A regional geological disaster risk assessment method and system based on multi-source data fusion

Abstract

The application discloses a regional geological disaster risk assessment method and system based on multi-source data fusion, relates to the technical field of geological disaster risk assessment, and comprises the following steps: collecting original signals of remote sensing image data, surface deformation sequence data and multi-point sensing data in a large-scale region, embedding time synchronization identification in the collection process, and uniformly arranging the time sequence of various data; on the basis of the time-synchronized original signals, high-frequency weak signals are subjected to energy section extraction, and the detail signals of surface micro-sliding regions, crack edge regions and deformation mutation regions are subjected to local weighted identification. Through time synchronization and weighted fidelity processing, the application guarantees that multi-source monitoring data maintains consistent time sequence and complete key details before and after compression, and realizes the continuous presentation of surface deformation details by combining time continuous tracking and cross-source integration, so that the stability and foresight of regional geological disaster risk assessment are improved.

Description

Technical Field

[0001] This invention relates to the field of geological disaster risk assessment technology, specifically to a regional geological disaster risk assessment method and system based on multi-source data fusion. Background Technology

[0002] Regional geological hazard risk assessment based on multi-source data fusion refers to the process of comprehensively analyzing and quantitatively judging the hazard and potential losses of geological hazards at a regional scale by utilizing data resources from multiple sources and of different types. Its core lies in unifying the modeling of heterogeneous data from multiple sources, such as remote sensing imagery, topographic data, geological structure information, hydrological and meteorological data, earthquake monitoring records, engineering construction distribution, and socio-economic indicators, through spatial fusion, temporal synchronization, and feature correlation. Using methods such as machine learning, Bayesian inference, analytic hierarchy process (AHP), or graph neural networks, multidimensional correlation analysis is conducted on the inducing factors, disaster-prone environment, and characteristics of disaster-bearing bodies to identify disaster-prone areas, potential risk areas, and disaster propagation paths, thereby forming a regional-level geological hazard risk distribution map, providing a scientific basis for early warning decision-making, engineering site selection, and prevention and control planning.

[0003] The existing technology has the following shortcomings: In large-scale regional geological disaster monitoring, the amount of data is extremely large. To improve transmission and storage efficiency, lossy compression algorithms are often used to process remote sensing images, surface deformation sequences, and sensor data. However, this type of compression weakens subtle texture information without significantly affecting the overall image quality. It smooths or covers pixel gradients of small-scale deformation features such as landslide cracks and ground fissures, causing key early risk signals to be gradually lost during data transmission. Such feature loss is often difficult to recover through subsequent algorithms. The system may misjudge the surface state as stable during the automatic identification stage, missing weak displacement areas in the disaster-prone phase. This can easily lead to the failure of early disaster warnings, delaying prevention and control efforts, and even causing regional risk assessment biases during the disaster expansion stage, creating a potential disaster amplification effect.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a regional geological disaster risk assessment method and system based on multi-source data fusion to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a regional geological hazard risk assessment method based on multi-source data fusion, comprising the following steps: Step 1: Collect raw signals of remote sensing image data, surface deformation sequence data and multi-point sensor data over a large area. During the acquisition process, embed time synchronization identifiers and perform unified arrangement of the time series of various types of data to establish a time consistency comparison basis for the subsequent compression stage. Step 2: Based on the original signal after time synchronization, energy segmentation is performed on the high-frequency weak signal. The detailed signals of the surface micro-slip area, crack edge area and deformation abrupt area are locally weighted and labeled to provide data basis for key fidelity allocation in the subsequent compression stage. Step 3: Based on the weighted identifiers formed by energy segment extraction, perform partitioned coding control during the data compression stage, and implement differentiated adjustment of the compression ratio for key areas to maintain the continuous features of key textures and prevent the weakening of details during the data dimensionality reduction process. Step 4: Combining the compression results after partition coding adjustment, the coding sequence of key areas is continuously tracked in time, and the signal brightness of adjacent time periods is subjected to differential balancing processing in order to correct the small drift deviations in the compression process and maintain the temporal extensibility of deformation details. Step 5: Based on the difference balance results after continuous time tracking, perform cross-source data mapping and integration, dynamically match high-sensitivity detail areas in remote sensing image data, surface deformation sequence data and multi-point sensor data, and generate multi-source complementary detail restoration output to achieve continuous maintenance and dynamic extension of early risk signals, thereby completing the regional geological disaster risk assessment.

[0007] Preferably, the steps for collecting raw signals from remote sensing imagery, surface deformation sequence data, and multi-point sensor data over a large area include: For remote sensing image data, surface deformation sequence data and multi-point sensing data, the acquisition path, sampling frequency and time recording method are set respectively, and a unified time reference signal is introduced during the acquisition process to synchronize the time recording of each acquisition channel; After the data collection is completed, the data collection time of the remote sensing images, the observation time of the surface deformation sequence, and the sampling time of the multi-point sensing signals are converted into readable time stamp signals, with a unified time reference as the core, and then embedded into the corresponding original signal data. Using the embedded time stamp as an index, the raw signal data from different sources are rearranged in a unified time axis order, and the time interval is filled in by interpolation when the sampling frequency is inconsistent. In a uniformly compiled dataset, the sampling interval between adjacent time nodes is analyzed, a global time mapping table is constructed, and the time nodes of remote sensing images, surface deformation sequences, and multi-point sensing data are cross-compared and the node distribution is adjusted to form a basis for time consistency comparison.

[0008] Preferably, the steps of extracting energy segments and performing local weighted labeling of the high-frequency weak signal based on the original signal after time synchronization include: In the original signal data after time synchronization, the energy distribution of remote sensing image data, surface deformation sequence data and multi-point sensing data is divided, and data from different sources are sorted in layers at the same time node to determine the time segment and spatial range corresponding to high-frequency information. Based on the energy distribution division results, the time axis and spatial distribution are divided synchronously, and the high-frequency energy segments corresponding to each time node are continuously extracted and matched with spatial location to form a set of high-frequency energy segments; For high-frequency energy fragment sets, local weighted identification is performed on the detailed signals of surface micro-slip regions, crack edge regions, and deformation abrupt regions, and a weight mapping relationship is established in the time axis and spatial grid. The locally weighted identifier is fused with the original signal after time synchronization, and embedded into the corresponding time nodes in chronological order to form a comprehensive signal set containing the original signal value, energy distribution characteristics, and weighting factors, which is used for data allocation in the subsequent compression stage.

