Geological disaster monitoring data real-time fusion method and system

Abstract

The invention relates to the technical field of data fusion, and discloses a geological disaster monitoring data real-time fusion method and system, and the method comprises the steps: carrying out the heterogeneous adaptation analysis of multi-source geological monitoring data of a target region, and obtaining a standardized data sequence; constructing a dynamic semantic relation graph according to a coupling relation among multiple monitoring parameters in the standardized data sequence; based on a space-time constraint rule of the dynamic semantic relation graph, performing space-time two-dimensional proofreading on the standardized data sequence to obtain a space-time consistent data set; performing vectorization mapping on time-frequency domain features in the time-space consistent data set to obtain a multi-dimensional feature vector set; carrying out collaborative optimization on the real-time weight of the dynamic semantic relation graph and the data signal-to-noise ratio of the multi-dimensional feature vector set to obtain an adaptive fusion weight; based on the adaptive fusion weight, performing multi-level coupling on the multi-dimensional real-time feature vector set to obtain real-time fusion data; according to the invention, the efficiency of real-time fusion of geological disaster monitoring data can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data fusion, and in particular to a geological disaster monitoring data real-time fusion method and system. BACKGROUND

[0002] The multi-source data in the geological disaster monitoring scene has the characteristics of heterogeneous format and different dimensions, the existing heterogeneous data adaptation mechanism is imperfect, the time sequence calibration and dimension normalization processing lack precision, and it is difficult to generate standardized data sequences with strong consistency and high reliability. At the same time, the existing technology fails to deeply mine the dynamic coupling relationship between multiple monitoring parameters, the parameter correlation model constructed lacks timeliness and adaptability, and cannot dynamically reflect the real-time changes of parameter relationship, resulting in that the key correlation information contained in the data cannot be effectively utilized, and hidden dangers are caused for subsequent data fusion.

[0003] In the data fusion process, the existing technology lacks a parameter correlation-based spatio-temporal constraint rule system, it is difficult to accurately calibrate the standardized data in the spatio-temporal double dimensions, and the construction quality of the spatio-temporal consistent data set is poor; the vectorization mapping method of time-frequency domain features is not perfect, the comprehensiveness and accuracy of feature extraction are insufficient, and the fusion weight allocation does not realize the collaborative optimization with the data signal-to-noise ratio and the real-time weight of parameter correlation, resulting in insufficient accuracy of the fusion data after multi-level coupling, and weak real-time response ability of the overall processing flow. Therefore, how to improve the real-time fusion efficiency of geological disaster monitoring data has become a problem to be solved. SUMMARY

[0004] The present application provides a geological disaster monitoring data real-time fusion method and system to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides a geological disaster monitoring data real-time fusion method, which comprises: S1. Heterogeneous adaptation analysis is performed on the multi-source geological monitoring data of a target region to obtain a standardized data sequence of the target region; S2. According to the coupling relationship between multiple monitoring parameters in the standardized data sequence, a dynamic semantic relationship graph of the target region is constructed; S3. Based on the spatio-temporal constraint rules of the dynamic semantic relationship graph, the standardized data sequence is calibrated in the spatio-temporal double dimensions to obtain a spatio-temporal consistent data set of the target region; S4. The time-frequency domain features in the spatio-temporal consistent data set are vectorized and mapped to obtain a multi-dimensional feature vector set of the target region; S5. The real-time weight of the dynamic semantic relationship graph and the data signal-to-noise ratio of the multi-dimensional feature vector set are collaboratively optimized to obtain an adaptive fusion weight of the target region; S6. performing multi-level coupling on the multi-dimensional real-time feature vector set based on the adaptive fusion weight, to obtain real-time fusion data of the target region.

[0006] In a preferred embodiment, the heterogeneous adaptive analysis of the multi-source geological monitoring data of the target region to obtain the standardized data sequence of the target region comprises: collecting multi-source geological monitoring data of the target region; performing time series calibration on the multi-source geological monitoring data to obtain time-aligned data of the target region; performing dimensionless normalization processing on the time-aligned data to obtain a dimensionless data set of the target region; performing sequence reorganization on the dimensionless data set to obtain the standardized data sequence of the target region.

[0007] In a preferred embodiment, the dynamic semantic relationship graph of the target region is constructed according to the coupling relationship between the multiple monitoring parameters in the standardized data sequence, comprising: performing coupling relationship extraction on the coupling features between the multiple monitoring parameters in the standardized data sequence to obtain a candidate relationship set between the multiple monitoring parameters, and performing topological construction on the candidate relationship set to obtain an initial relationship network structure of the target region; performing dynamic time window constraint on the initial relationship network structure to obtain a time-varying semantic network of the target region; performing rule verification on the time-varying semantic network to obtain a verified semantic network of the target region; performing topological reconstruction on the verified semantic network to obtain a dynamic semantic relationship graph of the target region.

[0008] In a preferred embodiment, the dynamic time window constraint on the initial relationship network structure to obtain the time-varying semantic network of the target region comprises: performing sliding time window division on the continuous time axis of the standardized data sequence to obtain a continuous time window sequence of the target region; based on the continuous time window sequence, performing segmented sampling on the standardized data sequence to obtain a windowed data subset of the target region; based on the windowed data subset, performing one-by-one matching on the node connection positions in the initial relationship network structure to obtain an instantaneous relationship subgraph of the target region; performing dynamic linking on the instantaneous relationship subgraph to obtain a time-varying semantic network of the target region.

[0009] In a preferred embodiment, the spatio-temporal constraint rule based on the dynamic semantic relation graph performs spatio-temporal two-dimensional checking on the standardized data sequence to obtain a spatio-temporal consistent data set of the target region, including: performing spatial dimension checking on the standardized data sequence to obtain spatial dimension consistent data of the standardized data sequence; fusing the standardized data sequence and the spatial dimension consistent data to obtain spatio-temporal alignment data of the standardized data sequence; performing associated constraint analysis on the dynamic semantic relation graph to obtain a spatio-temporal constraint rule of the dynamic semantic relation graph; performing constraint-driven interpolation on the spatio-temporal alignment data according to the spatio-temporal constraint rule to obtain a spatio-temporal consistent data set of the target region.

[0010] In a preferred embodiment, the vectorization mapping of the time-frequency domain features in the spatio-temporal consistent data set obtains a multi-dimensional feature vector set of the target region, including: performing multi-scale feature fusion on the time domain features and the frequency domain features in the spatio-temporal consistent data set to obtain a fusion feature set of the spatio-temporal consistent data set; interactively coupling the feature dimension of the spatio-temporal consistent data set and the fusion feature set to obtain an embedding vector set of the target region; performing feature semantic regularization on the embedding vector set to obtain a semantic clear vector group of the spatio-temporal consistent data set; performing topological integration on the vector structure of the semantic clear vector group to obtain a multi-dimensional feature vector set of the target region.

[0011] In a preferred embodiment, the real-time weight of the dynamic semantic relation graph and the data signal-to-noise ratio of the multi-dimensional feature vector set are cooperatively optimized to obtain an adaptive fusion weight of the target region, including: performing associated mapping on the real-time weight of the dynamic semantic relation graph and the data signal-to-noise ratio of the multi-dimensional feature vector set to obtain an initial matching relation of the target region; based on the initial matching relation, jointly evaluating the real-time weight and the data signal-to-noise ratio to obtain a cooperative evaluation index of the target region; according to the cooperative evaluation index, performing strategy calculation on the spatio-temporal consistent data set to obtain an adaptive fusion weight of the target region.

