A multi-source data coupling system and method

Abstract

The present application relates to the technical field of multi-source data processing, in particular to a multi-source data coupling system and method, comprising: obtaining target area original monitoring data from multiple data sources, quality evaluating each data source data and generating quality scores; combining geological activity stage parameters to establish a data source contribution dynamic weight distribution model, calculating real-time dynamic contribution weights of each data source and weighting and fusing the original data; introducing geological prior knowledge to perform adaptive feature screening on the fused data, retaining key features and performing collaborative analysis with the geological prior knowledge to obtain a geological state coupling result; constructing a closed-loop optimization mechanism, using the coupling result to calibrate the data quality scores and update the weight model for the next cycle fusion. The method can adapt to dynamic changes of geological activities, improve the reliability of the fused data, realize adaptive coupling optimization of multi-source monitoring data, and adapt to complex geological monitoring scenarios.

Description

Technical Field

[0001] This invention relates to the field of multi-source data processing technology, and in particular to a multi-source data coupling system and method. Background Technology

[0002] In the field of geological monitoring, multi-source data coupling is an important means to achieve accurate analysis of the geological status of a target area. Currently, the industry generally adopts a multi-source monitoring data fusion technology, which obtains raw monitoring data from multiple data sources, performs simple weighted fusion, and then analyzes it in conjunction with relevant geological information to obtain geological status assessment results. In existing technologies, weight allocation often adopts a preset fixed weight method or determines the contribution weight of each data source based solely on a single data quality indicator, without considering the impact of dynamic changes in the geological activity stages of the target area on the effectiveness of the data sources.

[0003] Existing technical solutions have shortcomings. Fixed weights or single-indicator weight allocation methods cannot adapt to the monitoring needs of different stages of geological activity. As a result, the role of some high-value data sources is not fully utilized, and the interference from low-quality data sources is difficult to suppress effectively, affecting the reliability of fused data. At the same time, the data coupling process is mostly a one-way processing flow. The results of the fusion analysis cannot be used to evaluate data quality and allocate weights in reverse. This makes it difficult to iteratively optimize the coupling effect with the monitoring cycle, making it difficult to adapt to complex and ever-changing geological monitoring scenarios and unable to achieve adaptive coupling of multi-source data. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a multi-source data coupling system and method.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a multi-source data coupling method, comprising:

[0006] Raw monitoring data of the target area is obtained from multiple data sources, a raw monitoring data quality assessment unit is constructed, the quality of raw monitoring data from each data source is assessed, and a data quality score for each data source is generated.

[0007] Based on the data quality scores of each data source and combined with the geological activity stage parameters of the target area, a dynamic weight allocation model for the contribution of data sources is established to calculate the real-time dynamic contribution weight of each data source.

[0008] Based on the real-time dynamic contribution weights, the original monitoring data from each data source are weighted and fused to generate a weighted fused monitoring dataset.

[0009] Introducing prior geological knowledge of the target area, including lithological distribution characteristics and historical deformation patterns, adaptive feature filtering processing is performed on the weighted fusion monitoring dataset;

[0010] Through adaptive feature filtering, key feature monitoring data are retained to form a filtered feature monitoring dataset;

[0011] The filtered feature monitoring dataset is combined with the geological prior knowledge for collaborative analysis to generate coupled analysis results of the geological state of the target area.

[0012] A closed-loop optimization mechanism is constructed, and the data quality scores of each data source are recalibrated using the geological state coupling analysis results, and the dynamic weight allocation model of the contribution of the data source is updated.

[0013] The updated dynamic weighting model for data source contribution is used for weighted fusion of the original monitoring data in the next cycle, thereby achieving adaptive coupling optimization of multi-source monitoring data.

[0014] As a further aspect of the present invention, obtaining the original monitoring data of the target area from multiple data sources specifically includes:

[0015] The original monitoring data includes topographic change data, geological deformation time series data, lithological distribution data, and historical activity record data;

[0016] Topographic change data of the target area is obtained from the first type of data source, the topographic change data including digital elevation model difference data collected at different time points;

[0017] Geological deformation time series data of the target area are obtained from the second type of data source, wherein the geological deformation time series data is a time series of surface displacement continuously collected through a sensor network;

[0018] Lithological distribution data of the target area is obtained from a third type of data source, wherein the lithological distribution data is a spatial distribution map of rock types obtained through geological exploration;

[0019] Historical activity records of the target area are obtained from the fourth type of data source. The historical activity records include spatiotemporal information of historical earthquake events, landslide events, and fault activity records.

[0020] The topographic change data, geological deformation time series data, lithological distribution data, and historical activity record data are registered and aligned according to a unified spatiotemporal benchmark to form the initial raw monitoring dataset.

[0021] As a further aspect of the present invention, the construction of the original monitoring data quality assessment unit, which assesses the quality of the original monitoring data from each of the data sources and generates a data quality score for each data source, specifically involves:

[0022] For terrain change data, assess its spatial resolution, temporal coverage integrity, and elevation measurement accuracy to generate a terrain data quality sub-score;

[0023] For geological deformation time series data, the continuity of the time series, the proportion of missing values, and the level of sensor acquisition noise are evaluated to generate a deformation data quality sub-score;

[0024] For lithological distribution data, assess its exploration point density, lithological classification, and spatial interpolation uncertainty, and generate a lithological data quality sub-score.

[0025] For historical activity record data, assess the completeness of recorded events, spatiotemporal location, and reliability of event intensity descriptions to generate historical data quality sub-scores;

[0026] By combining the quality sub-scores corresponding to each data source, a comprehensive data quality score for each data source is calculated using a preset score aggregation algorithm.

[0027] As a further aspect of the present invention, the dynamic weight allocation model for the contribution of data sources is established based on the data quality scores of each data source and combined with the geological activity stage parameters of the target area, and the real-time dynamic contribution weight of each data source is calculated, specifically as follows:

[0028] Obtain geological activity stage parameters that characterize the current intensity of geological activity in the target area;

[0029] A basic weighting factor is set for the dynamic weighting model of data source contribution, and the basic weighting factor is positively correlated with the comprehensive data quality score of each data source;

[0030] In the dynamic weight allocation model for the contribution of the data source, an adjustment coefficient for the geological activity stage parameter is introduced. When the geological activity stage parameter indicates an intensification of activity, the basic weight factor of the data source that is sensitive to time-series changes is increased, and vice versa.

[0031] The comprehensive data quality score, basic weight factor, and adjustment coefficient of geological activity stage parameters are combined to calculate the real-time dynamic contribution weight of each data source in the coupled analysis process, and ensure that the sum of the real-time dynamic contribution weights of all data sources is one.