[0009] Preferably, the steps for performing partition coding control during the data compression stage based on the weighted identifier formed by energy segment extraction include: Based on the weighted identifiers formed by energy segment extraction, and combined with time synchronization identifiers and spatial coordinates, remote sensing image data, surface deformation sequence data and multi-point sensor data are divided into zones to form regional correspondences of key fidelity zones, general fidelity zones and low fidelity zones. Based on the partitioning results, a correspondence is established between the weighted identifier and each partition, and a corresponding compression ratio parameter is set for different partitions according to the weight distribution; Based on the compression ratio parameter and weighted identification relationship, the remote sensing image data, surface deformation sequence data and multi-point sensor data in each partition are subject to partition coding control, and the compression ratio of the same spatial location is kept consistent under the time node sorting. Based on the compression results generated by partition coding regulation, data from adjacent time nodes are integrated and processed in conjunction with time synchronization identifiers to maintain the continuous feature expression of key areas in the time series.

[0010] Preferably, during the partitioned encoding control process, the compression ratio parameter of the key fidelity area remains consistent at each time node, and the brightness change and signal amplitude are balanced according to the time synchronization mark in the data integration processing of adjacent time nodes, so as to maintain the continuous expression of key textures in the time series.

[0011] Preferably, the steps of performing temporal continuous tracking and difference balancing on the encoded sequence of key regions based on the compression results after partition coding adjustment include: In the compressed data formed by partition coding control, key areas are selected based on the weighting identifier and compression ratio setting results, and the coding sequences corresponding to the key areas are sorted by time according to the time synchronization identifier to form a set of continuously arranged coding sequences. Based on the set of encoded sequences, the encoded sequences of adjacent time nodes are continuously tracked along the time axis, and the encoded segments of the same geographical location are kept consistent in the spatial coordinate level, while the compression parameter information of each time node is preserved. Based on the results of continuous time tracking, the signal brightness difference balancing process is performed on the encoded sequences of adjacent time periods, and the brightness change amplitude is adjusted under the constraint of time mapping relationship to form a continuously changing time series; Based on the time series formed by difference balancing, the temporal extension of the encoded sequence in key areas is maintained, and the brightness and signal amplitude are kept smoothly connected in continuous time periods.

[0012] Preferably, in the step of performing time-continuous tracking and differential balancing of the coding sequence of key areas, the differential balancing is performed based on the brightness change relationship of the coding sequence corresponding to adjacent time nodes, and while maintaining the consistency of time synchronization markers, the brightness or amplitude of remote sensing image data, surface deformation sequence data and multi-point sensing data are adjusted respectively, so as to maintain the temporal consistency and continuous evolution characteristics of the coding sequence of key areas within a continuous time period.

[0013] Preferably, the steps for performing cross-source data mapping and integration based on the difference balancing results after continuous time tracking include: In the compressed data formed by continuous time tracking and difference balancing, high-sensitivity detail areas corresponding to remote sensing image data, surface deformation sequence data and multi-point sensing data are extracted, and a time consistency index relationship is established based on the time synchronization identifier. Based on the temporal consistency index relationship, spatial coordinate alignment and feature mapping are performed on high-sensitivity detail areas to uniformly map remote sensing image data, surface deformation sequence data and multi-point sensor data to the same spatial reference system. Based on the spatial coordinate alignment and feature mapping results, dynamic matching and integration are performed on high-sensitivity detail regions of various data sources in the time series to form a multi-source feature set corresponding to both time and space dimensions; Based on the multi-source feature set formed by dynamic matching and integration, a multi-source complementary detailed restoration output is generated, and a regional geological hazard risk assessment is performed based on the detailed restoration output.

[0014] Preferably, in the step of generating multi-source complementary detailed restoration output, the high-sensitivity detailed regions of remote sensing image data, surface deformation sequence data and multi-point sensor data are reconstructed in time sequence according to the time synchronization identifier, and the high-sensitivity detailed regions are superimposed and expressed under a unified spatial reference system to form a multi-dimensional restoration result covering a continuous time period, and the regional geological disaster risk distribution is determined based on the multi-dimensional restoration result.

[0015] A regional geological hazard risk assessment system based on multi-source data fusion includes a multi-source time-series acquisition module, a high-frequency energy identification module, a zonal compression and control module, a time-series difference balancing module, and a cross-source detail fusion module; The multi-source time-series acquisition module collects raw signals of remote sensing image data, surface deformation sequence data and multi-point sensor data over a large area. It embeds time synchronization identifiers during the acquisition process and performs unified arrangement of the time series of various types of data. The high-frequency energy identification module extracts energy segments from the weak high-frequency signals based on the original signals after time synchronization, and locally weights and identifies the detailed signals of the surface micro-slip area, crack edge area and deformation abrupt area. The partitioned compression and control module performs partitioned coding control during the data compression stage based on the weighted identifier formed by energy segment extraction, and implements differentiated adjustment of the compression ratio for key areas; The timing difference balancing module combines the compression results after partition coding control to continuously track the coding sequence of key areas in time and perform difference balancing processing on the signal brightness of adjacent time periods. The cross-source detail fusion module performs cross-source data mapping and integration based on the difference balance results after continuous time tracking. It dynamically matches high-sensitivity detail regions in remote sensing image data, surface deformation sequence data, and multi-point sensor data to generate multi-source complementary detail restoration output.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention establishes a stable and consistent temporal correlation between remote sensing image data, surface deformation sequence data, and multi-point sensor data by introducing time synchronization identifiers and uniformly arranging the time series of multi-source data during the data acquisition stage, before compression processing. Building upon this, it further combines energy segmentation extraction and local weighted identifiers of high-frequency weak signals, giving higher fidelity priority to areas of surface micro-slippage, crack edges, and abrupt deformation changes during data dimensionality reduction, effectively preventing key details from being weakened or obscured during transmission and storage. This processing method ensures that subtle changes during the geological hazard incubation stage are continuously preserved and stably transmitted, providing a more reliable data foundation for subsequent risk identification.

[0017] This invention introduces continuous time tracking and difference balancing processing on the basis of compressed results, and further performs cross-source data mapping and integration, enabling data from different sources to form dynamic correspondences in time and space. Through multi-source complementary detailed restoration output, the continuous evolution process of surface deformation is fully presented, and early risk signals maintain extensibility and coherence in the time series, thereby reducing risk misjudgments caused by data distortion. This scheme can improve the stability and continuity of regional geological hazard risk assessment results, providing more forward-looking support for prevention and control decisions. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of a regional geological hazard risk assessment method based on multi-source data fusion according to the present invention.