[0012] In a preferred embodiment, the strategy calculation on the spatio-temporal consistent data set according to the cooperative evaluation index to obtain an adaptive fusion weight of the target region includes: The cooperative evaluation index is cooperatively analyzed with the dynamic semantic relationship graph to obtain a mapping rule set of the target region; The real-time weight and the data signal-to-noise ratio are rule-applied and mapped based on the mapping rule set to obtain a candidate weight set of the target region; The candidate weight set is subjected to consistency verification to obtain an effective weight of the target region; The effective weight is subjected to adaptive optimization to obtain an adaptive fusion weight of the target region, wherein a calculation formula of the adaptive fusion weight is as follows: ; In the formula, is a component of the adaptive fusion weight, is an i-th component of the real-time weight, is a normalized value of an i-th component of the cooperative evaluation index, is an i-th component of the data signal-to-noise ratio, is an adjustment index of the real-time weight, is an adjustment index of the cooperative evaluation index, is an adjustment index of the data signal-to-noise ratio, is a small normal number preventing a denominator from being zero, is a total number of the multiple monitoring parameters, is an i-th component of the real-time weight, is a normalized value of an i-th component of the cooperative evaluation index, is an i-th component of the data signal-to-noise ratio. In a preferred embodiment, based on the adaptive fusion weight, the multiple real-time feature vector sets are subjected to multi-level coupling to obtain real-time fusion data of the target region, which comprises: The multiple real-time feature vector sets are subjected to hierarchical decomposition to obtain a multi-level feature sequence of the target region; Based on the adaptive fusion weight, the multi-level feature sequence is subjected to feature interaction fusion to obtain a hierarchical fusion sequence of the target region; The hierarchical fusion sequences are subjected to cross-level aggregation to obtain a preliminary fusion data sequence of the target region; The preliminary fusion data sequence is subjected to time-domain continuity smoothing to obtain real-time fusion data of the target region.

[0013] In a preferred embodiment, based on the adaptive fusion weight, the multiple real-time feature vector sets are subjected to multi-level coupling to obtain real-time fusion data of the target region, which comprises: The multiple real-time feature vector sets are subjected to hierarchical decomposition to obtain a multi-level feature sequence of the target region; Based on the adaptive fusion weight, the multi-level feature sequence is subjected to feature interaction fusion to obtain a hierarchical fusion sequence of the target region; The hierarchical fusion sequences are subjected to cross-level aggregation to obtain a preliminary fusion data sequence of the target region; The preliminary fusion data sequence is subjected to time-domain continuity smoothing to obtain real-time fusion data of the target region.

[0014] ​To solve the above problems, the application also provides a geological disaster monitoring data real-time fusion system, the system comprises: A data standardization module is configured to perform heterogeneous adaptive analysis on multi-source geological monitoring data of a target region to obtain a standardized data sequence of the target region. A semantic relationship graph construction module is configured to construct a dynamic semantic relationship graph of the target region according to coupling relationships between multiple monitoring parameters in the standardized data sequence. A space-time dual-dimension data correction module is configured to perform space-time dual-dimension correction on the standardized data sequence based on space-time constraint rules of the dynamic semantic relationship graph to obtain a space-time consistent data set of the target region. A time-frequency domain feature vectorization module is configured to vectorize and map time-frequency domain features in the space-time consistent data set to obtain a multi-dimensional feature vector set of the target region. An adaptive weight optimization module is configured to cooperatively optimize real-time weights of the dynamic semantic relationship graph and data signal-to-noise ratios of the multi-dimensional feature vector set to obtain adaptive fusion weights of the target region. A multi-level fusion output module is configured to perform multi-level coupling on the multi-dimensional real-time feature vector set based on the adaptive fusion weights to obtain real-time fusion data of the target region.

[0015] Compared with the prior art, the application has the following beneficial effects: 1. The application performs heterogeneous adaptive analysis on multi-source geological monitoring data, and through time sequence calibration, dimension normalization and sequence recombination, effectively improves the consistency and reliability of the standardized data sequence, and lays a solid foundation for subsequent data fusion work. Meanwhile, the dynamic semantic relationship graph is constructed based on the coupling relationships between multiple monitoring parameters in the standardized data sequence, and through dynamic time window constraint, rule verification and topology reconstruction, the real-time dynamic changes of parameter correlation can be accurately captured, and the associated information contained in the data can be fully mined and efficiently utilized.

[0016] 2. The application performs space-time dual-dimension correction based on the space-time constraint rules of the dynamic semantic relationship graph to ensure the accuracy and integrity of the space-time consistent data set; through multi-scale feature fusion, semantic regularization and topology integration, the time-frequency domain feature vectorization is completed, so that the multi-dimensional feature vector set can fully and clearly present the core features of the data. In addition, through the cooperative optimization of the real-time weights of the dynamic semantic relationship graph and the data signal-to-noise ratio, the adaptive fusion weights are obtained, and then through multi-level coupling and time domain continuity smoothing processing, the accuracy and stability of the real-time fusion data are significantly improved, and the real-time response capability of data fusion is guaranteed, which provides strong data support for accurate decision-making of geological disaster monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 FIG. 1 shows a flowchart of a geological disaster monitoring data real-time fusion method according to an embodiment of the present application; Figure 2 FIG. 2 shows a functional module diagram of a geological disaster monitoring data real-time fusion system according to an embodiment of the present application; The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0018] It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

[0019] The present application provides a geological disaster monitoring data real-time fusion method. The execution subject of the geological disaster monitoring data real-time fusion method includes but is not limited to at least one of electronic devices such as a server, a terminal, etc., which can be configured to execute the method provided by the present application. In other words, the geological disaster monitoring data real-time fusion method can be executed by software or hardware installed in a terminal device or a server device. The server includes but is not limited to a single server, a server cluster, a cloud server, or a cloud server cluster, etc. The server can be a stand-alone server, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (CDN), and big data and artificial intelligence platforms, etc. basic cloud computing services.

[0020] Referring to FIG. 1, a flowchart of a geological disaster monitoring data real-time fusion method according to an embodiment of the present application is shown. In the present embodiment, the geological disaster monitoring data real-time fusion method includes: Figure 1 S1. Heterogeneous adaptation analysis is performed on multi-source geological monitoring data of a target region to obtain a standardized data sequence of the target region; S1. Heterogeneous adaptation analysis is performed on multi-source geological monitoring data of a target region to obtain a standardized data sequence of the target region; In the present embodiment, the heterogeneous adaptation analysis on the multi-source geological monitoring data of the target region to obtain the standardized data sequence of the target region includes: Collecting multi-source geological monitoring data of a target region; Performing time sequence calibration on the multi-source geological monitoring data to obtain time-aligned data of the target region; Performing dimensionless normalization processing on the time-aligned data to obtain a dimensionless data set of the target region; Performing sequence recombination on the dimensionless data set to obtain a standardized data sequence of the target region.

[0021] Determine the specific monitoring range of the target region, and determine the types of geological disasters that need to be monitored and the core monitoring indicators in the range. Scientifically deploy multiple types of monitoring equipment in the monitoring range, including geological stress monitoring equipment, displacement monitoring equipment, groundwater monitoring equipment, and weather monitoring equipment. All monitoring equipment continuously collects data at a predetermined fixed sampling frequency. During the collection process, the key auxiliary information corresponding to each raw data is recorded simultaneously, including the exact collection time, specific collection location coordinates, and corresponding device unique number, to avoid data confusion. All raw data collected by the equipment is fully summarized, and invalid random code data caused by obvious equipment failure is removed to ensure that the summarized data can completely cover all monitoring dimensions of the target region, and finally form a multi-source geological monitoring data of the target region.