[0032] As a further aspect of the present invention, based on the real-time dynamic contribution weight, the original monitoring data from each data source are weighted and fused to generate a weighted fused monitoring dataset, specifically as follows:

[0033] Read the real-time dynamic contribution weights corresponding to topographic change data, geological deformation time series data, lithological distribution data, and historical activity record data;

[0034] The value of each spatial raster cell in the terrain change data is multiplied by its corresponding real-time dynamic contribution weight to obtain the weighted terrain change data.

[0035] Multiply the value of each time series data point in the geological deformation time series data by its corresponding real-time dynamic contribution weight to obtain the weighted geological deformation time series data.

[0036] The distribution probability of each lithology type in the lithology distribution data is multiplied by its corresponding real-time dynamic contribution weight to obtain the weighted lithology distribution data;

[0037] The intensity and impact factor of each historical event in the historical activity record data are multiplied by its corresponding real-time dynamic contribution weight to obtain the weighted historical activity record data.

[0038] The weighted topographic change data, geological deformation time series data, lithological distribution data, and historical activity record data are overlaid and spatially fused to generate a unified weighted fusion monitoring dataset.

[0039] As a further aspect of the present invention, the introduction of prior geological knowledge of the target area, including lithological distribution characteristics and historical deformation patterns, and the application of adaptive feature filtering processing to the weighted fused monitoring dataset are specifically as follows:

[0040] The lithological distribution characteristics of the target area are retrieved from the geological prior knowledge base. These lithological distribution characteristics identify the spatial distribution of different rock types and their physical and mechanical properties.

[0041] The historical deformation patterns of the target area are retrieved from the geological prior knowledge base. These historical deformation patterns describe the typical deformation patterns and spatial distribution characteristics of the target area in past geological activities.

[0042] Set the initial global threshold for adaptive feature selection;

[0043] Based on the lithological distribution characteristics, the initial global threshold is dynamically adjusted by weighting and fusing monitoring data from different lithological zones. A low feature screening threshold is used in lithologically unstable areas, while a high feature screening threshold is used in stable areas.

[0044] Based on the aforementioned historical deformation patterns, in areas with historically active deformation, the feature selection threshold for monitoring features in the weighted fusion monitoring data that conform to the historical deformation patterns is further lowered and these features are prioritized for retention.

[0045] Based on the dynamically adjusted feature screening threshold, each monitoring feature data in the weighted fusion monitoring dataset is screened. Data below the threshold is regarded as noise or secondary features and filtered out, while data above the threshold is retained as key feature monitoring data, thus forming the screened feature monitoring dataset.

[0046] As a further aspect of the present invention, the filtered feature monitoring dataset is collaboratively analyzed with the geological prior knowledge to generate a coupled analysis result of the geological state of the target area, specifically as follows:

[0047] Spatial correlation analysis was performed between the filtered feature monitoring dataset and the lithological distribution characteristics in geological prior knowledge to identify the spatial correlation between the monitoring features and specific lithologies.

[0048] The filtered feature monitoring dataset is matched with historical deformation patterns in geological prior knowledge in a spatiotemporal pattern to determine whether the current monitoring features are consistent with known historical deformation patterns.

[0049] Based on the results of integrated spatial correlation analysis and spatiotemporal pattern matching, the potential intensity of geological activity at different spatial locations within the target area is assessed.

[0050] Based on the spatial distribution of the potential intensity of geological activity, and combined with the spatiotemporal evolution trend of the feature values ​​in the filtered feature monitoring data, a coupled analysis result of geological state is generated, which includes the location of potential geological risk areas, risk level, and activity pattern prediction.

[0051] As a further aspect of the present invention, the construction of the closed-loop optimization mechanism, which utilizes the geological state coupling analysis results to recalibrate the data quality scores of each data source and update the dynamic weight allocation model for the contribution of the data sources, specifically involves:

[0052] In the results of the geological state coupling analysis, high-confidence analysis areas were identified that were subsequently verified as accurate by on-site inspections or independent monitoring methods.

[0053] The role of the raw monitoring data provided by each data source was traced back during the coupled analysis process in the high-confidence analysis area;

[0054] Based on the analytical contribution of each data source in the high-confidence analysis region, the initial comprehensive data quality score is adjusted. Data sources with large contributions have their data quality scores increased, while data sources with small contributions or that provide interfering information have their data quality scores decreased.

[0055] Using the corrected data quality scores of each data source, the basic weight factors in the dynamic weight allocation model of the data source contribution are recalculated;

[0056] The recalculated basic weighting factors are substituted into the dynamic weighting model of the data source contribution to generate an updated dynamic weighting model of the data source contribution, which is used to weight and fuse the newly acquired raw monitoring data.

[0057] As a further aspect of the present invention, it also includes a step of quantifying and transferring uncertainty in the original monitoring data before weighted fusion, specifically:

[0058] For the raw monitoring data of each data source, based on its data acquisition principle, processing flow and the data quality score obtained from the evaluation, its data uncertainty is quantified, and an uncertainty spatial distribution map of each data source is generated.

[0059] When weighting and fusing the raw monitoring data from each data source, the uncertainty spatial distribution maps corresponding to each data source are simultaneously weighted and fused according to their respective real-time dynamic contribution weights.

[0060] Generate a fusion uncertainty distribution map corresponding to the space of the weighted fusion monitoring dataset;

[0061] When performing adaptive feature filtering, the fused uncertainty distribution map is used as an auxiliary criterion. A stricter feature filtering threshold is applied to monitoring features from high uncertainty regions, while a relatively lenient feature filtering threshold is applied to monitoring features from low uncertainty regions.

[0062] As a further aspect of the present invention, the present invention also includes a multi-source data coupling system, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, implements the steps of the multi-source data coupling method described above.

[0063] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0064] Based on the quality scores of the raw monitoring data from each data source, a dynamic weight allocation model for the contribution of each data source is established by combining parameters of the geological activity stage in the target area. This model calculates the real-time dynamic contribution weight of each data source. This dynamic weight allocation method can adapt to the characteristic differences of different stages of geological activity, adjust the contribution ratio of each data source according to the real-time status of geological activity, and make the weight allocation more in line with actual monitoring needs. It effectively avoids the limitations of fixed weights or single indicator weight allocation, making the weighted fused monitoring dataset more reflective of the true geological conditions of the target area and improving the relevance and reliability of the fused data.