[0020] Figure 2 This is a schematic diagram of a regional geological disaster risk assessment system based on multi-source data fusion according to the present invention. Detailed Implementation

[0021] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0022] This invention provides, for example Figure 1 The method for regional geological hazard risk assessment based on multi-source data fusion, as shown, includes the following steps: Step 1: Collect raw signals of remote sensing image data, surface deformation sequence data and multi-point sensor data over a large area. During the acquisition process, embed time synchronization identifiers and perform unified arrangement of the time series of various types of data to establish a time consistency comparison basis for the subsequent compression stage. The specific implementation method for this step is as follows: When conducting large-scale regional data acquisition, acquisition paths, sampling frequencies, and time recording methods are set separately for remote sensing image data, surface deformation sequence data, and multi-point sensor data. Remote sensing image data acquisition is performed in a regional coverage manner, using satellite or airborne remote sensing platforms to acquire multispectral images at different time points, ensuring the spatial integrity of the images, and recording the acquisition time information at the instant each image is generated. Surface deformation sequence data acquisition is achieved by continuously observing changes in surface height or reflection signals, forming time-series information containing minute surface displacements. Multi-point sensor data comes from stress gauges, rain gauges, temperature sensors, groundwater pressure sensors, and seismic vibration recorders deployed within the monitoring area. These devices output monitoring data streams at fixed intervals. To maintain the consistency of the temporal characteristics of these three types of data, a unified time reference signal is introduced during the acquisition process, for example, using a standard time signal as a reference to synchronize the time records of all acquisition channels. The sampling period and time interval for each type of data are configured according to its physical characteristics. For example, remote sensing imagery is collected in hours or days, surface deformation in minutes, and multi-point sensor data in seconds. By configuring the sampling period and acquisition interval as integer multiples, the three types of data can form aligned sampling nodes on the time axis. After acquisition, each data record is accompanied by start and end time identifiers and acquisition number to ensure that all raw signals are traceable in the time dimension.

[0023] After the initial signal acquisition is completed, time synchronization identifiers are embedded into the three types of data. This process, centered on a unified time reference, converts the acquisition time of remote sensing images, the observation time of surface deformation sequences, and the sampling time of multi-point sensor signals into readable time identifier signals, which are then embedded into the corresponding original signal data. For remote sensing images, the time identifier embedding is based on the image generation time, adding time records accurate to the second to the image file's metadata. For surface deformation sequences, time identifier embedding is achieved by adding a time series index to each set of height change data, ensuring that every displacement record at the same surface location has a clear time correspondence. For multi-point sensor data, time identifier embedding adds a timestamp to each sensor output record, ensuring that the time synchronization deviation of all sensors is controlled within the millisecond range. After embedding, all data files have time identifier fields in a unified format. These fields are expressed using a unified time encoding method, enabling unified parsing and positioning of data from different sources on the timeline. In this way, remote sensing images, deformation sequences, and sensor data all possess accurately matchable time information before entering subsequent processing stages.

[0024] After the time stamp embedding is completed, the original signal data with embedded time information is uniformly arranged in time series. This arrangement uses the time stamp as the core index, rearranging data from different sources according to a unified time axis order. First, the remote sensing image data is arranged chronologically from early to late, and deformation records for corresponding time periods are inserted into the same time node based on the time index of the surface deformation sequence data. Subsequently, the multi-point sensor data is segmented and organized according to its sampling interval, and aligned on the time axis to the corresponding time points of the remote sensing images and surface deformation. In this process, for data with different sampling frequencies, the time interval is filled in by interpolation to ensure that different data types have corresponding data records at each unified time node. In the uniformly arranged dataset, each time node contains synchronous records of three types of data: remote sensing images corresponding to spatial images, surface deformation records corresponding to surface change curves, and multi-point sensor data corresponding to changes in environmental physical quantities. In this way, the entire time series forms a time-aligned structure for multi-source data, enabling various types of data to undergo difference analysis and joint judgment using the same time reference in subsequent processing.

[0025] After completing the unified time series arrangement, a time consistency reference is established across the entire timeline to provide a continuous time mapping relationship for subsequent data compression and fusion stages. Specifically, by analyzing the sampling intervals between adjacent time nodes from different data sources, a global time mapping table is constructed, cross-referencing the acquisition nodes of remote sensing images, the recording nodes of surface deformation, and the sampling nodes of multi-point sensor data. During this process, the average, maximum, and minimum offsets of the time intervals are calculated, and the node distribution of the time mapping table is adjusted according to the offsets to maintain a constant correspondence between time nodes. In the generated time consistency reference table, each row records a time node, including the index number of the three types of data at that time and the corresponding time identifier. Through this time reference table, during subsequent compression operations, the time consistency of different data sources before and after compression can be maintained based on a unified time node, thus preventing time series drift caused by compression. In the subsequent fusion and identification processes, the time positioning of all data is based on this reference table to achieve precise alignment on the timeline.

[0026] Through the above steps, the synchronized acquisition and temporal consistency construction of remote sensing image data, surface deformation sequence data, and multi-point sensor data were completed. The entire process achieved continuous operation from raw signal acquisition to time stamp embedding and unified time series arrangement, ensuring the consistency of multi-source data across the temporal dimension. The established temporal consistency reference not only ensures the temporal stability of various data types during subsequent compression stages but also provides accurate temporal references for energy extraction, weighted labeling, and cross-source mapping. This approach ensures that various monitoring data can work collaboratively within the same time frame during regional geological hazard risk assessment, providing complete data temporal support for the accurate identification of early geological hazard risk characteristics.

[0027] Step 2: Based on the original signal after time synchronization, energy segmentation is performed on the high-frequency weak signal. The detailed signals of the surface micro-slip area, crack edge area and deformation abrupt area are locally weighted and labeled to provide data basis for key fidelity allocation in the subsequent compression stage. The specific implementation method for this step is as follows: After time synchronization, the energy distribution of remote sensing image data, surface deformation sequence data, and multi-point sensor data is divided to determine the time period and spatial range of high-frequency information. This process is based on time synchronization identifiers, stratifying data from different sources at the same time node. For remote sensing image data, areas of subtle changes in surface morphology are identified by analyzing the grayscale variation amplitude and surface reflectance differences between adjacent image frames; these areas typically reflect potential surface displacement or structural changes. For surface deformation sequence data, energy concentration zones of minor surface deformation are identified by calculating the degree of difference between adjacent observations in the time series. For multi-point sensor data, instantaneous fluctuations corresponding to surface disturbances or underground stress releases are identified by analyzing the amplitude variations and response delays of the sensor signals. After this stage of processing, high-frequency information in the raw signal is clearly located, forming a high-frequency information index containing time location, spatial coordinates, and energy level, laying the foundation for the next step of energy extraction.

[0028] After obtaining the distribution characteristics of high-frequency information, the energy segmentation extraction is refined by synchronously dividing the temporal axis and spatial distribution correspondence. Using the time-synchronized original signal as a benchmark, high-frequency energy segments corresponding to each time node are continuously extracted and mapped one-to-one with their spatial locations. For remote sensing image data, pixel groups with large brightness gradient differences are extracted from areas with dense surface texture changes, forming high-energy segments containing surface crack edges and subtle sliding trajectories. For surface deformation sequence data, time periods of sudden increases in surface displacement are extracted, and their corresponding coordinates in the spatial grid are recorded, forming a set of local signals reflecting abrupt surface changes. For multi-point sensing data, high-frequency fluctuation segments are extracted from points with frequent signal changes within a short period and mapped to geographic coordinates. Through this method, the energy segmentation extraction process not only achieves independent capture of weak high-frequency signals from various data sources but also establishes a unified correspondence at the temporal and spatial levels, thereby enabling the concentrated expression of high-frequency details in surface micro-slippage, crack edges, and abrupt deformation areas.

