of the normalized layer pair; and a weighting unit for weighting the calculation of the cosine similarity and root mean square error to obtain the dynamic time-normalized distance.

[0101] The alignment layer pair is dynamically normalized at multiple scales to obtain a normalized layer pair. The multi-scale dynamic normalization unit may further include: a coarse-grained alignment subunit for performing coarse-grained alignment on the alignment layer pair to obtain a coarse-grained alignment layer pair; and a fine-grained optimization subunit for performing fine-grained optimization on the coarse-grained alignment layer pair to obtain a normalized layer pair.

[0102] Specifically, when the dynamic time warp distance is greater than the adaptive distance threshold, a spatial coordinate-bound difference warning signal is generated. The difference warning signal generation module 40 may further include: a baseline distance threshold acquisition unit for acquiring historical dynamic time warp distances within a preset time range, and using the 90th percentile of the historical dynamic time warp distance as the baseline distance threshold; an adaptive distance threshold acquisition unit for calculating the quality score of the current data, and adaptively adjusting the baseline distance threshold according to the quality score to obtain an adaptive distance threshold; and a multi-level difference warning signal generation unit for generating a spatial coordinate-bound multi-level difference warning signal based on the distance difference when the dynamic time warp distance is greater than the adaptive distance threshold, and marking the difference area on the map.

[0103] The specific configuration of the multi-source dynamic layer data stream access module 10 will be described in detail below. As mentioned above, the multi-source dynamic layer data stream access module 10 may further include: a multi-source dynamic layer data stream preset unit for presetting multi-source dynamic layer data streams, which include: infrastructure sensor data, human activity density heatmaps, and environmental parameter time-series change layers.

[0104] The GIS layer analysis system based on automatic map merging and parsing provided in this embodiment of the invention can execute the GIS layer analysis method based on automatic map merging and parsing provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0105] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0106] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A GIS layer analysis method based on automatic map merging and parsing, characterized in that, The method includes: Real-time access to multi-source dynamic layer data streams that update at a frequency greater than the preset frequency; Based on the multi-source dynamic layer data stream, a spatiotemporal causal knowledge graph is constructed, with spatial elements as nodes and association rules as directed edges. Multi-layer joint simulation is performed based on the spatiotemporal causal knowledge graph to generate a prediction topic layer containing time series confidence intervals; Calculate the dynamic time warping distance between the predicted thematic layer and the real-time monitoring layer, and generate a spatial coordinate binding difference warning signal when the dynamic time warping distance is greater than the adaptive distance threshold; Specifically, multi-layer joint simulation is performed based on the spatiotemporal causal knowledge graph to generate a prediction topic layer containing time series confidence intervals, including: The spatiotemporal causal knowledge graph is subjected to time discretization to generate a time propagation sequence; Based on the time propagation sequence, starting from the state of each node at the current moment, state propagation simulation is performed along the directed edge to obtain the first predicted thematic layer; Based on the first predicted thematic layer, random noise is added to the dynamic nodes in a simulation to obtain the second predicted thematic layer. Add confidence intervals to the second predicted thematic layer to obtain a predicted thematic layer containing time series confidence intervals; The calculation of the dynamic time-warped distance between the predicted thematic layer and the real-time monitoring layer includes: Perform time-series alignment preprocessing on the predicted thematic layer and the real-time monitoring layer to obtain aligned layer pairs; The aligned layer pairs are dynamically normalized at multiple scales to obtain normalized layer pairs; Calculate the cosine similarity and root mean square error of the regularized layer pairs; The cosine similarity and root mean square error are weighted and calculated to obtain the dynamic time-warped distance.

2. The GIS layer analysis method based on automatic map merging and parsing as described in claim 1, characterized in that, Based on the aforementioned multi-source dynamic layer data stream, using spatial features as nodes and association rules as directed edges, a spatiotemporal causal knowledge graph is constructed, including: Extract static spatial elements and corresponding attribute information from the multi-source dynamic layer data stream to construct static nodes of the knowledge graph; Extract dynamic spatial elements and corresponding spatiotemporal change features from the multi-source dynamic layer data stream to construct dynamic nodes of the knowledge graph; Based on the interaction relationship between the static spatial elements and the dynamic spatial elements, association rules are constructed, and directed edges of the knowledge graph are established based on the association rules. Based on the static nodes, the dynamic nodes, and the directed edges, a spatiotemporal causal knowledge graph is constructed.