[0022] Extract the collection time stamp of each data from the multi-source geological monitoring data one by one. According to the core functional requirements of the monitoring system, select the sampling time of the core monitoring equipment with stable performance and accurate sampling frequency as the unified reference time axis, which needs to include continuous and uniformly distributed time nodes. Compare the time stamps of all non-core device collected data with the time nodes on the reference time axis one by one. For data with time deviation, adjust the time mark accurately in combination with the fixed sampling interval of the corresponding device and the preset reasonable data transmission delay rule, to ensure that the time information of each data can be accurately matched to the corresponding node on the reference time axis. Make all data collected by different sources and different devices completely consistent in the time dimension, and finally obtain the time-aligned data of the target region.

[0023] Statistically determine the maximum and minimum values of each parameter in all data in the time-aligned data, to ensure that the statistical results can truly reflect the actual data distribution of the parameter. For each data under each parameter, perform dimension conversion according to the fixed processing procedure. Subtract the minimum value of the parameter from the specific value of the data, and then divide the difference by the difference between the maximum and minimum values of the parameter. Through this unified calculation method, all types of monitoring parameter data with different dimensions are converted to the unified value interval of 0 to 1. Classify and integrate all data that no longer have the original dimension properties after dimension conversion, and arrange them in order according to the monitoring parameter type to form a dimensionless data set of the target region.

[0024] With the previously determined reference time axis as the core sequencing basis, all data in the dimensionless data set are strictly arranged in time sequence, ensuring that the data can be presented in chronological order. For the dimensionless data of various monitoring parameters at the same time node, the data are associated and combined according to the preset parameter classification rule, so that each time node corresponds to a data group containing complete monitoring parameter information, avoiding parameter omission. At the same time, the combined data group is structured, the arrangement order of different parameters in each data group is determined and kept consistent, ensuring that the structure of the entire data set is unified and the logic is clear. After sequencing and combination, the standardized data sequence of the target region is finally formed.

[0025] The beneficial effects are that by comprehensively collecting multi-source geological monitoring data of the target region, the multi-parameter dimension of geological disaster monitoring can be completely covered, the information limitation of a single data source is broken, and comprehensive and rich basic data support is provided for subsequent data fusion; by time sequence calibration of the multi-source geological monitoring data, the time deviation caused by different monitoring equipment and different collection frequencies is accurately eliminated, time-aligned data with high consistency in time dimension are obtained, time reference consistency is ensured for subsequent time and space two-dimensional correction and dynamic semantic relationship graph construction, and fusion errors caused by time asynchrony are avoided; by dimensionless normalization processing, the data incomparability problem caused by dimension difference of different monitoring parameters is effectively eliminated, the dimensionless data set has a unified data reference, and the rationality and accuracy of multi-parameter fusion are significantly improved; by serializing and recombining the dimensionless data set, the dispersed heterogeneous data are converted into structured and ordered standardized data sequence, the processing complexity of subsequent core steps such as dynamic semantic relationship graph construction and time-frequency domain feature vector mapping is reduced, and the efficiency of data flow and processing is improved.

[0026] S2. According to the coupling relationship between the multiple monitoring parameters in the standardized data sequence, a dynamic semantic relationship graph of the target region is constructed; In the embodiment of the application, the dynamic semantic relationship graph of the target region is constructed according to the coupling relationship between the multiple monitoring parameters in the standardized data sequence, comprising: The coupling features between the multiple monitoring parameters in the standardized data sequence are coupled to obtain a candidate relationship set between the multiple monitoring parameters, and the candidate relationship set is topologically constructed to obtain an initial relationship network structure of the target region; The initial relationship network structure is dynamically time-window constrained to obtain a time-varying semantic network of the target region; The time-varying semantic network is subjected to rule verification to obtain a verified semantic network of the target region; The verified semantic network is topologically reconstructed to obtain the dynamic semantic relationship graph of the target region.

[0027] The initial relationship network structure is dynamically time-window constrained to obtain a time-varying semantic network of the target region. The continuous time axis of the standardized data sequence is divided by a sliding time window to obtain a continuous time window sequence of the target region. Based on the continuous time window sequence, the standardized data sequence is segmented and sampled to obtain a windowed data subset of the target region. Based on the windowed data subset, the node connection positions in the initial relationship network structure are matched one by one to obtain an instantaneous relationship subgraph of the target region. The instantaneous relationship subgraph is dynamically linked to obtain a time-varying semantic network of the target region.

[0028] The change trends of all monitoring parameters in the standardized data sequence are comprehensively analyzed, and the change correlation of different parameters at the same time node and time interval is focused on. When a parameter value fluctuates, whether other parameters will change accordingly is tracked. The mutual influence and correlation characteristics are extracted as coupling features. All extracted coupling features are classified and sorted to determine the correlation types and strengths between parameters, forming a candidate relationship set between multiple monitoring parameters. Then each monitoring parameter is taken as an independent node, and the correlation in the candidate relationship set is taken as the connection edge between nodes. The network framework is built according to the logical order of parameter correlation to form an initial relationship network structure of the target region.

[0029] A fixed time length is set as the window size of the sliding time window, and a fixed moving step is determined. Starting from the beginning of the continuous time axis of the standardized data sequence, the window is moved backward according to the set moving step. The time interval in the corresponding time period is intercepted after each movement, until the window covers the entire continuous time axis. All time intervals arranged in time order together form a continuous time window sequence of the target region.

[0030] For each time window in the continuous time window sequence, the corresponding time interval range is determined, and all monitoring parameter data contained in the time interval are extracted from the standardized data sequence to ensure that each time window can correspond to a complete data group matching the time period. All extracted data groups are arranged in order according to their corresponding time windows to form a windowed data subset of the target region.

[0031] Each group of data in the windowed data subset is extracted, corresponding to the monitoring data of a time window respectively. For each group of data, the actual existing correlation between the monitoring parameters in the time window is analyzed. Each node in the initial relationship network structure is corresponded to the correlation object of the parameter in the current window one by one. The node connection consistent with the correlation state reflected by the current window data is retained, and the inconsistent connection is removed, forming a transient relationship subgraph that can only reflect the parameter relationship in the specific time window.

[0032] In the order of the sequence of continuous time windows, all generated transient relationship subgraphs are sequentially connected to ensure that the nodes of the previous transient relationship subgraph and the corresponding nodes in the next transient relationship subgraph maintain consistent correspondence. At the same time, the change of node connection between adjacent transient relationship subgraphs is recorded, so that all transient relationship subgraphs form a coherent network structure that can reflect the dynamic change of parameter correlation with time, and the time-varying semantic network of the target region is obtained.

[0033] According to the verified parameter correlation logic and data rationality criteria in the field of geological disaster monitoring, each node connection in the time-varying semantic network is checked one by one to determine whether the correlation between the nodes conforms to the parameter action law under the actual monitoring scene. False connections that do not exist actual correlation are removed, and node connection relationships that have logical contradictions are corrected to ensure that all parameter correlations in the network have rationality and effectiveness. After comprehensive inspection and correction, the verified semantic network of the target region is obtained.