[0029] It should be noted that: High-frequency weak signals refer to subtle changes in time or space, characterized by rapid changes but relatively small amplitudes, often submerged in background noise or overall trends. In this scheme, the specific definition can be limited by two dimensions: "change frequency" and "signal amplitude." On one hand, "high frequency" means the signal has a high number of changes or a large gradient change per unit time, such as the first-order difference or gradient value between adjacent sampling points exceeding a set frequency threshold (e.g., being in the high-frequency range of the overall spectrum, or having a change rate 1.5 to 3 times higher than the average change rate). On the other hand, "weak" means the amplitude of this type of signal is significantly smaller than the overall signal's main change amplitude, for example, its amplitude is in the low percentile range of the global amplitude distribution (e.g., within the 10% to 30% range), but it is still identifiable relative to the local noise level.

[0030] In summary, when a signal simultaneously meets the conditions of "the rate of change is higher than the set frequency threshold" and "the amplitude is low but higher than the noise baseline", it can be identified as a high-frequency weak signal. Such signals usually correspond to the early response characteristics of surface microslippage, crack initiation, or stress abrupt change.

[0031] After high-frequency energy segmentation extraction, the extracted detail signals are locally weighted to highlight the fidelity priority of key detail areas in subsequent data compression. This step is based on the high-energy segments obtained in the previous stage, assigning different weight values ​​to the corresponding signal segments in the time axis and spatial grid according to their importance. For remote sensing image data, higher weights are given to pixel blocks at the edges of surface cracks, the leading edge of landslides, and areas of concentrated surface deformation to maintain higher fidelity in subsequent data compression. For surface deformation sequence data, higher time weights are given to periods of abrupt deformation changes or continuously accelerating trends to ensure that the surface deformation process at these critical moments is fully preserved. For multi-point sensing data, higher weights are assigned to signals from monitoring points with high stress change rates, high vibration frequencies, or strong humidity and temperature fluctuations to prevent the weakening of the coupling information between environmental changes and surface deformation. The generation of local weighted labels not only enhances the signal representation capability of key areas but also provides a direct basis for differentiated coding in the data compression stage. After this processing, a hierarchical weight mapping relationship is formed in the entire dataset, enabling the high-risk region to be distinguished from the stable region during signal processing.

[0032] After completing the local weighting and labeling, the weighted results are fused with the time-synchronized original signal to generate a comprehensive signal set with fidelity priority information, providing data support for the subsequent compression stage. The fusion process uses time synchronization labeling as the main thread, embedding the results of energy segment extraction and weighting labeling into the corresponding time nodes of the original signal in chronological order while maintaining spatial coordinate consistency. Each time node contains three types of information: the original signal value, energy distribution characteristics, and weighting factors. This allows the compression stage to allocate different compression ratios to data from different regions and time periods based on this information. In this way, during subsequent compression operations, compression parameters can be controlled according to the priority in the weighting labeling, ensuring the integrity and continuity of high-frequency detail signals in areas of surface micro-slippage, crack edges, and abrupt deformation, thus preventing the weakening of key texture features during dimensionality reduction. This fusion process not only ensures the accurate correspondence of energy segment extraction results in both time and space but also achieves dynamic labeling of key fidelity areas, laying the foundation for efficient fusion of multi-source data and subsequent detail preservation.

[0033] Through the above steps, high-frequency weak signals can be completely extracted and systematically identified based on the original time-synchronized signal. The entire process prioritizes time consistency, determining the distribution of high-frequency signals through energy segmentation extraction, and then strengthening the expression of key detail regions through local weighted identification, ultimately generating a comprehensive dataset with temporal, spatial, and fidelity weighting information. This implementation method prioritizes the preservation of detailed signals from surface micro-slippage, crack edges, and abrupt deformation areas during subsequent compression stages, fundamentally solving the problem of weakened early-stage geological hazard signals during traditional compression processes, and providing a high-precision data foundation for regional geological hazard risk assessment.

[0034] Step 3: Based on the weighted identifiers formed by energy segment extraction, perform partitioned coding control during the data compression stage, and implement differentiated adjustment of the compression ratio for key areas to maintain the continuous features of key textures and prevent the weakening of details during the data dimensionality reduction process. The specific implementation method for this step is as follows: Before the data compression phase begins, the remote sensing image data, surface deformation sequence data, and multi-point sensor data to be compressed are partitioned based on the weighted identifier results generated by energy segmentation extraction. The partitioning is based on both temporal synchronization identifiers and spatial coordinates, dividing the entire monitoring area into three categories: high-fidelity zones, general-fidelity zones, and low-fidelity zones. High-fidelity zones correspond to areas with high weighted identifier values, typically including areas with dense surface cracks, frequent micro-slip activity, and areas experiencing abrupt deformation. General-fidelity zones correspond to areas with medium weighted values, mostly distributed on flat terrain or slowly deforming surfaces. Low-fidelity zones correspond to areas with low weighted values, mostly long-term stable surfaces or monitoring boundary areas. During the partitioning process, the spatial locations of various data sources are mapped one-to-one with time nodes based on the weighted identifier map generated in the previous stage, ensuring consistency in the partitioning results of different types of data at the same time. This partitioning operation provides clear spatial and temporal boundaries for subsequent partitioning, coding, and control, enabling the compression process to adopt different data processing strategies for different areas.

[0035] After partitioning, corresponding compression ratio parameters are set for the data characteristics of each partition. This step establishes a correspondence between weight values ​​and compression ratios based on the weighted identifiers from the previous stage, allowing high-weight areas to receive higher-fidelity encoding allocations. Specifically, for key fidelity areas, a low compression ratio strategy is adopted to preserve the grayscale levels, texture continuity, and temporal variation details of deformation signals in the image data as completely as possible; for general fidelity areas, a medium compression ratio is used to effectively reduce the dimensionality of the data while maintaining necessary spatial and temporal resolution; for low-fidelity areas, a high compression ratio is used to effectively control the overall data volume. During this process, the compression ratio for each partition is not a fixed value but varies continuously within a certain range according to the distribution of weighted identifiers, creating a smooth spatial transition and avoiding visual or numerical discontinuities caused by abrupt changes between regions. Through this differentiated compression ratio setting, key areas maintain high texture continuity and deformation detail recognition after compression, while non-key areas achieve dimensionality reduction without affecting the overall analysis accuracy, thus balancing compression efficiency and information fidelity.