3. The GIS layer analysis method based on automatic map merging and parsing as described in claim 2, characterized in that, Based on the interaction relationship between the static spatial elements and the dynamic spatial elements, association rules are constructed, and directed edges of the knowledge graph are established based on the association rules, including: If the spatial co-occurrence probability of any two static spatial elements is greater than a first preset threshold, a first dynamic association rule is generated, and a first directed edge of the knowledge graph is established based on the first dynamic association rule. If the time lag correlation between any two dynamic spatial elements is greater than the second preset threshold, a second dynamic association rule is generated, and a second directed edge of the knowledge graph is established based on the second dynamic association rule. If the dependency correlation between any static spatial element and any dynamic spatial element is greater than a third preset threshold, a third dynamic association rule is generated, and a third directed edge of the knowledge graph is established based on the third dynamic association rule.

4. The GIS layer analysis method based on automatic map merging and parsing as described in claim 1, characterized in that, Add confidence intervals to the second predicted thematic layer to obtain a predicted thematic layer containing time series confidence intervals, including: Obtain the basic number of simulations, N, where N≥1000; Perform multi-layer joint simulation according to the basic number of simulations N, and obtain N second prediction thematic layers; For the spatiotemporal state values ​​of each dynamic spatial element in the N second prediction thematic layers, the probability distribution function is calculated using Gaussian kernel density; For non-critical areas, the 5% to 95th percentile of the probability distribution function is taken as the 90% confidence interval; for critical areas, the 10% to 90th percentile of the probability distribution function is taken as the 80% confidence interval. The average spatiotemporal state values ​​of each dynamic spatial element in the N second prediction thematic layers are calculated to obtain the second mean prediction thematic layer; Add the confidence interval to the second mean prediction thematic layer to obtain a prediction thematic layer containing time series confidence intervals.

5. The GIS layer analysis method based on automatic map merging and parsing as described in claim 1, characterized in that, Multi-scale dynamic normalization is performed on the aligned layer pairs to obtain normalized layer pairs, including: Perform coarse-grained alignment on the alignment layer pair to obtain a coarse-grained alignment layer pair; Fine-grained optimization is performed on the coarse-grained aligned layer pairs to obtain regularized layer pairs.

6. The GIS layer analysis method based on automatic map merging and parsing as described in claim 1, characterized in that, When the dynamic time warp distance is greater than the adaptive distance threshold, a spatial coordinate binding difference warning signal is generated, including: Obtain the historical dynamic time warping distance within a preset time range, and use the 90th percentile of the historical dynamic time warping distance as the benchmark distance threshold; Calculate the quality score of the current data, and adaptively adjust the baseline distance threshold based on the quality score to obtain the adaptive distance threshold; When the dynamic time warping distance is greater than the adaptive distance threshold, a multi-level difference warning signal bound to spatial coordinates is generated based on the distance difference, and the difference area is marked on the map.

7. The GIS layer analysis method based on automatic map merging and parsing as described in claim 1, characterized in that, The multi-source dynamic layer data stream includes: infrastructure sensor data, human activity density heatmap, and environmental parameter time-series change layers.

8. A GIS layer analysis system based on automatic map merging and parsing, characterized in that, The system is used to implement the GIS layer analysis method based on automatic map merging and parsing as described in any one of claims 1-7, and the system comprises: The multi-source dynamic layer data stream access module is used to access multi-source dynamic layer data streams that are updated at a frequency greater than a preset frequency in real time. The spatiotemporal causal knowledge graph construction module is used to construct a spatiotemporal causal knowledge graph based on the multi-source dynamic layer data stream, using spatial elements as nodes and association rules as directed edges. The multi-layer co-simulation module is used to perform multi-layer co-simulation based on the spatiotemporal causal knowledge graph to generate a prediction topic layer containing time series confidence intervals; The difference warning signal generation module is used to calculate the dynamic time warping distance between the predicted thematic layer and the real-time monitoring layer, and generate a spatial coordinate-bound difference warning signal when the dynamic time warping distance is greater than the adaptive distance threshold.

Images