[0034] The node distribution and connection structure of the verified semantic network are optimized and adjusted. According to the closeness of parameter correlation, the arrangement of nodes is re-planned so that parameter nodes with close correlation are located closer in the network. At the same time, redundant connection paths in the network are simplified, core and key parameter connections are retained, core correlation relationships are strengthened, dynamic characteristics of parameter relationship changes over time are completely retained, and the network can clearly and accurately present the correlation state of multiple monitoring parameters under different time dimensions. Finally, the dynamic semantic relationship graph of the target region is formed.

[0035] The beneficial effect is that by extracting and topological construction of the coupling characteristics of multiple monitoring parameters in the standardized data sequence, the internal correlation logic between parameters can be accurately mined, a structured initial relationship network structure is formed, a clear correlation foundation is laid for subsequent dynamic optimization of the semantic relationship graph, and fusion deviation caused by fuzzy correlation in multi-parameter fusion is effectively avoided;Through the dynamic time window constraint mechanism, the continuous time axis is first divided and sampled in sections, then the windowed data subset is matched with the node connection position and the instantaneous relationship subgraph is dynamically linked, so that the initial relationship network can respond to the time sequence change in real time, generate a time-varying semantic network that fits the dynamic characteristics of the data, and greatly improve the adaptability of the semantic relationship graph to the time sequence volatility of the geological monitoring data;Through rule checking, false correlation and logical conflicts in the time-varying semantic network are removed, ensuring the accuracy and compliance of the semantic relationship, and topological reconstruction further optimizes the rationality and efficiency of the network structure, and finally the dynamic semantic relationship graph not only retains the coupling nature between multiple parameters, but also has real-time dynamic updating characteristics, providing high-quality semantic support for subsequent spatio-temporal constraint rule extraction and adaptive fusion weight calculation.

[0036] S3. Based on the spatio-temporal constraint rule of the dynamic semantic relationship graph, the standardized data sequence is subjected to spatio-temporal double-dimensional correction to obtain a spatio-temporally consistent data set of the target region; In the embodiment of the present application, the spatio-temporal constraint rule based on the dynamic semantic relationship graph is used to perform spatio-temporal double-dimensional correction on the standardized data sequence to obtain a spatio-temporally consistent data set of the target region, which comprises: The standardized data sequence is subjected to spatial dimension verification to obtain spatial dimension consistent data of the standardized data sequence; The standardized data sequence and the spatial dimension consistent data are fused to obtain spatio-temporally aligned data of the standardized data sequence; The dynamic semantic relationship graph is subjected to correlation constraint analysis to obtain a spatio-temporal constraint rule of the dynamic semantic relationship graph; According to the spatio-temporal constraint rule, the spatio-temporally aligned data is subjected to constraint-driven interpolation to obtain a spatio-temporally consistent data set of the target region.

[0037] The spatial coordinates of each data in the standardized data sequence corresponding to the monitoring point and the monitoring area range to which it belongs are determined, and the spatial position information of each data is checked one by one to see whether it is consistent with the preset monitoring layout of the target region, and data that exceeds the preset monitoring range or has obvious errors in spatial coordinates is removed, and coordinate offset data caused by equipment positioning deviation is corrected, so that all retained data meet the distribution requirements of the actual monitoring scene in the spatial position, and spatial dimension consistent data of the standardized data sequence is obtained.

[0038] The spatial dimension consistent data and the original standardized data sequence are accurately matched according to timestamps, and for each data in the standardized data sequence, if the spatial information thereof has been verified and does not need to be corrected, the data is directly retained; if the spatial information thereof is excluded or corrected, the accurate spatial information of the corresponding time node in the spatial dimension consistent data is used for replacement and supplement, so that each data maintains the original time dimension continuity and has the accurate spatial attribute verified, and time-space alignment data of the standardized data sequence is formed.

[0039] The association logic between the monitoring parameter nodes in the dynamic semantic relationship graph is combed, the mutual dependency relationship of different parameters on the spatial distribution is analyzed, the spatial range in which the parameter association takes effect is determined, the synchronous response law of each parameter on the time change is summarized, the time interval requirement in which the parameter association is established is determined, the spatial dependency relationship and the time response law are arranged into clear constraint criteria, and the time-space constraint rule of the dynamic semantic relationship graph is formed.

[0040] The data integrity of each time node and each spatial position in the time-space alignment data is checked one by one, for the case that there is data loss or abnormality, according to the association requirement of the parameter in the time-space constraint rule, referring to the effective data of the adjacent time nodes at the same spatial position and the associated parameter data of the adjacent spatial positions at the same time node, the missing data is supplemented and the abnormal data that does not conform to the constraint rule is corrected according to the reasonable response logic between the parameters, so that all data meet the parameter association criteria in the time-space dimension, and the time-space consistent data set of the target region is obtained.

[0041] The beneficial effects are that by performing spatial dimension verification on the standardized data sequence, the spatial positioning deviation and data heterogeneity problems caused by different monitoring points and different collection devices are accurately excluded, the data with high consistency in the spatial dimension is obtained, the incompatibility problem of multi-source monitoring data in the spatial dimension is solved from the root, and the spatial reference guarantee is provided for subsequent time-space fusion; by fusing the standardized data sequence and the spatial dimension consistent data, the time dimension and the spatial dimension are aligned, the time-space alignment data is formed, the data association barrier in the time-space dimension is completely broken through, the data has a unified time-space reference system, and the dimension conflict in subsequent data processing is greatly reduced; by performing association constraint analysis on the dynamic semantic relationship graph, the time-space constraint rule conforming to the internal law of the geological monitoring parameter is extracted, the data interpolation process no longer depends on blind algorithm, but is guided by the constraint condition conforming to the geological logic, and the rationality and scientificity of the interpolated data are ensured; according to the time-space constraint rule, the interpolation is driven and constrained, the data missing gap can be efficiently filled, and it is ensured that the filled data strictly follows the time-space association rule, and finally the time-space consistent data set has completeness, consistency and reliability.

[0042] S4. Vectorizing the time-frequency domain features in the spatiotemporal consistent dataset to obtain a multi-dimensional feature vector set of the target region; In the embodiment of the present application, the vectorizing the time-frequency domain features in the spatiotemporal consistent dataset to obtain a multi-dimensional feature vector set of the target region comprises:

[0043] When extracting the time domain features in the spatiotemporal consistent dataset, a plurality of fixed windows are divided according to different time lengths, and the core information such as the trend of data change and fluctuation amplitude in each window is extracted to form a multi-scale time domain feature group; when extracting the frequency domain features, the time domain data of each time window is converted into frequency domain data to capture the feature distribution of different frequency bands to form a multi-scale frequency domain feature group, and then the time domain features and the frequency domain features under the same scale are associated and integrated to retain complementary information and eliminate feature conflicts, and the integration results of all scales are arranged in order to obtain the fusion feature set of the spatiotemporal consistent dataset.

[0044] The feature dimension information of the spatiotemporal consistent dataset is determined, including the monitoring attribute, data type and value range corresponding to each dimension to form a feature dimension description set; the dimension of each fusion feature in the fusion feature set is accurately matched with the corresponding dimension in the feature dimension description set, and the feature information of the two is added element by element in the order of dimensions, so that each dimension contains both the original feature information of the spatiotemporal consistent dataset and the enhanced information of the fusion features, and the integration result of each dimension constitutes an independent vector, and all vectors are combined in the order of dimensions to obtain the embedding vector set of the target region.