[0036] After setting the compression ratio, the next stage is zonal coding and adjustment. This stage uses the time-synchronized original signal as a foundation, embedding the correspondence between the compression ratio and weighting identifiers into the coding process. For each zonal region, before compression, the corresponding parts of the remote sensing imagery, surface deformation sequences, and multi-point sensor data are synchronously loaded according to time nodes to ensure temporal continuity of the coding. Based on this, according to the compression ratio parameters set in the previous step, low-ratio compression is applied to the image signals and numerical sequences of key fidelity areas, while medium-to-high ratio compression is applied to general fidelity areas and low-fidelity areas, respectively. During the coding process, the data compression of each zonal region maintains temporal consistency, ensuring a constant compression ratio for the same geographical location at different time nodes, so that the surface deformation process can still be fully expressed after compression. Simultaneously, for high-frequency information within key areas, local weight adjustments are made to improve the coding accuracy of edge transition areas, thereby maintaining the continuous features of key textures. Through this zonal coding and adjustment operation, the entire dataset achieves priority preservation of local details and balanced distribution of global information while being compressed, enabling the data to still accurately reflect surface changes after dimensionality reduction.

[0037] After the partitioned coding and adjustment are completed, the compression results are integrated for temporal consistency to maintain the continuity of detailed features in key areas over continuous time periods. This process uses time synchronization markers as the main thread, comparing the compression results of the same geographical location at adjacent time nodes, and balancing the brightness changes and signal amplitude of each node according to time differences to avoid discontinuities in the time series caused by differences in compression ratios. For key high-fidelity areas, the compression ratio is kept stable throughout the entire time series to prevent breaks in high-frequency details in the time dimension. For general high-fidelity areas and low-fidelity areas, the signal changes on the time axis are kept smooth and extended by adjusting the compression parameters between continuous time segments. After integration, the entire dataset maintains continuous features in both spatial and temporal dimensions. In the spatial dimension, the key textures of surface micro-slip areas, crack edge areas, and abrupt deformation areas are completely continued; in the temporal dimension, the gradual process of surface deformation is coherently expressed. The compressed data after this integration process not only effectively controls its volume but also accurately recovers the detailed features and change trends of key areas during subsequent decompression and fusion processes.

[0038] Through the above steps, regional coding control can be achieved based on the weighted identifiers extracted from energy segments during the data compression stage, allowing for differentiated adjustment of compression ratios for different regions. The entire process, based on weighted identifiers, forms a closed compression management chain through four stages: spatial division, ratio setting, coding control, and temporal integration. This ensures that key textures in critical areas remain continuous during dimensionality reduction, preventing the weakening of detailed content. This implementation not only ensures the long-term preservation of high-frequency information from surface micro-slippage, crack edges, and abrupt deformation areas, but also provides a complete and reliable data foundation for subsequent multi-source data fusion, feature extension, and risk assessment.

[0039] Step 4: Combining the compression results after partition coding adjustment, the coding sequence of key areas is continuously tracked in time, and the signal brightness of adjacent time periods is subjected to differential balancing processing in order to correct the small drift deviations in the compression process and maintain the temporal extensibility of deformation details. The specific implementation method for this step is as follows: In the compressed data after partitioned coding and adjustment, the coding sequences corresponding to key fidelity areas are selected as the target objects for continuous time tracking. The key areas are determined based on the weighting identifier and compression ratio settings of the previous stage; these areas include micro-slip zones, crack edge zones, and abrupt deformation zones. For each type of key area, the compressed data is sorted according to the time synchronization identifier, forming a set of coded sequences arranged chronologically. Each coded sequence corresponds to a specific time node, containing the compression result after partitioned coding and adjustment at that node. Through this sorting operation, a continuous data chain of key areas in the time dimension is established, enabling logical connections between compressed outputs from different time periods. Based on this, a preliminary comparison of coding parameters, data distribution, and brightness characteristics between time nodes is performed to identify possible minor offsets or brightness differences between adjacent time nodes. These deviations often originate from the accumulation of subtle quantization errors caused by the differentiated compression ratios of the previous stage. This step lays the data organization foundation for subsequent continuous time tracking and difference balancing.

[0040] After time series sorting is completed, a continuous time tracking operation is performed on the encoded sequences of key areas. This continuous time tracking uses time synchronization markers as a reference, progressively connecting encoded sequences between adjacent time nodes to establish a continuous time mapping relationship. Specifically, for each key area, multiple consecutive time nodes are selected on the time axis, and the corresponding compression results are compared in spatial coordinates to ensure that encoded segments at the same geographical location can be tracked sequentially along the time axis. To ensure the integrity of the tracking process, compression parameter information for each time node is retained during processing, including encoding ratio, compression level, and brightness mapping relationship, thereby ensuring consistent parameter connection between each node in the tracking chain. Through continuous tracking, a sequence of compressed results for key areas can be formed on the time axis, giving the compressed data a continuous and traceable logical chain in the time dimension. Thus, when entering the difference balancing stage, the system can clearly define the correspondence between adjacent time periods, ensuring that adjustments to time differences have a precise reference.

[0041] Based on continuous time tracking, differential balancing is performed on the signal brightness of adjacent time periods. Differential balancing uses the tracked time series as input, adjusting and smoothing brightness differences by calculating the amplitude of brightness changes between adjacent time nodes, ensuring the compressed data maintains a continuous trend over time. For remote sensing image data of key areas, differential balancing focuses on critical components such as surface texture, shadows, and terrain brightness. By adjusting brightness distribution, it ensures that brightness changes at the same surface location in adjacent image frames conform to temporal evolution patterns. For surface deformation sequence data, differential balancing corrects the brightness amplitude difference of displacement signals between adjacent time periods, maintaining the continuity of deformation change curves. For multi-point sensing data, differential balancing maintains the stability of the time series by correcting signal amplitude differences within time intervals. In this process, differential balancing is not performed independently but corresponds one-to-one with the time mapping relationship obtained in the previous stage. Each brightness adjustment is performed based on the time chain to ensure that the direction and trend of brightness changes at each time node remain consistent. This processing effectively corrects time drift caused by differences in compression ratios, maintaining stable and continuous changes in key areas over time.