[0065] A closed-loop optimization mechanism is constructed, using the generated geological state coupling analysis results of the target area to recalibrate the data quality scores of each data source. Based on the calibrated scores, the dynamic weight allocation model for data source contribution is updated, and the updated model is then applied to the weighted fusion of the original monitoring data in the next cycle. This cyclical optimization approach enables adaptive adjustment of the data coupling process. As the monitoring cycle progresses, it continuously corrects data quality assessment biases and unreasonable weight allocations, gradually optimizing the coupling effect. This makes subsequent feature selection and collaborative analysis more aligned with geological realities and adaptable to the dynamic changes in complex geological monitoring scenarios. Attached Figure Description

[0066] Figure 1 This is a flowchart of a multi-source data coupling method according to the present invention;

[0067] Figure 2 The flowchart for acquiring and registering raw monitoring data;

[0068] Figure 3 A flowchart for dynamically assigning weights to the contribution of data sources;

[0069] Figure 4 Analysis of temporal deformation trends in different risk zones;

[0070] Figure 5 This is a spatial distribution map of terrain change data. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0072] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0073] See Figure 1This study acquires raw monitoring data of the target area from multiple data sources, constructs a raw monitoring data quality assessment unit, evaluates the quality of raw monitoring data from each data source, and generates a data quality score for each data source. Based on the data quality scores of each data source and combined with the geological activity stage parameters of the target area, a dynamic weight allocation model for data source contribution is established to calculate the real-time dynamic contribution weight of each data source. According to the calculated real-time dynamic contribution weight, the raw monitoring data from each data source are weighted and fused to generate a unified weighted fused monitoring dataset. Geological prior knowledge of the target area, including lithological distribution characteristics and historical deformation patterns, is introduced to perform adaptive feature filtering processing on the weighted fused monitoring dataset. This processing retains key feature monitoring data, forming a filtered feature monitoring dataset. The filtered feature monitoring dataset is then collaboratively analyzed with geological prior knowledge to generate a geological state coupling analysis result for the target area. A closed-loop optimization mechanism is constructed to recalibrate the data quality scores of each data source using the generated geological state coupling analysis result, thereby updating the dynamic weight allocation model for data source contribution. The updated model is used for weighted fusion of raw monitoring data in the next cycle, achieving adaptive coupling optimization of multi-source monitoring data.

[0074] See Figure 2 In one embodiment of the present invention, the raw monitoring data includes topographic change data, geological deformation time-series data, lithological distribution data, and historical activity record data. Topographic change data of the target area is obtained from a first type of data source, which includes digital elevation model difference data collected at different time points. Geological deformation time-series data of the target area is obtained from a second type of data source, which is a continuous time series of surface displacements collected through a sensor network. Lithological distribution data of the target area is obtained from a third type of data source, which is a spatial distribution map of rock types obtained through geological exploration. Historical activity record data of the target area is obtained from a fourth type of data source, which includes spatiotemporal information of historical earthquake events, landslide events, and fault activity records. After obtaining the above data, the topographic change data, geological deformation time-series data, lithological distribution data, and historical activity record data are registered and aligned according to a unified spatiotemporal benchmark to form an initial raw monitoring dataset.

[0075] In practical implementation, an example scenario of geological hazard monitoring on a slope in a mountainous area in Southwest China can be used as an illustration. The original monitoring data includes topographic change data, geological deformation time series data, lithological distribution data, and historical activity records. Topographic change data of the target area is obtained from the first type of data source. The topographic change data includes digital elevation models (DEMs) at different time points generated by multiple periods of aerial photogrammetry. By calculating the elevation difference between corresponding grids of two adjacent DEMs, DEM difference data reflecting changes in landform are generated. In the example scenario, the DEM difference data between October 2025 and January 2026 shows that there is a significant subsidence area in the upper part of the slope.

[0076] In some embodiments, geological deformation time-series data of the target area is obtained from a second type of data source. This data consists of continuous surface displacement time series collected by GPS sensors and fissure gauges deployed on the slope. The GPS sensors provide a three-dimensional coordinate sequence, and the fissure gauges record the fissure width variation sequence. In the example scenario, data continuously collected by the sensor network from November 2025 to February 2026 constitutes a set of geological deformation time-series data containing hundreds of monitoring points with a time resolution down to the hour. Lithological distribution data of the target area is obtained from a third type of data source. This data is a spatial distribution map of rock types obtained through previous drilling, ground-penetrating radar detection, and outcrop surveys. The map is stored in a polygonal vector format, marking the spatial range of different rock and soil bodies such as strongly weathered sandstone, mudstone, and sedimentary layers. In the example scenario, the lithological distribution map reveals that the monitored slope is mainly composed of upper strongly weathered sandstone and underlying mudstone. Historical activity records for the target area are obtained from the fourth type of data source. These records come from local chronicles, seismic network records, and geological disaster survey reports. They include spatiotemporal information on historically recorded earthquake events, landslide events, and fault activity records. In the example scenario, historical activity records show that the slope experienced a surface landslide during the rainy season in 1998, and its affected area spatially overlaps with the current deformation area.

[0077] Optionally, the raw data obtained from the above four types of data sources differ in spatial and temporal references. Topographic change data is based on local independent coordinate systems, geological deformation time-series data is based on the WGS84 coordinate system, lithology distribution maps are based on the National Geodetic Coordinate System 2000, and historical event records use textual descriptions of relative locations and Gregorian calendar dates. It is understandable that registering and aligning topographic change data, geological deformation time-series data, lithology distribution data, and historical activity record data according to a unified spatiotemporal reference is a necessary step in forming the initial raw monitoring dataset. In practice, spatial benchmark unification is achieved through coordinate transformation, converting all data to the National Geodetic Coordinate System 2000. Temporal benchmark unification uses the Gregorian calendar as the standard, converting historical lunar or descriptive timestamps into unified Gregorian timestamps. For geological deformation time-series data with time-series characteristics, each data point is associated with precise UTC time. The registration process involves converting and resampling all data—including digital elevation model difference data raster, GNSS monitoring point locations, lithological polygon boundaries, and historical event location descriptions—to a single spatial grid framework with clearly defined geographic coordinates, ultimately forming a spatiotemporally consistent initial raw monitoring dataset. In some embodiments, the accuracy of spatial registration is evaluated and controlled by the following formula:

[0078] ;

[0079] in: This represents the mean square error of the overall planar position after registration. This represents the coordinate residual in the east direction. This represents the coordinate residual in the north direction. In practice, this is achieved by controlling the number and distribution of ground control points. The value is no greater than the size of 1 pixel.

[0080] See Figure 3In one embodiment of the present invention, for topographic change data, its spatial resolution, temporal span coverage integrity, and elevation measurement accuracy are evaluated to generate a topographic data quality sub-score. For geological deformation time series data, its time series continuity, missing value ratio, and sensor acquisition noise level are evaluated to generate a deformation data quality sub-score. For lithology distribution data, its exploration point density, lithology classification accuracy, and spatial interpolation uncertainty are evaluated to generate a lithology data quality sub-score. For historical activity record data, its completeness of recorded events, spatiotemporal positioning accuracy, and reliability of event intensity description are evaluated to generate a historical data quality sub-score. The quality sub-scores corresponding to each data source are combined, and a preset scoring aggregation algorithm is used to calculate the comprehensive data quality score for each data source. Geological activity stage parameters characterizing the current intensity of geological activity in the target area are obtained. A basic weight factor for the dynamic weight allocation model of data source contribution is set, and the basic weight factor is positively correlated with the comprehensive data quality score of each data source. In the dynamic weight allocation model for data source contributions, an adjustment coefficient for geological activity stage parameters is introduced. When the geological activity stage parameters indicate intensified activity, the base weight factor of data sources sensitive to time-series changes is increased, and vice versa. The comprehensive data quality score, base weight factor, and adjustment coefficient of geological activity stage parameters are calculated simultaneously to obtain the real-time dynamic contribution weight of each data source in the coupled analysis process, ensuring that the sum of the real-time dynamic contribution weights of all data sources is equal.