[0045] The semantic connotation of each vector in the embedding vector set is analyzed one by one to determine the geological monitoring semantic direction corresponding to each vector, and the redundant information irrelevant to the core semantics in the vector is eliminated; the values of all vectors are uniformly adjusted to the same reasonable interval to strengthen the feature strength corresponding to the core semantics of each vector, and at the same time, the consistency of semantic expression between vectors is calibrated to ensure that the semantics of different vectors can be directly compared, and all the processed vectors form the semantic clear vector group of the spatiotemporal consistent dataset.

[0046] ​​​​The semantic correlation strength and feature dependency relationship between each vector in the semantic clear vector group are combed, and a correlation topological structure between vectors is constructed; according to the correlation strength, vectors with high semantic correlation are clustered and integrated, and the arrangement order of the vectors is optimized, so that the topological structure of the vector group is adapted to the feature logic of the geological monitoring data, while the core feature information of each vector is retained, and the integrated vector group can comprehensively and orderly present multi-dimensional monitoring features, and finally a multi-dimensional feature vector set of the target region is formed.

[0047] The beneficial effects are that by performing multi-scale feature fusion on the time domain features and the frequency domain features in the spatiotemporal consistent data set, the dynamic change trend of the geological monitoring data in the time dimension and the periodicity of the frequency dimension can be comprehensively captured, the limitations of single-scale feature analysis are broken, the fused feature set has both time sequence continuity and frequency domain correlation, and the core information related to the evolution of geological disasters is completely retained; the feature dimensions of the spatiotemporal consistent data set and the fused feature set are interactively coupled, the deep correlation between different dimensional features and multi-scale fused features is realized, the embedded vector set effectively integrates multi-element information, avoids information fragmentation caused by feature isolation, and greatly improves the representativeness and discriminability of the features; the embedded vector set is subjected to feature semantic regularization, fuzzy and redundant semantic information in the vector is eliminated, the geological monitoring physical meaning corresponding to each vector is determined, the interference of semantic ambiguity on subsequent fusion calculation is avoided, and the semantic clear vector group has stronger explainability; the semantic clear vector group is subjected to topological integration, the dispersed vectors are structurally reorganized according to the internal correlation logic of the geological monitoring parameters, an ordered and regular multi-dimensional feature vector set is formed, the processing complexity of subsequent adaptive fusion weight calculation and multi-level coupling is reduced, and the logical coherence between the feature vectors is ensured. The overall process performs progressive processing through “multi-scale fusion-interactive coupling-semantic regularization-topological integration”, significantly improves the integrity, clarity and structure of the geological monitoring data features, and provides high-quality feature support for accurate calculation of adaptive fusion weights and efficient coupling of multi-dimensional features.

[0048] S5. The real-time weight of the dynamic semantic relationship graph and the data signal-to-noise ratio of the multi-dimensional feature vector set are cooperatively optimized to obtain the adaptive fusion weight of the target region. In the embodiment of the present application, the real-time weight of the dynamic semantic relationship graph and the data signal-to-noise ratio of the multi-dimensional feature vector set are cooperatively optimized to obtain the adaptive fusion weight of the target region, which comprises: The real-time weight of the dynamic semantic relationship graph and the data signal-to-noise ratio of the multi-dimensional feature vector set are associatedly mapped to obtain the initial matching relationship of the target region. Based on the initial matching relationship, the real-time weight and the data signal-to-noise ratio are jointly evaluated to obtain the cooperative evaluation index of the target region. According to the cooperative evaluation index, the spatio-temporal consistent data set is strategy calculated to obtain the adaptive fusion weight of the target region.

[0049] The strategy calculation of the spatio-temporal consistent data set according to the cooperative evaluation index to obtain the adaptive fusion weight of the target region comprises: The cooperative evaluation index and the dynamic semantic relationship graph are cooperatively analyzed to obtain a mapping rule set of the target region. The real-time weight and the data signal-to-noise ratio are rule applied and mapped based on the mapping rule set to obtain a candidate weight set of the target region. The candidate weight set is consistency checked to obtain an effective weight of the target region. The effective weight is adaptively optimized to obtain the adaptive fusion weight of the target region, wherein the calculation formula of the adaptive fusion weight is as follows: ; In the formula, is the adaptive fusion weight component, is the first component in the real-time weight, is the first component in the cooperative evaluation index, is the normalized value of the first component in the cooperative evaluation index, is the first component in the data signal-to-noise ratio, is the first component in the data signal-to-noise ratio, is the adjustment index of the real-time weight, is the adjustment index of the cooperative evaluation index, is the adjustment index of the data signal-to-noise ratio, is the adjustment index of the data signal-to-noise ratio, is a minimum normal number to prevent the denominator from being zero, is the total number of the multiple monitoring parameters, is the first component in the real-time weight, is the normalized value of the first component in the cooperative evaluation index, is the first component in the data signal-to-noise ratio, is the first component in the data signal-to-noise ratio.

[0050] ​​The real-time weight corresponding to each monitoring parameter in the dynamic semantic relationship graph is combed, and the attribution object of each real-time weight component is determined, that is, each component uniquely corresponds to a monitoring parameter of the target region. Meanwhile, the data signal-to-noise ratio corresponding to each feature vector in the multi-dimensional feature vector set is extracted, so as to ensure that each data signal-to-noise ratio component also corresponds to the monitoring parameter of the target region, and the attribution parameter of the data signal-to-noise ratio component is consistent with the attribution parameter of the real-time weight component. The real-time weight component and the data signal-to-noise ratio component corresponding to the same monitoring parameter are paired one by one, and the parameter identifier, real-time weight value and data signal-to-noise ratio value in each paired combination are clearly labeled, thereby establishing an association relationship between the real-time weight and the data signal-to-noise ratio corresponding to each monitoring parameter. The association relationship combination of all monitoring parameters forms the initial matching relationship of the target region. Based on each association combination in the initial matching relationship, the matching condition between the strength of the real-time weight and the level of the data signal-to-noise ratio is analyzed in depth, whether the monitoring parameter with a higher real-time weight corresponds to a higher data signal-to-noise ratio, whether the monitoring parameter with a lower real-time weight corresponds to a lower data signal-to-noise ratio, and whether the matching condition is consistent with the importance of the monitoring parameter in geological disaster monitoring. The matching condition of each association combination is converted into a quantitative value under a unified standard, and the size of the quantitative value directly reflects the level of the matching degree. All monitoring parameters corresponding to the quantitative value are arranged in sequence according to the parameter order, thereby forming a cooperative evaluation index of the target region that can comprehensively reflect the matching effect of the real-time weight and the data signal-to-noise ratio.

[0051] The association logic of each quantitative value in the cooperative evaluation index and the corresponding monitoring parameter in the dynamic semantic relationship graph is combined, and the internal relationship between the matching state reflected by different quantitative values and the parameter association strength, parameter action law is analyzed in depth. For different value intervals of the cooperative evaluation index, the corresponding real-time weight adjustment direction and adjustment amplitude standard are determined, and the reference priority of the data signal-to-noise ratio in the weight adjustment is determined in combination with the closeness of the parameter association in the dynamic semantic relationship graph. These adjustment directions, amplitude standards and reference priorities are arranged into clear and directly applicable rule items, each rule item clearly corresponds to a specific cooperative evaluation index range and parameter association scene, and all rule items are integrated to form a mapping rule set of the target region.