[0042] After completing the difference balancing process, the corrected coding sequence undergoes temporal extension maintenance to ensure the complete transmission of surface deformation details in key areas over long time series. This step, based on the balanced continuous time series, assesses the temporal extension of the coding results for each key area, maintaining smooth transitions in brightness and signal amplitude across continuous time periods. During temporal extension maintenance, the corrected coding sequence is first recombined according to time nodes to form a continuous sequence covering the entire monitoring period. Subsequently, the coding results at time period boundaries undergo smoothing processing to ensure no abrupt changes in the time series at node transitions. For areas with micro-slippage, extension maintenance ensures the continuous expression of displacement trends on the time axis; for crack edge areas, continuous time tracking and brightness balancing maintain the continuous visibility of crack propagation processes; for areas with abrupt deformation changes, temporal extension processing ensures a natural transition in brightness changes before and after the abrupt change, enabling the deformation process to possess complete and continuous evolutionary characteristics on the time axis. Through this process, the coding sequence of the entire key area forms a smooth and dynamic change process in the time dimension, so as to accurately reflect the temporal evolution characteristics of geological disasters in the subsequent data fusion and risk assessment stages.

[0043] Through the above steps, based on the compressed results after partitioned coding and adjustment, it is possible to achieve continuous temporal tracking and difference balancing of the coded sequences in key areas. The entire process, with time synchronization identification as the main thread and compression results as the foundation, establishes a complete temporal dimension continuous control mechanism through four stages: time sorting, continuous tracking, difference balancing, and temporal extension. This ensures that key areas maintain temporal consistency and detail extensibility in the compressed data. This implementation method effectively corrects minor drift deviations generated during compression, ensuring that detailed information in areas of surface micro-slippage, crack edges, and abrupt deformation is completely preserved in the temporal dimension, providing a high-fidelity, temporally continuous data foundation for subsequent cross-source data integration and dynamic risk assessment.

[0044] Step 5: Based on the difference balance results after continuous time tracking, perform cross-source data mapping and integration, dynamically match the high-sensitivity detail areas in remote sensing image data, surface deformation sequence data and multi-point sensor data, and generate multi-source complementary detail restoration output to achieve continuous maintenance and dynamic extension of early risk signals, thereby completing the regional geological disaster risk assessment. The specific implementation method for this step is as follows: In the compressed data after continuous time tracking and difference balancing, high-sensitivity detail regions corresponding to various data sources are extracted, and a temporal and spatial consistency index relationship is established. This process, based on time synchronization identifiers, aligns balanced remote sensing image data, surface deformation sequence data, and multi-point sensor data along the temporal dimension, ensuring a one-to-one correspondence between data at the same time point. For remote sensing image data, pixel groups with high brightness variation rates and continuous texture structures are extracted from high-sensitivity feature areas such as surface cracks, landslide fronts, and collapse boundaries. For surface deformation sequence data, spatial units with abrupt or continuously accelerating displacement rates are extracted. For multi-point sensor data, monitoring point signals with frequent stress changes, drastic groundwater level fluctuations, or concentrated seismic vibration energy are extracted. After extraction, the time axes of the three data sources are uniformly calibrated according to the time synchronization identifiers, ensuring strict consistency of various high-sensitivity detail regions at the same time point along the temporal dimension. Through this process, a multi-source high-sensitivity region index system with time as the main axis is formed, providing a basic reference for cross-source mapping and integration.

[0045] It should be noted that: The determination of high-sensitivity detail regions is quantitatively defined by preset change thresholds: For remote sensing image data, when the grayscale change rate of the same pixel location in adjacent time frames is greater than 1.5 to 3 times the overall average grayscale change rate, and the change continuously covers no less than a preset number of pixels in the spatial neighborhood (such as a continuous 5×5 pixel area), it is determined to be a region with a high brightness change rate and continuous texture structure; For surface deformation sequence data, when the displacement change rate per unit time exceeds more than twice the historical average rate of the monitoring point, or shows a monotonically increasing trend in multiple consecutive time nodes and the cumulative increase exceeds a set threshold (such as exceeding twice the historical standard deviation), it is determined to be a region of sudden displacement rate change or continuous acceleration; For multi-point sensing data, when the change amplitude of the monitoring signal within a unit time window exceeds 1.5 to 2 times its long-term average, or when there are multiple consecutive (such as more than 3) fluctuation peaks exceeding the threshold within a sliding time window, it is determined to be a monitoring point signal with frequent stress changes or concentrated energy.

[0046] By using the aforementioned quantitative standards based on relative thresholds and statistical characteristics, a unified determination of highly sensitive detail regions can be achieved, ensuring consistency and comparability of different data sources in terms of recognition scale.

[0047] After obtaining the temporally consistent index, spatial coordinate alignment and feature mapping are performed on high-sensitivity detail areas. This step, based on the indexing system established in the previous stage, uniformly maps remote sensing image data, surface deformation sequence data, and multi-point sensor data in the spatial dimension. For remote sensing image data, its position in the regional grid is determined through geographic coordinate association; for surface deformation sequence data, the spatial position of each deformation unit is mapped to the coordinate point corresponding to the image pixel; for multi-point sensor data, its sampling position is mapped to the same spatial reference system according to the geographical distribution of monitoring points. After coordinate alignment, high-sensitivity information from various data sources at the same spatial location is matched accordingly, enabling complementary association of high-sensitivity details from different data sources within the same area. For example, when remote sensing images show changes in surface texture but the deformation sequence does not show corresponding fluctuations, this can be supplemented by stress change characteristics in the sensor data; when the deformation sequence experiences a sudden change at a certain moment but the image data does not show a significant change, this can be confirmed by mapping subtle changes in image brightness with sensor signal synchronization information. This spatial mapping process enables the matching of multi-source data within the same spatial framework, providing a spatial foundation for dynamic integration.

[0048] After spatial mapping, dynamic matching and integration are performed on highly sensitive detail areas in both temporal and spatial dimensions to achieve complementary fusion of multi-source data. This step, based on the spatial alignment results of the previous stage, dynamically corresponds and overlays the highly sensitive features of various data types on a temporal scale. For the same time point, pixel areas with high brightness change rates in remote sensing images are matched with spatial units showing prominent displacement changes in the surface deformation sequence, and correlation enhancement is performed based on energy fluctuation information in multi-point sensing data. For continuous time periods, temporal extension relationships are established by tracking the change trajectories of highly sensitive features in various data types between consecutive nodes, forming a continuous chain of features from adjacent time periods on the time axis. Through dynamic matching and integration, multi-source data achieve a consistent trend expression in the temporal dimension and complementary feature coverage in the spatial dimension. During the integration process, highly sensitive areas that repeatedly appear in different data sources are maintained in the fusion result through dynamic weight adjustment; highly sensitive features that are prominent only in a single data source are supplemented and balanced through temporal correlation and spatial proximity relationships, thereby ensuring the integrity and continuity of the fusion output. This process enables spatiotemporal complementarity and feature enhancement among multi-source data, allowing detailed features of surface deformation, crack propagation, and stress evolution to be uniformly presented in the fusion results.