[0081] In practical implementation, taking the aforementioned slope monitoring scenario in the southwestern mountainous area as an example, quality assessments are conducted on the acquired and registered topographic change data, geological deformation time-series data, lithological distribution data, and historical activity record data. For topographic change data, its spatial resolution, temporal span coverage integrity, and elevation measurement accuracy are evaluated, generating a sub-score for topographic data quality. Spatial resolution is assessed based on the grid size of the digital elevation model; temporal span coverage integrity is assessed based on whether the available data periods cover the entire monitoring cycle; and elevation measurement accuracy is assessed based on the calibration error of aerial surveys. For geological deformation time-series data, its time series continuity, missing value ratio, and sensor acquisition noise level are evaluated, generating a sub-score for deformation data quality. Time series continuity is measured by calculating the proportion of valid data acquisition time to the total duration; the missing value ratio is calculated by the proportion of invalid or interrupted data points in the sequence; and the sensor acquisition noise level is determined by analyzing the standard deviation of monitoring data under stable conditions.

[0082] In some embodiments, for lithological distribution data, the density of exploration points, lithological classification accuracy, and spatial interpolation uncertainty are evaluated to generate a lithological data quality sub-score. Exploration point density is calculated based on the number of effective boreholes or geophysical points per square kilometer. Lithological classification accuracy is assessed by comparing the consistency between lithological identification results at field verification points and map annotation results. Spatial interpolation uncertainty is quantified by calculating the kriging variance of lithological inferences for unexplored areas using geostatistical methods. For historical activity record data, the completeness of recorded events, spatiotemporal positioning accuracy, and reliability of event intensity descriptions are evaluated to generate a historical data quality sub-score. The completeness of recorded events is assessed by verifying the records of the same event from multiple independent historical sources. Spatiotemporal positioning accuracy is judged based on the deviation between the location described in the record and the location of known geological bodies. The reliability of event intensity descriptions is assessed based on the quantification and consistency of the descriptions of event scale and impact range in the record. The overall data quality score for each data source is calculated by combining the quality sub-scores corresponding to each data source using a pre-defined scoring aggregation algorithm. The scoring aggregation algorithm can employ a weighted geometric average or a rule fusion method based on fuzzy logic.

[0083] Optionally, geological activity stage parameters characterizing the intensity of current geological activity in the target area can be obtained. These parameters can be quantitative indicators such as the frequency of earthquakes per unit time calculated through a deployed microseismic monitoring network, the average tilt rate monitored by a surface tiltmeter array, or the spatial standard deviation of the surface deformation rate field retrieved through remote sensing. A basic weight factor is set for the dynamic weight allocation model of data source contribution. This basic weight factor is positively correlated with the comprehensive data quality score of each data source. An adjustment coefficient for the geological activity stage parameters is introduced into the model. When the geological activity stage parameters indicate intensified activity, the basic weight factor of the data source sensitive to temporal changes is increased; conversely, it is decreased. It can be understood that data sources sensitive to temporal changes typically include sensor networks providing temporal data on geological deformation and aerial photography systems providing multi-period topographic change data. In the example scenario, if the monitored monthly average deformation rate exceeds a preset threshold, it is determined that the geological activity stage parameters indicate intensified activity, and the basic weight factor corresponding to the geological deformation temporal data will be increased.

[0084] In some embodiments, the comprehensive data quality score, the basic weighting factor, and the adjustment coefficients of the geological activity stage parameters are simultaneously calculated to obtain the real-time dynamic contribution weight of each data source in the coupled analysis process, ensuring that the sum of the real-time dynamic contribution weights of all data sources is equal. The calculation of the real-time dynamic contribution weight can be expressed by the following formula:

[0085] ;

[0086] in: Indicates the first Real-time dynamic contribution weight of each data source Indicates the first A comprehensive data quality score for each data source. This represents the normalized parameters of the geological activity stages. It is a sensitivity coefficient related to parameters of geological activity stages, used to adjust parameters. The strength of the influence on the weights, For the total number of data sources involved in the coupling, the denominator ensures that the sum of all weights is 1. When When the value increases, it indicates that the activity is intensifying. Increasing the value will make the original Higher values ​​and data sources sensitive to activity (such as geological deformation time series data) yield higher [values]. value.

[0087] In one embodiment of the present invention, real-time dynamic contribution weights corresponding to topographic change data, geological deformation time-series data, lithological distribution data, and historical activity record data are read. The value of each spatial raster cell in the topographic change data is multiplied by its corresponding real-time dynamic contribution weight to obtain weighted topographic change data. The value of each time series data point in the geological deformation time-series data is multiplied by its corresponding real-time dynamic contribution weight to obtain weighted geological deformation time-series data. The distribution probability of each lithological type in the lithological distribution data is multiplied by its corresponding real-time dynamic contribution weight to obtain weighted lithological distribution data. The intensity and influence factor of each historical event in the historical activity record data are multiplied by its corresponding real-time dynamic contribution weight to obtain weighted historical activity record data. All weighted topographic change data, geological deformation time-series data, lithological distribution data, and historical activity record data are overlaid and spatially fused to generate a unified weighted fused monitoring dataset. The lithological distribution characteristics of the target area are retrieved from a geological prior knowledge base; these lithological distribution characteristics identify the spatial distribution and physical and mechanical properties of different rock types. Historical deformation patterns of the target area are retrieved from a geological prior knowledge base. These patterns describe typical deformation modes and spatial distribution characteristics of the target area during past geological activities. An initial global threshold for adaptive feature selection is set. Based on lithological distribution characteristics, the initial global threshold is dynamically adjusted for weighted fusion monitoring data from different lithological zones. A lower feature selection threshold is used in lithologically unstable areas, while a higher threshold is used in stable areas. Considering historical deformation patterns, in historically active deformation areas, monitoring features in the weighted fusion monitoring data that conform to historical deformation patterns are further reduced in the feature selection threshold and prioritized for retention. Based on the dynamically adjusted feature selection threshold, each monitoring feature data in the weighted fusion monitoring dataset is filtered. Data below the threshold is considered noise or secondary features and filtered out, while data above the threshold is retained as key feature monitoring data, forming a filtered feature monitoring dataset.