[0052] For each monitoring parameter, its corresponding real-time weight component and data signal-to-noise ratio component are extracted from the initial matching relationship, and the quantized value corresponding to the parameter in the cooperative evaluation index is obtained. By comparing with the rule entries in the mapping rule set, the rule completely matching the interval to which the quantized value belongs and the associated state of the parameter in the dynamic semantic relationship graph is found. According to the adjustment direction and amplitude standard specified in the found rule, combined with the reference priority of the data signal-to-noise ratio, the real-time weight component of the parameter is adjusted specifically to obtain the preliminary weight value of the parameter after rule adaptation. The preliminary weight values of all monitoring parameters are arranged in order according to the inherent order of the parameters to form a candidate weight set of the target region, in which each value corresponds to a specific monitoring parameter.

[0053] All preliminary weight values in the candidate weight set are comprehensively checked to confirm whether each value is within a reasonable value range required for data fusion, and to verify whether each preliminary weight value and the corresponding monitoring parameter in the dynamic semantic relationship graph are consistent in associated importance, so as to ensure that the size of the weight value can accurately reflect the degree of the parameter's role in the associated network. Preliminary weight values that are out of the reasonable range or have obvious contradictions with the associated importance of the parameters are removed. For the weight missing positions caused by removal, logical weight values are supplemented and generated by referring to the weight value rules of adjacent monitoring parameters and the associated characteristics of the missing parameters and other parameters, to finally form an effective weight of the target region in which all values are valid and consistent with the associated state of the parameters.

[0054] In combination with the actual data distribution characteristics of each monitoring parameter in the spatio-temporal consistent data set of the target region, including the fluctuation range, change trend and mutual influence degree of the data, and in reference to the real-time dynamic changes of parameter association in the dynamic semantic relationship graph, each value in the effective weight is finely adjusted. For the weight value corresponding to the parameter with stable data distribution and prominent associated effect, the value size is appropriately strengthened to improve its contribution proportion in the fusion; for the weight value corresponding to the parameter with large data fluctuation and weak associated effect, the value size is appropriately weakened to avoid interference with the fusion result. After the fine adjustment, the weight set, each value of which can accurately adapt to the current data characteristics and parameter association state, accurately reflects the actual contribution degree of the corresponding parameter in the data fusion, and finally forms the adaptive fusion weight of the target region.

[0055] The formula is used to calculate the adaptive fusion weight component corresponding to each parameter in the target regional multi-monitoring parameter. The core significance is to organically integrate the real-time weight of the dynamic semantic relationship diagram, the normalized value of the collaborative evaluation index, and the data signal-to-noise ratio of the multi-dimensional feature vector set, and accurately quantify the actual contribution of each monitoring parameter in the data fusion process. The adaptive fusion weight component obtained by the formula can adapt to the real-time characteristics of the monitoring data and the parameter correlation state, provide a core basis for subsequent multi-level coupling of the multi-dimensional real-time feature vector set based on adaptive fusion weight, and generate accurate real-time fusion data, ensuring that the final fusion data can fully reflect the effective information of each parameter and improving the accuracy and reliability of geological disaster monitoring data fusion. The first step is to calculate the numerator of a single monitoring parameter. The first component in the real-time weight is raised to the power of the adjustment index of the real-time weight; the normalized value of the first component in the collaborative evaluation index is added by 1 and raised to the power of the adjustment index of the collaborative evaluation index; the first component in the data signal-to-noise ratio is added by a small normal number to prevent the denominator from being zero, and raised to the power of the adjustment index of the data signal-to-noise ratio and then inverted; and the three operation results are multiplied to obtain the numerator value corresponding to the parameter. The second step is to calculate the denominator. For all monitoring parameters, the numerator value of each parameter is obtained by the above-mentioned numerator calculation method, and the sum of the numerator values is obtained to obtain the total denominator value. The third step is to calculate the adaptive fusion weight component. The numerator value of a single monitoring parameter is divided by the total denominator value, and the result is the adaptive fusion weight component corresponding to the parameter. The components of all parameters together constitute the adaptive fusion weight of the target region.

[0056] ​​​​The beneficial effect is that by correlating and mapping the real-time weight of the dynamic semantic relation graph with the data signal-to-noise ratio of the multi-dimensional feature vector set, the isolated analysis limitation of the two types of key parameters is broken, the initial matching relationship between the two is established, a precise correlation basis is provided for subsequent weight optimization, and the weight deviation caused by single parameter decision is avoided; the collaborative evaluation index is obtained based on the initial matching relationship, the comprehensive consideration of the parameter importance and the data quality is realized, the internal correlation priority of the geological monitoring parameters is taken into account, and the reliability of the data is fully considered, so that the comprehensiveness and scientificity of the evaluation result are greatly improved; the mapping rule set is obtained through collaborative analysis of the collaborative evaluation index and the dynamic semantic relation graph, the weight calculation rule is deeply matched with the semantic correlation logic of the geological monitoring data, the blindness of rule making is avoided, and the pertinence and rationality of the weight generation are ensured; the candidate weight set is generated based on the mapping rule set, the abstract rule is converted into a specific weight scheme that can be implemented, the precise adaptation of the rule and the parameter is realized, and the weight can accurately match the data characteristics of different monitoring scenes; the consistency of the candidate weight set is verified, the logical conflict and unreasonable value between the weights are effectively eliminated, the internal consistency and feasibility of the weight set are ensured, and reliable weight support is provided for subsequent fusion; through the adaptive optimization of the effective weight, the adaptive fusion weight is finally obtained, the weight has dynamic adjustment capability, can respond to data characteristic changes and semantic relation evolution in real time, completely gets rid of the limitation of fixed weight, and significantly improves the adaptability of the fusion weight to complex geological monitoring data.

[0057] S6. Based on the adaptive fusion weight, the multi-dimensional real-time feature vector set is coupled in multiple levels to obtain real-time fusion data of the target region.

[0058] In the embodiment of the application, based on the adaptive fusion weight, the multi-dimensional real-time feature vector set is coupled in multiple levels to obtain real-time fusion data of the target region, which comprises: deconstructing the multi-dimensional real-time feature vector set in levels to obtain a multi-level feature sequence of the target region; based on the adaptive fusion weight, the multi-level feature sequence is interactively fused to obtain a level fusion sequence of the target region; the level fusion sequences are aggregated across levels to obtain a preliminary fusion data sequence of the target region; the preliminary fusion data sequence is smoothed in time domain to obtain real-time fusion data of the target region.

[0059] All feature vectors in the multi-dimensional real-time feature vector set are comprehensively analyzed for feature attributes, and the type of geological monitoring information, data dimension and core representation meaning carried by each feature vector are determined. The feature information is strictly classified according to the abstract degree and parameter association closeness. The features directly reflecting the physical properties of the original monitoring data are classified as bottom-level features, including the original numerical value of the monitoring parameter, the basic change amplitude and other concrete information. The features reflecting the interaction and correlation law between different monitoring parameters are classified as middle-level features, including the parameter coupling strength, correlation response delay and other derived information. The features that can comprehensively reflect the overall geological environment state and evolution trend of the target region are classified as high-level features, including the regional geological situation, risk tendency and other macro information. The feature vectors in each level are sequentially arranged according to the category and time sequence order of the corresponding monitoring parameters, forming a multi-level feature sequence of the target region with clear structure, distinct levels and complete information.