[0049] After dynamic matching and integration, a multi-source complementary detailed restoration output is generated and used for regional geological hazard risk assessment. This process, based on the dynamic matching results of the previous stage, reconstructs the fused high-sensitivity detailed areas into a continuous dynamic sequence in chronological order. For the remote sensing imagery, a continuous image sequence of surface changes is generated, which can intuitively reflect the temporal evolution of surface morphology; for the surface deformation sequence, a displacement trend curve changing over time is formed to identify areas of accelerated deformation; for the multi-point sensor data, a response curve of monitoring point parameters changing over time is generated to reveal potential underground disturbance patterns. By fusing and overlaying these three types of outputs, a multi-dimensional restoration result containing temporal, spatial, and physical response information is formed. This result not only maintains the detailed features of key areas after compression and balancing processing but also enhances the ability to identify weak risk signals through a multi-source complementary mechanism. Finally, a regional geological hazard risk distribution map is generated based on the fusion results, dividing high-risk areas, potential-risk areas, and stable areas, realizing a complete risk assessment process from data acquisition, compression and control, time tracking to cross-source fusion.

[0050] Through the above steps, based on the difference balancing results after continuous temporal tracking, cross-source mapping and dynamic integration of multi-source data can be completed, generating multi-source complementary detailed restoration output. The entire process, with temporal continuity as the main thread and spatial mapping as the foundation, constructs a multi-source data collaborative fusion mechanism through four stages: temporal alignment, spatial correspondence, dynamic matching, and restoration output. This enables remote sensing image data, surface deformation sequence data, and multi-point sensor data to achieve unified expression in both temporal and spatial dimensions. This implementation method not only achieves the continuous maintenance and dynamic extension of early risk signals but also provides a complete, continuous, and high-fidelity multi-source fusion data foundation for regional geological disaster risk assessment, thereby improving the timeliness and accuracy of disaster identification and prediction.

[0051] Beneficial effect 1: This invention establishes a stable and consistent temporal correlation between remote sensing image data, surface deformation sequence data, and multi-point sensor data by introducing time synchronization identifiers and uniformly arranging the time series of multi-source data during the data acquisition stage, before compression processing. Building upon this, it further combines energy segmentation extraction and local weighted identifiers of high-frequency weak signals, giving higher fidelity priority to areas of surface micro-slippage, crack edges, and abrupt deformation changes during data dimensionality reduction, effectively preventing key details from being weakened or obscured during transmission and storage. This processing method ensures that subtle changes during the geological hazard incubation stage are continuously preserved and stably transmitted, providing a more reliable data foundation for subsequent risk identification.

[0052] Benefit 2: This invention introduces continuous time tracking and difference balancing processing on the basis of compressed results, and further performs cross-source data mapping and integration, enabling data from different sources to form dynamic correspondences in time and space. Through multi-source complementary detailed restoration output, the continuous evolution process of surface deformation is fully presented, and early risk signals maintain extensibility and coherence in the time series, thereby reducing risk misjudgments caused by data distortion. This scheme can improve the stability and continuity of regional geological hazard risk assessment results, providing more forward-looking support for prevention and control decisions.

[0053] This invention provides, for example Figure 2 The regional geological hazard risk assessment system based on multi-source data fusion is shown, including a multi-source time-series acquisition module, a high-frequency energy identification module, a zonal compression and control module, a time-series difference balancing module, and a cross-source detail fusion module; The multi-source time-series acquisition module collects raw signals of remote sensing image data, surface deformation sequence data and multi-point sensor data over a large area. It embeds time synchronization identifiers during the acquisition process and performs unified arrangement of the time series of various types of data. The high-frequency energy identification module extracts energy segments from the weak high-frequency signals based on the original signals after time synchronization, and locally weights and identifies the detailed signals of the surface micro-slip area, crack edge area and deformation abrupt area. The partitioned compression and control module performs partitioned coding control during the data compression stage based on the weighted identifier formed by energy segment extraction, and implements differentiated adjustment of the compression ratio for key areas; The timing difference balancing module combines the compression results after partition coding control to continuously track the coding sequence of key areas in time and perform difference balancing processing on the signal brightness of adjacent time periods. The cross-source detail fusion module performs cross-source data mapping and integration based on the difference balance results after continuous time tracking. It dynamically matches high-sensitivity detail regions in remote sensing image data, surface deformation sequence data, and multi-point sensor data to generate multi-source complementary detail restoration output.

[0054] The present invention provides a regional geological disaster risk assessment method based on multi-source data fusion, which is implemented through the aforementioned regional geological disaster risk assessment system based on multi-source data fusion. For details of the specific method and process of the regional geological disaster risk assessment system based on multi-source data fusion, please refer to the aforementioned embodiment of the regional geological disaster risk assessment method based on multi-source data fusion, which will not be repeated here.

[0055] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A regional geological hazard risk assessment method based on multi-source data fusion, characterized in that, Includes the following steps: Step 1: Collect raw signals of remote sensing image data, surface deformation sequence data and multi-point sensor data over a large area. During the acquisition process, embed time synchronization identifiers and perform unified arrangement of the time series of various types of data. Step 2: Based on the original signal after time synchronization, energy segmentation is performed on the high-frequency weak signal, and the detailed signals of the surface micro-slip area, crack edge area and deformation abrupt area are locally weighted and labeled. Step 3: Based on the weighted identifiers formed by energy segment extraction, perform partition coding control during the data compression stage, and implement differentiated adjustment of compression ratios for key areas; Step 4: Combining the compression results after partition coding adjustment, perform time-continuous tracking of the coding sequence of key areas, and perform differential balancing processing on the signal brightness of adjacent time periods. Step 5: Based on the difference balance results after continuous time tracking, perform cross-source data mapping and integration, dynamically match high-sensitivity detail areas in remote sensing image data, surface deformation sequence data and multi-point sensing data, and generate multi-source complementary detail restoration output.

2. The regional geological hazard risk assessment method based on multi-source data fusion according to claim 1, characterized in that, The steps for acquiring raw signals from remote sensing imagery, surface deformation sequence data, and multi-point sensor data over a large area include: For remote sensing image data, surface deformation sequence data and multi-point sensing data, the acquisition path, sampling frequency and time recording method are set respectively, and a unified time reference signal is introduced during the acquisition process to synchronize the time recording of each acquisition channel; After the data collection is completed, the data collection time of the remote sensing images, the observation time of the surface deformation sequence, and the sampling time of the multi-point sensing signals are converted into readable time stamp signals, with a unified time reference as the core, and then embedded into the corresponding original signal data. Using the embedded time stamp as an index, the raw signal data from different sources are rearranged in a unified time axis order, and the time interval is filled in by interpolation when the sampling frequency is inconsistent. In a uniformly compiled dataset, the sampling interval between adjacent time nodes is analyzed, a global time mapping table is constructed, and the time nodes of remote sensing images, surface deformation sequences, and multi-point sensing data are cross-compared and the node distribution is adjusted to form a basis for time consistency comparison.