[0088] In practical implementation, the monitoring scenario targeting a slope in a mountainous area in Southwest China continues to be used. Real-time dynamic contribution weights are read from topographic change data, geological deformation time-series data, lithological distribution data, and historical activity records. These weights are calculated by a dynamic weight allocation model of data source contribution. For each spatial raster cell in the topographic change data, the value is multiplied by its corresponding real-time dynamic contribution weight to obtain weighted topographic change data. For example, if a raster value in the digital elevation model difference data is -0.15 meters, and its data source's real-time dynamic contribution weight is 0.3, then the weighted value of that raster is -0.045 meters. Similarly, for each time series data point in the geological deformation time-series data, the value is multiplied by its corresponding real-time dynamic contribution weight to obtain weighted geological deformation time-series data. For example, if the subsidence rate of a certain global navigation satellite system monitoring point at a specific time is 5 mm / day, and its data source's real-time dynamic contribution weight is 0.4, then the weighted subsidence rate is 2 mm / day.

[0089] In some embodiments, the probability of each lithological type distribution in the lithological distribution data is multiplied by its corresponding real-time dynamic contribution weight to obtain weighted lithological distribution data. For example, if the probability of a loose sedimentary layer at a certain spatial location is 0.8, and the real-time dynamic contribution weight of its data source is 0.25, then the weighted probability value is 0.2. The intensity and influence factor of each historical event in the historical activity record data are multiplied by their corresponding real-time dynamic contribution weight to obtain weighted historical activity record data. For example, if the intensity index of a historical landslide event is 8, and the real-time dynamic contribution weight of its data source is 0.15, then the weighted intensity index is 1.2. All weighted topographic change data, geological deformation time series data, lithological distribution data, and historical activity record data are overlaid and spatially fused to generate a unified weighted fused monitoring dataset. This dataset integrates all monitoring information from different sources, with different physical meanings, but after weight normalization. Each spatial location or spatiotemporal unit contains a fused multidimensional feature vector.

[0090] Optionally, lithological distribution characteristics of the target area can be retrieved from a geological prior knowledge base. These characteristics identify the spatial distribution and physical and mechanical properties of different rock types. For example, in the slope area of ​​the example scenario, prior knowledge identifies the slope toe as relatively intact mudstone, while the upper and middle parts of the slope are strongly weathered sandstone. Historical deformation patterns of the target area can also be retrieved from the geological prior knowledge base. These patterns describe typical deformation patterns and spatial distribution characteristics of the target area during past geological activities. For example, historical records show that landslides in this area often exhibit a traction-type pattern of rear-edge fracturing and front-edge bulging, primarily occurring along specific weak interlayers. An initial global threshold for adaptive feature filtering is set; this threshold serves as a benchmark value for judging the significance of monitoring features. Combining lithological distribution characteristics, the weighted fusion monitoring data from different lithological zones is dynamically adjusted, using lower feature filtering thresholds in unstable lithological areas and higher thresholds in stable areas. For example, in the strongly weathered sandstone distribution area, the initial global threshold is lowered by 30%, while in the intact mudstone area, the threshold is increased by 20%.

[0091] It is understandable that, considering historical deformation patterns, in areas with historically active deformation, the feature selection threshold for monitoring features in the weighted fusion monitoring data that conform to historical deformation patterns is further lowered for priority retention. For example, in areas where tensile fracturing has occurred historically, if the current fusion data also shows similar tensile deformation characteristics, the feature selection threshold for that local area is further reduced by a certain percentage on top of the lithological adjustment. Based on the dynamically adjusted feature selection threshold, each monitoring feature data in the weighted fusion monitoring dataset is filtered, with data below the threshold considered noise or secondary features and filtered out, while data above the threshold is retained as key feature monitoring data, forming a filtered feature monitoring dataset. In some embodiments, the calculation of the dynamically adjusted feature selection threshold can be expressed as:

[0092] ;

[0093] in: Indicates spatial location The feature selection threshold for the final application. This represents the initial global threshold. This represents the adjustment factor based on the lithological stability assessment at this location (less than 1 in unstable areas, greater than 1 in stable areas). This represents the adjustment coefficient based on the matching degree between the historical deformation pattern at that location and the current data pattern (less than 1 for a high matching degree, and equal to 1 otherwise). The feature value of each location in the weighted fusion monitoring dataset will be compared with the calculated... By making comparisons, adaptive filtering can be achieved.

[0094] In one embodiment of the invention, a spatial correlation analysis is performed between the screened feature monitoring dataset and the lithological distribution characteristics in prior geological knowledge to identify the spatial correlation between the monitoring features and specific lithologies. The screened feature monitoring dataset is then matched with historical deformation patterns in prior geological knowledge to determine whether the current monitoring features match known historical deformation patterns. The results of the combined spatial correlation analysis and spatiotemporal pattern matching are used to assess the potential intensity of geological activity at different spatial locations within the target area. Based on the spatial distribution of the potential intensity of geological activity, combined with the spatiotemporal evolution trends of feature values ​​in the screened feature monitoring dataset, a geological state coupling analysis result is generated, including the location, risk level, and activity pattern prediction of potential geological risk areas. In the geological state coupling analysis result, high-confidence analysis areas confirmed as accurate by subsequent field verification or independent monitoring methods are identified. The role of the original monitoring data provided by each data source in the coupling analysis process of the high-confidence analysis areas is reviewed. Based on the analytical contribution of each data source in the high-confidence analysis areas, their initial comprehensive data quality scores are revised; data sources with large contributions have their data quality scores increased, while data sources with small contributions or providing interfering information have their data quality scores decreased. Using the corrected data quality scores from each data source, the basic weight factors in the dynamic weight allocation model for data source contribution are recalculated. These recalculated basic weight factors are then substituted into the dynamic weight allocation model to generate an updated model, which is used for weighted fusion of subsequently acquired raw monitoring data.

[0095] In practical implementation, the filtered feature monitoring dataset is collaboratively analyzed with prior geological knowledge, and a closed-loop optimization mechanism is constructed. Taking a slope in a mountainous area in Southwest China as the target area as an example, the filtered feature monitoring dataset is spatially correlated with the lithological distribution characteristics in prior geological knowledge to identify the spatial correlation between monitoring features and specific lithologies. For example, the boundaries of significant deformation areas shown in the filtered feature monitoring dataset are superimposed and compared with the contact zones of strongly weathered sandstone and mudstone in the lithological distribution map to calculate the spatial consistency between deformation gradient changes and lithological boundaries. The filtered feature monitoring dataset is then spatiotemporally matched with historical deformation patterns in prior geological knowledge to determine whether the current monitoring features match known historical deformation patterns. For example, the direction and rate of the currently monitored surface displacement vector are compared with the "rear edge fracturing, front edge bulging" pattern described in historical records to analyze the similarity of their spatial patterns.