[0060] For each level in the multi-level feature sequence, the adaptive fusion weight component corresponding to all feature vectors in the level is extracted one by one, establishing a one-to-one correspondence between the feature vector and the weight component, ensuring that each feature vector can match the weight corresponding to its representation meaning and importance. Through the weight assignment process, the role of key feature vectors in fusion is strengthened, and the influence of secondary feature vectors is weakened. Then, all feature vectors in the same level are subjected to deep feature interaction processing, through element-by-element information comparison, complementary information between different feature vectors is extracted, and redundant information content is eliminated. At the same time, the contribution proportion of each feature vector information is adjusted according to the size of the weight component, so that the fusion result can highlight the core information and take into account the auxiliary information. After each level is processed as above, an independent fusion result is formed, and the fusion results of all levels are arranged in order of bottom, middle and high levels to obtain a hierarchical fusion sequence of the target region.

[0061] The internal logical relationship between the fusion sequences at each level is thoroughly combed. It is clear that the underlying fusion sequence is the basic support for the entire data fusion, and the concrete information it contains provides the original data basis for subsequent fusion. The middle-level fusion sequence is the key link connecting the bottom and the top, and the parameter correlation law it carries can deepen the understanding of geological data. The high-level fusion sequence is a macroscopic control of the overall geological state, and the situation information it contains can enhance the decision-making value of the fused data. According to the order from the bottom to the top, cross-level aggregation is carried out. First, take the underlying fusion sequence as the basic framework, and accurately embed the core feature information reflecting the parameter correlation law in the middle-level fusion sequence into the corresponding time node and parameter category position of the underlying fusion sequence. Then, the macro feature information reflecting the overall geological situation in the high-level fusion sequence is covered in the sequence integrated by the middle and bottom layers, realizing the organic nesting and deep complementation of feature information at different levels. Finally, a coherent data set covering the underlying concrete data, the middle-level correlation law, and the high-level macro situation is formed, and the preliminary fusion data sequence of the target region is obtained.

[0062] The detailed comparison of the data at each time node in the preliminary fusion data sequence with the data at the adjacent time nodes is carried out, and the change difference between the adjacent data is calculated to determine whether the change difference is within the reasonable range of natural evolution of geological monitoring data. Referring to the change law of historical geological monitoring data in the target region, the data fluctuation characteristics in the same geological environment, and the physical evolution mechanism of geological processes, the normal change threshold of different monitoring parameters in unit time is determined. When the data change difference at a certain time node exceeds the normal change threshold, it is identified as a mutation data. At this time, based on the effective data of the adjacent time nodes, combined with the historical change trend of the monitoring parameter and the cooperative change law of related parameters, the mutation data is adjusted smoothly, so that the adjusted data not only conforms to the objective law of geological evolution, but also maintains the continuity with the previous and subsequent data. After adjusting all the mutation data one by one and checking the entire sequence data, the entire data sequence presents a continuous and smooth change trend in the time dimension, and finally the real-time fusion data of the target region that can accurately and truly reflect the geological monitoring state of the target region is formed.

[0063] The beneficial effect is that by hierarchical decomposition of the multi-dimensional real-time feature vector set, different dimensions, different importance levels of features can be accurately split, a structured multi-level feature sequence is formed, the processing limitation of mixing multi-dimensional features is broken, a clear feature hierarchical basis is provided for subsequent targeted fusion, and the fusion inefficiency problem caused by feature dimension cross and priority ambiguity is avoided; the multi-level feature sequence is interactively fused based on the adaptive fusion weight, so that the fusion process can accurately respond to the importance difference of different levels and different types of features, the key features with higher weight occupy the dominant position in the fusion, and the secondary features or interference features with lower weight are reasonably adapted, thereby ensuring the prominence of the core information; the hierarchical aggregation of the fusion sequence of each level is performed, the information barrier between levels is broken, the local feature correlation dispersed in different levels is logically integrated into a whole, a preliminary fusion data sequence covering the core information of geological monitoring is formed, the information omission caused by single-level fusion is effectively avoided, and the data can fully reflect the multi-dimensional correlation state of the target regional geological environment; the time domain coherence of the preliminary fusion data sequence is smoothed, the interference factors such as accidental fluctuation and device noise in the time series data are efficiently filtered, the change continuity and trend consistency of the data in the time dimension are maintained, the real-time fusion data obtained finally has both accuracy and time series stability, and can truly reflect the dynamic law of geological disaster evolution.

[0064] As Figure 2 shown, it is a functional module diagram of a reference information generation system based on artificial intelligence and smart home according to an embodiment of the present application.

[0065] The geological disaster monitoring data real-time fusion system 100 can be installed in an electronic device. According to the functions to be realized, the geological disaster monitoring data real-time fusion system 100 can include a data standardization module 101, a semantic relationship graph construction module 102, a space-time dual-dimensional data correction module 103, a time-frequency domain feature vectorization module 104, an adaptive weight optimization module 105, and a multi-level fusion output module 106. The modules of the present application can also be referred to as units, which refer to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, and are stored in the memory of the electronic device.

[0066] In the present embodiment, the functions of each module / unit are as follows: The data standardization module 101 is configured to perform heterogeneous adaptive analysis on multi-source geological monitoring data of a target region to obtain a standardized data sequence of the target region. The semantic relationship graph construction module 102 is configured to construct a dynamic semantic relationship graph of the target region according to the coupling relationship between the multiple monitoring parameters in the standardized data sequence. The space-time dual dimension correction module 103 is configured to correct the standardized data sequence based on the space-time constraint rule of the dynamic semantic relation graph to obtain a space-time consistent data set of the target region. The time-frequency domain feature vectorization module 104 is configured to vectorize and map time-frequency domain features in the space-time consistent data set to obtain a multi-dimensional feature vector set of the target region. The adaptive weight optimization module 105 is configured to cooperatively optimize real-time weights of the dynamic semantic relation graph and a data signal-to-noise ratio of the multi-dimensional feature vector set to obtain an adaptive fusion weight of the target region. The multi-level fusion output module 106 is configured to perform multi-level coupling on the multi-dimensional real-time feature vector set based on the adaptive fusion weight to obtain real-time fusion data of the target region.

[0067] In several embodiments provided in the present application, it should be understood that the disclosed method and system can be implemented in other ways. For example, the system embodiments described above are only illustrative, for example, the division of the modules is only a logical function division, and actual implementation can have another division mode.

[0068] The modules described as separate components can or can not be physically separated, and the components displayed as modules can or can not be physical units, that is, they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs.

[0069] In addition, each functional module in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of hardware plus software function modules.

[0070] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0071] The embodiment of the present application can acquire and process related data based on artificial intelligence technology. Artificial intelligence is the use of digital computers or digital computer controlled machines to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0072] Finally, it should be noted that the above examples are merely intended to illustrate the technical solutions of the present application and not to limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for real-time fusion of geological disaster monitoring data, characterized in that, The method includes: S1. Perform heterogeneous adaptation analysis on the multi-source geological monitoring data of the target area to obtain the standardized data sequence of the target area; S2. Based on the coupling relationship between multiple monitoring parameters in the standardized data sequence, construct a dynamic semantic relationship graph of the target region; S3. Based on the spatiotemporal constraint rules of the dynamic semantic relationship graph, the standardized data sequence is checked in both spatiotemporal dimensions to obtain a spatiotemporally consistent dataset of the target region; S4. Vectorize the time-frequency domain features in the spatiotemporally consistent dataset to obtain a multidimensional feature vector set of the target region; S5. The real-time weights of the dynamic semantic relationship graph and the data signal-to-noise ratio of the multi-dimensional feature vector set are optimized collaboratively to obtain the adaptive fusion weights of the target region; S6. Based on the adaptive fusion weights, the multi-dimensional real-time feature vector set is coupled in multiple levels to obtain the real-time fusion data of the target region.