3. The regional geological hazard risk assessment method based on multi-source data fusion according to claim 2, characterized in that, The steps for extracting energy segments and performing local weighted labeling of weak high-frequency signals based on the original time-synchronized signal include: In the original signal data after time synchronization, the energy distribution of remote sensing image data, surface deformation sequence data and multi-point sensing data is divided, and data from different sources are sorted in layers at the same time node to determine the time segment and spatial range corresponding to high-frequency information. Based on the energy distribution division results, the time axis and spatial distribution are divided synchronously, and the high-frequency energy segments corresponding to each time node are continuously extracted and matched with spatial location to form a set of high-frequency energy segments; For high-frequency energy fragment sets, local weighted identification is performed on the detailed signals of surface micro-slip regions, crack edge regions, and deformation abrupt regions, and a weight mapping relationship is established in the time axis and spatial grid. The locally weighted identifier is fused with the original signal after time synchronization, and the corresponding time nodes are embedded in chronological order to form a comprehensive signal set containing the original signal value, energy distribution characteristics, and weighting factors.

4. The regional geological hazard risk assessment method based on multi-source data fusion according to claim 3, characterized in that, The steps for performing partition coding control during the data compression stage based on the weighted identifier formed by energy segment extraction include: Based on the weighted identifiers formed by energy segment extraction, and combined with time synchronization identifiers and spatial coordinates, remote sensing image data, surface deformation sequence data and multi-point sensor data are divided into zones to form regional correspondences of key fidelity zones, general fidelity zones and low fidelity zones. Based on the partitioning results, a correspondence is established between the weighted identifier and each partition, and a corresponding compression ratio parameter is set for different partitions according to the weight distribution; Based on the compression ratio parameter and weighted identification relationship, the remote sensing image data, surface deformation sequence data and multi-point sensor data in each partition are subject to partition coding control, and the compression ratio of the same spatial location is kept consistent under the time node sorting. Based on the compression results generated by the partition coding regulation, the data of adjacent time nodes are integrated and processed in combination with the time synchronization identifier to maintain the continuous feature expression of key areas in the time series.

5. The regional geological hazard risk assessment method based on multi-source data fusion according to claim 4, characterized in that, During the partitioned coding control process, the key fidelity zone maintains a consistent compression ratio parameter at each time node, and the brightness changes and signal amplitude are balanced according to the time synchronization identifier during the data integration processing of adjacent time nodes.

6. The regional geological hazard risk assessment method based on multi-source data fusion according to claim 4, characterized in that, The steps for performing temporal continuous tracking and difference balancing on the coded sequences of key regions based on the compression results after partition coding adjustment include: In the compressed data formed by partition coding control, key areas are selected based on the weighting identifier and compression ratio setting results, and the coding sequences corresponding to the key areas are sorted by time according to the time synchronization identifier to form a set of continuously arranged coding sequences. Based on the set of encoded sequences, the encoded sequences of adjacent time nodes are continuously tracked along the time axis, and the encoded segments of the same geographical location are kept consistent in the spatial coordinate level, while the compression parameter information of each time node is preserved. Based on the results of continuous time tracking, the signal brightness difference balancing process is performed on the encoded sequences of adjacent time periods, and the brightness change amplitude is adjusted under the constraint of time mapping relationship to form a continuously changing time series; Based on the time series formed by difference balancing, the temporal extension of the encoded sequence in key areas is maintained, and the brightness and signal amplitude are kept smoothly connected in continuous time periods.

7. A regional geological hazard risk assessment method based on multi-source data fusion according to claim 6, characterized in that, In the step of performing time-continuous tracking and differential balancing of the coded sequences in key areas, the differential balancing is based on the brightness change relationship of the corresponding coded sequences at adjacent time nodes, and the brightness or amplitude is adjusted for remote sensing image data, surface deformation sequence data and multi-point sensing data respectively while maintaining the consistency of time synchronization identifiers.

8. A regional geological hazard risk assessment method based on multi-source data fusion according to claim 6, characterized in that, The steps for performing cross-source data mapping and integration based on the difference balancing results after continuous time tracking include: In the compressed data formed by continuous time tracking and difference balancing, high-sensitivity detail areas corresponding to remote sensing image data, surface deformation sequence data and multi-point sensing data are extracted, and a time consistency index relationship is established based on the time synchronization identifier. Based on the temporal consistency index relationship, spatial coordinate alignment and feature mapping are performed on high-sensitivity detail areas to uniformly map remote sensing image data, surface deformation sequence data and multi-point sensor data to the same spatial reference system. Based on the spatial coordinate alignment and feature mapping results, dynamic matching and integration are performed on high-sensitivity detail regions of various data sources in the time series to form a multi-source feature set corresponding to both time and space dimensions; Based on the multi-source feature set formed by dynamic matching and integration, a multi-source complementary detailed restoration output is generated, and a regional geological hazard risk assessment is performed based on the detailed restoration output.

9. A regional geological hazard risk assessment method based on multi-source data fusion according to claim 8, characterized in that, In the step of generating multi-source complementary detailed restoration output, the high-sensitivity detailed regions of remote sensing image data, surface deformation sequence data and multi-point sensor data are reconstructed in time sequence according to the time synchronization identifier, and the high-sensitivity detailed regions are superimposed and expressed under a unified spatial reference system to form a multi-dimensional restoration result covering a continuous time period. Based on the multi-dimensional restoration result, the distribution of regional geological disaster risks is determined.

10. A regional geological hazard risk assessment system based on multi-source data fusion, used to implement the regional geological hazard risk assessment method based on multi-source data fusion as described in any one of claims 1-9, characterized in that, It includes a multi-source time-series acquisition module, a high-frequency energy identification module, a zone compression and control module, a time-series difference balancing module, and a cross-source detail fusion module; The multi-source time-series acquisition module collects raw signals of remote sensing image data, surface deformation sequence data and multi-point sensor data over a large area. It embeds time synchronization identifiers during the acquisition process and performs unified arrangement of the time series of various types of data. The high-frequency energy identification module extracts energy segments from the weak high-frequency signals based on the original signals after time synchronization, and locally weights and identifies the detailed signals of the surface micro-slip area, crack edge area and deformation abrupt area. The partitioned compression and control module performs partitioned coding control during the data compression stage based on the weighted identifier formed by energy segment extraction, and implements differentiated adjustment of the compression ratio for key areas; The timing difference balancing module combines the compression results after partition coding control to continuously track the coding sequence of key areas in time and perform difference balancing processing on the signal brightness of adjacent time periods. The cross-source detail fusion module performs cross-source data mapping and integration based on the difference balance results after continuous time tracking. It dynamically matches high-sensitivity detail regions in remote sensing image data, surface deformation sequence data, and multi-point sensor data to generate multi-source complementary detail restoration output.

Images