[0096] In some embodiments, the results of spatial correlation analysis and spatiotemporal pattern matching are used to assess the potential intensity of geological activity at different spatial locations within the target area. This is typically achieved through an evaluation function that integrates spatial correlation coefficients and spatiotemporal pattern matching scores. The spatial correlation coefficients reflect the degree of correlation between monitored anomalies and geological conditions, while the spatiotemporal pattern matching scores reflect the similarity between current dynamics and historical patterns. Based on the spatial distribution of potential intensity of geological activity, and combined with the spatiotemporal evolution trends of feature values ​​in the filtered feature monitoring dataset, a geological state coupling analysis result is generated, including the location of potential geological risk areas, risk levels, and activity pattern predictions. Risk levels can be classified according to potential intensity, and activity pattern predictions are inferred based on matching with historical patterns. See Table 1 for an example of the geological state coupling analysis results, which shows the analysis results for several typical locations within the target area.

[0097] Table 1: Results of Geological State Coupling Analysis

[0098] ;

[0099] Optionally, in the coupled geological state analysis results, high-confidence analysis areas that are confirmed accurate by subsequent field verification or independent monitoring methods are identified. For example, the location with coordinates (345670, 2876543) in Table 1 was confirmed to have significant deformation by subsequent UAV detailed inspection and on-site crack measurement. The role of the original monitoring data provided by each data source in the coupled analysis process of the high-confidence analysis area is reviewed. Specifically, this is done by analyzing the proportion of the weighted feature data of each data source and its matching contribution with prior knowledge in the final coupled analysis results for that area. Based on the analytical contribution of each data source in the high-confidence analysis area, its initial comprehensive data quality score is adjusted. Data sources with large contributions have their data quality scores increased, while data sources with small contributions or providing interfering information have their data quality scores decreased.

[0100] It is understandable that the evaluation of contribution can be based on a quantitative indicator, such as the proportion of key feature monitoring data provided by the data source in a high-confidence region and ultimately selected and retained, relative to all key feature monitoring data in that region. Using the corrected data quality scores of each data source, the basic weight factors in the dynamic weight allocation model for data source contribution are recalculated. In some embodiments, the correction of the data quality scores can be based on the following formula:

[0101] ;

[0102] in: Indicates the first The overall data quality score after correction for each data source. This indicates its initial overall data quality score. Indicates the first The average analytical contribution percentage of each data source within the confirmed set of high-confidence analysis regions. This represents the average contribution percentage of all data sources to the analysis. This is a preset correction strength coefficient. This formula means that if a data source's average contribution is over multiple high-confidence regions... Above average Its data quality score A positive correction will be made, while a negative correction will be made. The recalculated basic weight factors will be substituted into the dynamic weight allocation model of the data source contribution to generate an updated dynamic weight allocation model of the data source contribution. This updated model will be used to weight and fuse newly acquired raw monitoring data, thereby completing a closed-loop optimization for one analysis cycle.

[0103] See Figure 4 In the temporal deformation trend analysis of different risk zones, the dynamic evolution characteristics of surface displacement showed a significant positive correlation with the risk level. Specifically, the surface displacement value in the high-risk zone (solid dot line) increased continuously from approximately 2.1 mm in January 2025 to approximately 12.8 mm in January 2026, showing an overall near-linear accelerating growth trend, with the growth rate gradually increasing, reflecting the continuous intensification of geological activity and the sustained increase in instability risk in this area; the surface displacement value in the medium-risk zone (solid box line) increased slowly from approximately 0.8 mm to approximately 2.3 mm, with a gentle growth process and small fluctuations, indicating that this area is in a stable creep stage and the risk level is controllable; the surface displacement value in the low-risk zone (solid triangle line) remained at an extremely low level of 0~0.3 mm, with no obvious trend of change, indicating that the geological structure of this area is stable and there is basically no significant deformation activity. From the perspective of temporal evolution, the displacement curves of the three risk zones showed a clear divergence after July 2025: the displacement rate in the high-risk zone accelerated significantly, the rate of increase in the medium-risk zone slightly increased, while the low-risk zone remained stable. This phenomenon is consistent with the background of intensified geological activity in the target area during this period, verifying the rationality of the multi-source data coupling method for risk zone division. Meanwhile, the slope of the displacement curve in the high-risk zone continued to increase, reflecting the irreversible acceleration characteristic of its deformation process, while the medium- and low-risk zones exhibited stable or slow linear growth, consistent with the geological evolution patterns under different risk levels. At the parameter configuration level, this analysis adopted a monthly monitoring cycle and controlled the displacement accuracy within ±0.1mm to ensure the continuity and reliability of time series data. The risk level classification comprehensively considered the absolute amplitude of surface displacement, growth rate, and matching degree with historical deformation patterns. The threshold for high-risk areas was a displacement value exceeding 6mm and a monthly average growth rate exceeding 0.8mm; medium-risk areas were displacement values ​​between 1 and 6mm and a monthly average growth rate below 0.5mm; and low-risk areas were displacement values ​​below 1mm and without a significant growth trend.

[0104] In one embodiment of the present invention, for the raw monitoring data of each data source, based on its data acquisition principle, processing flow, and the data quality score obtained from the evaluation, its data uncertainty is quantified, and an uncertainty spatial distribution map of each data source is generated. When weighted fusion of the raw monitoring data from each data source, the uncertainty spatial distribution maps corresponding to each data source are simultaneously weighted and fused according to their respective real-time dynamic contribution weights. A fused uncertainty distribution map corresponding to the weighted fused monitoring dataset space is generated. When performing adaptive feature filtering processing, the fused uncertainty distribution map is used as an auxiliary criterion; a stricter feature filtering threshold is applied to monitoring features from high uncertainty regions, while a relatively lenient feature filtering threshold is applied to monitoring features from low uncertainty regions.

[0105] In practical implementation, taking a slope in a mountainous area in Southwest China as the target monitoring scenario, for the raw monitoring data of each data source, based on its data acquisition principle, processing flow, and data quality score obtained from evaluation, the data uncertainty is quantified, and a spatial distribution map of uncertainty for each data source is generated. For topographic change data, its uncertainty mainly stems from spatial interpolation errors and coordinate matching errors during the generation of multi-period digital elevation models. The uncertainty of the elevation difference of each grid cell can be calculated using the error propagation law, forming an uncertainty spatial distribution map consistent with the spatial resolution of the topographic change data. Each pixel value in the map represents the estimation error range of the topographic change value at that location. For geological deformation time series data, its uncertainty originates from the sensor's own measurement noise, multipath effects, and model residues in data processing. By analyzing the fluctuation characteristics of the time series and the nominal accuracy of the instrument, an uncertainty value that varies with time or is a comprehensive value can be assigned to the deformation time series data of each monitoring point. Spatial interpolation is then used to generate a spatial distribution map of uncertainty for geological deformation time series data covering the entire target area.