2. The real-time fusion method for geological disaster monitoring data as described in claim 1, characterized in that, The heterogeneous adaptation and parsing of multi-source geological monitoring data in the target area yields a standardized data sequence for the target area, including: Collect multi-source geological monitoring data of the target area; The multi-source geological monitoring data are time-series calibrated to obtain time-aligned data for the target region; The time-aligned data is subjected to dimension normalization to obtain a dimensionless data set for the target region. The dimensionless data set is serialized and recombined to obtain a standardized data sequence for the target region.

3. The real-time fusion method for geological disaster monitoring data as described in claim 1, characterized in that, The step of constructing a dynamic semantic relationship graph of the target region based on the coupling relationships among multiple monitoring parameters in the standardized data sequence includes: The coupling characteristics between multiple monitoring parameters in the standardized data sequence are extracted to obtain a candidate set of relationships between the multiple monitoring parameters. The candidate set of relationships is then used to construct a topology to obtain the initial relationship network structure of the target region. By applying dynamic time window constraints to the initial relational network structure, a time-varying semantic network for the target region is obtained. The time-varying semantic network is subjected to rule validation to obtain the validated semantic network for the target region; The verified semantic network is topologically reconstructed to obtain a dynamic semantic relationship graph of the target region.

4. The real-time fusion method for geological disaster monitoring data as described in claim 3, characterized in that, The step of applying dynamic time window constraints to the initial relational network structure to obtain the time-varying semantic network of the target region includes: The continuous time axis of the standardized data sequence is divided into sliding time windows to obtain the continuous time window sequence of the target region; Based on the continuous time window sequence, the standardized data sequence is segmented and sampled to obtain a windowed data subset of the target region; Based on the windowed data subset, the node connection positions in the initial relational network structure are matched one by one to obtain the instantaneous relational subgraph of the target region; The instantaneous relational subgraph is dynamically linked to obtain the time-varying semantic network of the target region.

5. The real-time fusion method for geological disaster monitoring data as described in claim 1, characterized in that, The spatiotemporal constraint rules based on the dynamic semantic relationship graph are used to perform spatiotemporal dual-dimensional calibration on the standardized data sequence to obtain a spatiotemporally consistent dataset for the target region, including: Spatial dimension verification is performed on the standardized data sequence to obtain spatially consistent data of the standardized data sequence. By fusing the standardized data sequence with data consistent with the spatial dimension, spatiotemporal aligned data of the standardized data sequence is obtained; The dynamic semantic relationship graph is subjected to association constraint analysis to obtain the spatiotemporal constraint rules of the dynamic semantic relationship graph; Based on the spatiotemporal constraint rules, constraint-driven interpolation is performed on the spatiotemporal aligned data to obtain a spatiotemporally consistent dataset for the target region.

6. The real-time fusion method for geological disaster monitoring data as described in claim 1, characterized in that, The step of vectorizing the time-frequency domain features in the spatiotemporally consistent dataset to obtain a multidimensional feature vector set for the target region includes: Multi-scale feature fusion is performed on the temporal and frequency domain features in the spatiotemporal consistent dataset to obtain the fused feature set of the spatiotemporal consistent dataset; The feature dimensions of the spatiotemporally consistent dataset and the fused feature set are interactively coupled to obtain the embedding vector set of the target region; The embedded vector set is subjected to feature semantic normalization to obtain the semantically clear vector set of the spatiotemporally consistent dataset; Topological integration is performed on the vector structure of the semantically clear vector group to obtain a multidimensional feature vector set of the target region.

7. The real-time fusion method for geological disaster monitoring data as described in claim 1, characterized in that, The step of collaboratively optimizing the real-time weights of the dynamic semantic relationship graph and the data signal-to-noise ratio of the multi-dimensional feature vector set to obtain the adaptive fusion weights for the target region includes: The real-time weights of the dynamic semantic relationship graph are correlated and mapped with the data signal-to-noise ratio of the multi-dimensional feature vector set to obtain the initial matching relationship of the target region; Based on the initial matching relationship, the real-time weights and the data signal-to-noise ratio are jointly evaluated to obtain the collaborative evaluation index of the target region; Based on the collaborative evaluation index, a strategy calculation is performed on the spatiotemporally consistent dataset to obtain the adaptive fusion weight of the target region.

8. The real-time fusion method for geological disaster monitoring data as described in claim 7, characterized in that, The step of performing policy calculations on the spatiotemporally consistent dataset based on the collaborative evaluation index to obtain the adaptive fusion weights for the target region includes: The collaborative evaluation indicators and the dynamic semantic relationship graph are analyzed collaboratively to obtain the mapping rule set for the target region; Based on the mapping rule set, the real-time weights and the data signal-to-noise ratio are mapped using rule application to obtain the candidate weight set for the target region; A consistency check is performed on the candidate weight set to obtain the effective weights for the target region. The effective weights are adapted and optimized to obtain the adaptive fusion weights for the target region. The calculation formula for the adaptive fusion weights is as follows: ； In the formula, The adaptive fusion weight components, The first of the real-time weights One portion, The first of the collaborative evaluation indicators The normalized values ​​of each component The first of the data signal-to-noise ratios One portion, The adjustment index for the real-time weights, This is the adjustment index for the aforementioned collaborative evaluation index. The adjustment index for the signal-to-noise ratio of the data is given. To prevent extremely small positive numbers with a denominator of zero, The total number of the multiple monitoring parameters, The first of the real-time weights One portion, The first of the collaborative evaluation indicators The normalized values ​​of each component The first of the data signal-to-noise ratios Each component.

9. The real-time fusion method for geological disaster monitoring data as described in claim 1, characterized in that, The step of coupling the multi-dimensional real-time feature vector set at multiple levels based on the adaptive fusion weights to obtain the real-time fused data of the target region includes: The multidimensional real-time feature vector set is hierarchically deconstructed to obtain a multi-level feature sequence of the target region; Based on the adaptive fusion weights, feature interaction fusion is performed on the multi-level feature sequences to obtain the hierarchical fusion sequence of the target region; The fusion sequences at each level are aggregated across levels to obtain a preliminary fusion data sequence for the target region. Temporal coherence smoothing is performed on the preliminary fused data sequence to obtain real-time fused data for the target region.

10. A real-time fusion system for geological disaster monitoring data, characterized in that, The system for implementing the real-time fusion method for geological disaster monitoring data as described in claim 1 includes: The data standardization module is used to perform heterogeneous adaptation and parsing on multi-source geological monitoring data of the target area to obtain a standardized data sequence of the target area. A semantic relationship graph construction module is used to construct a dynamic semantic relationship graph of the target region based on the coupling relationship between multiple monitoring parameters in the standardized data sequence. The spatiotemporal dual-dimensional data verification module is used to perform spatiotemporal dual-dimensional verification on the standardized data sequence based on the spatiotemporal constraint rules of the dynamic semantic relationship graph, so as to obtain the spatiotemporal consistent dataset of the target region. The time-frequency domain feature vectorization module is used to vectorize the time-frequency domain features in the spatiotemporally consistent dataset to obtain a multidimensional feature vector set of the target region. An adaptive weight optimization module is used to collaboratively optimize the real-time weights of the dynamic semantic relationship graph and the data signal-to-noise ratio of the multi-dimensional feature vector set to obtain the adaptive fusion weights of the target region. A multi-level fusion output module is used to perform multi-level coupling on the multi-dimensional real-time feature vector set based on the adaptive fusion weights to obtain real-time fused data of the target region.

Images