[0106] In some embodiments, the uncertainty of lithological distribution data mainly lies in the uncertainty of lithological boundaries and the credibility of lithological category interpretation. Based on the exploration point density and the confidence level of geological interpretation, a spatial distribution map of lithological distribution data uncertainty, expressing the probability or credibility of lithological category at each spatial location, can be generated through geostatistical simulation or fuzzy membership methods. For historical activity record data, the uncertainty is mainly related to the level of detail of the records themselves, memory bias due to their age, and contradictions between different historical materials. A credibility score can be assigned to each historical event record through expert scoring or a multi-criteria evaluation method based on information integrity and consistency, and this score can be converted into a spatial distribution map of historical activity record data uncertainty. When weighted fusion of the raw monitoring data from various data sources, the spatial distribution maps of uncertainty corresponding to each data source are simultaneously weighted and fused according to their respective real-time dynamic contribution weights. The transmission of uncertainty in weighted fusion follows the general law of error propagation.

[0107] Optionally, a post-fusion uncertainty distribution map corresponding to the weighted fusion monitoring dataset space is generated. This post-fusion uncertainty distribution map quantitatively characterizes the numerical reliability or possible error range of each data point in the final weighted fusion monitoring dataset after multi-source data coupling. During adaptive feature selection, the post-fusion uncertainty distribution map is used as an auxiliary criterion, applying a stricter feature selection threshold to monitoring features from high uncertainty regions and a relatively lenient feature selection threshold to monitoring features from low uncertainty regions. This step essentially integrates data reliability information into the feature selection process. Data from high uncertainty regions, even if numerically large, may be scrutinized more rigorously due to their low reliability, while small but reliable changes from low uncertainty regions are more likely to be retained. In some embodiments, adjusting the feature selection threshold based on the post-fusion uncertainty distribution map can be implemented through a function:

[0108] ;

[0109] in: Indicates spatial location At this point, the final feature selection threshold after uncertainty adjustment, This represents the threshold that has been dynamically adjusted based on prior geological knowledge in the embodiments. Indicates position The normalized uncertainty value (ranging from 0 to 1) on the uncertainty distribution plot after fusion. This is a preset adjustment parameter used to control the strength of the impact of uncertainty on threshold adjustment. When When the value is large (high uncertainty), The term is greater than 1, resulting in An increase in the threshold indicates the use of a stricter selection threshold, while a decrease indicates a more lenient threshold. Ultimately, the position... The monitoring characteristic values ​​need to be consistent with this The decision to retain or exclude is made by comparison.

[0110] See Figure 5 This image presents the spatial distribution characteristics of elevation differences in the target area. The data originates from the differential calculation of a multi-period digital elevation model (DEM) and is a typical visualization representation of a weighted fusion monitoring dataset in the topographic dimension. Spatially, the image uses cool colors (blue and green) to represent areas with negative elevation differences (i.e., decreasing surface elevation, corresponding to erosion or subsidence), and warm colors (yellow, brown, and gray) to represent areas with positive elevation differences (i.e., increasing surface elevation, corresponding to deposition or uplift). The numerical range covers -1.2m to 1.2m, perfectly corresponding to the color scale on the right. Specifically, significant negative elevation differences (-0.6m to -1.2m) were observed in the regions with X coordinates of 2–4km and Y coordinates of 6–10km, and in the regions with X coordinates of 7–10km and Y coordinates of 0–4km, indicating a significant decrease in surface elevation during Phase 2. Conversely, positive elevation differences (0.3m to 1.2m) were observed in the regions with X coordinates of 0–5km and Y coordinates of 0–4km, and in the regions with X coordinates of 6–10km and Y coordinates of 6–10km, reflecting an increase in surface elevation or material accumulation. The spatial heterogeneity of elevation differences in the figure not only reflects the differences in surface processes in the target area, but also provides key input for subsequent collaborative analysis with prior geological knowledge: by performing spatial correlation analysis between the figure and lithological distribution characteristics, the correlation between elevation changes and specific lithologies (such as weak rock layers) can be identified; by performing pattern matching with historical deformation patterns, it can be determined whether the current topographic changes conform to known historical activity patterns such as landslides and erosion, thereby generating geological state coupling analysis results that include the location and risk level of potential geological risk areas.

[0111] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A multi-source data coupling method, characterized in that, The method includes: Raw monitoring data of the target area is obtained from multiple data sources, a raw monitoring data quality assessment unit is constructed, the quality of raw monitoring data from each data source is assessed, and a data quality score for each data source is generated. Based on the data quality scores of each data source and combined with the geological activity stage parameters of the target area, a dynamic weight allocation model for the contribution of data sources is established to calculate the real-time dynamic contribution weight of each data source. Based on the real-time dynamic contribution weights, the original monitoring data from each data source are weighted and fused to generate a weighted fused monitoring dataset. Introducing prior geological knowledge of the target area, including lithological distribution characteristics and historical deformation patterns, adaptive feature filtering processing is performed on the weighted fusion monitoring dataset; Through adaptive feature filtering, key feature monitoring data are retained to form a filtered feature monitoring dataset; The filtered feature monitoring dataset is combined with the geological prior knowledge for collaborative analysis to generate coupled analysis results of the geological state of the target area. A closed-loop optimization mechanism is constructed, and the data quality scores of each data source are recalibrated using the geological state coupling analysis results, and the dynamic weight allocation model of the contribution of the data source is updated. The updated dynamic weighting model for data source contribution is used for weighted fusion of the original monitoring data in the next cycle, thereby achieving adaptive coupling optimization of multi-source monitoring data.

2. The multi-source data coupling method according to claim 1, characterized in that, The process of obtaining raw monitoring data for the target area from multiple data sources specifically involves: The original monitoring data includes topographic change data, geological deformation time series data, lithological distribution data, and historical activity record data; Topographic change data of the target area is obtained from the first type of data source, the topographic change data including digital elevation model difference data collected at different time points; Geological deformation time series data of the target area are obtained from the second type of data source, wherein the geological deformation time series data is a time series of surface displacement continuously collected through a sensor network; Lithological distribution data of the target area is obtained from a third type of data source, wherein the lithological distribution data is a spatial distribution map of rock types obtained through geological exploration; Historical activity records of the target area are obtained from the fourth type of data source. The historical activity records include spatiotemporal information of historical earthquake events, landslide events, and fault activity records. The topographic change data, geological deformation time series data, lithological distribution data, and historical activity record data are registered and aligned according to a unified spatiotemporal benchmark to form the initial raw monitoring dataset.

3. The multi-source data coupling method according to claim 2, characterized in that, The construction of the raw monitoring data quality assessment unit, which assesses the quality of raw monitoring data from each data source and generates a data quality score for each data source, specifically involves: For terrain change data, assess its spatial resolution, temporal coverage integrity, and elevation measurement accuracy to generate a terrain data quality sub-score; For geological deformation time series data, the continuity of the time series, the proportion of missing values, and the level of sensor acquisition noise are evaluated to generate a deformation data quality sub-score; For lithological distribution data, assess its exploration point density, lithological classification, and spatial interpolation uncertainty, and generate a lithological data quality sub-score. For historical activity record data, assess the completeness of recorded events, spatiotemporal location, and reliability of event intensity descriptions to generate historical data quality sub-scores; By combining the quality sub-scores corresponding to each data source, a comprehensive data quality score for each data source is calculated using a preset score aggregation algorithm.

4. The multi-source data coupling method according to claim 3, characterized in that, Based on the data quality scores of each data source and combined with the geological activity stage parameters of the target area, a dynamic weight allocation model for the contribution of data sources is established to calculate the real-time dynamic contribution weight of each data source, specifically: Obtain geological activity stage parameters that characterize the current intensity of geological activity in the target area; A basic weighting factor is set for the dynamic weighting model of data source contribution, and the basic weighting factor is positively correlated with the comprehensive data quality score of each data source; In the dynamic weight allocation model for the contribution of the data source, an adjustment coefficient for the geological activity stage parameter is introduced. When the geological activity stage parameter indicates an intensification of activity, the basic weight factor of the data source that is sensitive to time-series changes is increased, and vice versa. The comprehensive data quality score, basic weight factor, and adjustment coefficient of geological activity stage parameters are combined to calculate the real-time dynamic contribution weight of each data source in the coupled analysis process, and ensure that the sum of the real-time dynamic contribution weights of all data sources is one.

5. The multi-source data coupling method according to claim 4, characterized in that, Based on the aforementioned real-time dynamic contribution weights, the original monitoring data from each data source are weighted and fused to generate a weighted fused monitoring dataset, specifically as follows: Read the real-time dynamic contribution weights corresponding to topographic change data, geological deformation time series data, lithological distribution data, and historical activity record data; The value of each spatial raster cell in the terrain change data is multiplied by its corresponding real-time dynamic contribution weight to obtain the weighted terrain change data. Multiply the value of each time series data point in the geological deformation time series data by its corresponding real-time dynamic contribution weight to obtain the weighted geological deformation time series data. The distribution probability of each lithology type in the lithology distribution data is multiplied by its corresponding real-time dynamic contribution weight to obtain the weighted lithology distribution data; The intensity and impact factor of each historical event in the historical activity record data are multiplied by its corresponding real-time dynamic contribution weight to obtain the weighted historical activity record data. The weighted topographic change data, geological deformation time series data, lithological distribution data, and historical activity record data are overlaid and spatially fused to generate a unified weighted fusion monitoring dataset.

6. The multi-source data coupling method according to claim 5, characterized in that, The introduced geological prior knowledge of the target area, including lithological distribution characteristics and historical deformation patterns, is used to perform adaptive feature filtering processing on the weighted fused monitoring dataset, specifically as follows: The lithological distribution characteristics of the target area are retrieved from the geological prior knowledge base. These lithological distribution characteristics identify the spatial distribution of different rock types and their physical and mechanical properties. The historical deformation patterns of the target area are retrieved from the geological prior knowledge base. These historical deformation patterns describe the typical deformation patterns and spatial distribution characteristics of the target area in past geological activities. Set the initial global threshold for adaptive feature selection; Based on the lithological distribution characteristics, the initial global threshold is dynamically adjusted by weighting and fusing monitoring data from different lithological zones. A low feature screening threshold is used in lithologically unstable areas, while a high feature screening threshold is used in stable areas. Based on the aforementioned historical deformation patterns, in areas with historically active deformation, the feature selection threshold for monitoring features in the weighted fusion monitoring data that conform to the historical deformation patterns is further lowered and these features are prioritized for retention. Based on the dynamically adjusted feature screening threshold, each monitoring feature data in the weighted fusion monitoring dataset is screened. Data below the threshold is regarded as noise or minor features and filtered out, while data above the threshold is retained as key feature monitoring data, thus forming a screened feature monitoring dataset.

7. The multi-source data coupling method according to claim 6, characterized in that, The filtered feature monitoring dataset is then collaboratively analyzed with the geological prior knowledge to generate a coupled analysis result of the geological state of the target area. Specifically: Spatial correlation analysis was performed between the filtered feature monitoring dataset and the lithological distribution characteristics in geological prior knowledge to identify the spatial correlation between the monitoring features and specific lithologies. The filtered feature monitoring dataset is matched with historical deformation patterns in geological prior knowledge in a spatiotemporal pattern to determine whether the current monitoring features are consistent with known historical deformation patterns. Based on the results of integrated spatial correlation analysis and spatiotemporal pattern matching, the potential intensity of geological activity at different spatial locations within the target area is assessed. Based on the spatial distribution of the potential intensity of geological activity, and combined with the spatiotemporal evolution trend of the feature values ​​in the filtered feature monitoring data, a coupled analysis result of geological state is generated, which includes the location of potential geological risk areas, risk level, and activity pattern prediction.

8. The multi-source data coupling method according to claim 7, characterized in that, The aforementioned closed-loop optimization mechanism utilizes the geological state coupling analysis results to recalibrate the data quality scores of each data source and update the dynamic weight allocation model for the contribution of the data sources. Specifically: In the results of the geological state coupling analysis, high-confidence analysis areas were identified that were subsequently verified as accurate by on-site inspections or independent monitoring methods. The role of the raw monitoring data provided by each data source was traced back during the coupled analysis process in the high-confidence analysis area; Based on the analytical contribution of each data source in the high-confidence analysis region, the initial comprehensive data quality score is adjusted. Data sources with large contributions have their data quality scores increased, while data sources with small contributions or that provide interfering information have their data quality scores decreased. Using the corrected data quality scores of each data source, the basic weight factors in the dynamic weight allocation model of the data source contribution are recalculated; The recalculated basic weight factors are substituted into the dynamic weight allocation model of the data source contribution to generate an updated dynamic weight allocation model of the data source contribution, which is used to weight and fuse the newly acquired original monitoring data.

9. A multi-source data coupling method according to claim 8, characterized in that, It also includes the step of quantifying and transferring the uncertainty of the raw monitoring data before weighted fusion, specifically: For the raw monitoring data of each data source, based on its data acquisition principle, processing flow and the data quality score obtained from the evaluation, its data uncertainty is quantified, and an uncertainty spatial distribution map of each data source is generated. When weighting and fusing the raw monitoring data from each data source, the uncertainty spatial distribution maps corresponding to each data source are simultaneously weighted and fused according to their respective real-time dynamic contribution weights. Generate a fusion uncertainty distribution map corresponding to the space of the weighted fusion monitoring dataset; When performing adaptive feature filtering, the fused uncertainty distribution map is used as an auxiliary criterion. A stricter feature filtering threshold is applied to monitoring features from high uncertainty regions, while a relatively lenient feature filtering threshold is applied to monitoring features from low uncertainty regions.

10. A multi-source data coupling system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the multi-source data coupling method according to any one of claims 1 to 9.

Images