Method and system for dynamic prediction of underground settlement risk based on time-series remote sensing

Abstract

This invention discloses a method and system for dynamic prediction of subsidence risk based on time-series remote sensing, belonging to the field of prediction data processing technology. The method includes: acquiring a target ground area and sub-regions; collecting initial deformation sequences of the sub-regions; acquiring the stability ratio of the sub-regions; acquiring the difference index of adjacent sub-regions; acquiring the average stability ratio; if the difference index between the i-th sub-region and the j-th sub-region is lower than the product of the average stability ratio and a first preset threshold, then merging the i-th sub-region and the j-th sub-region to form a consistent prediction region, and merging to form the target deformation sequence of the consistent prediction region; acquiring the risk index of the consistent prediction region based on the target deformation sequence of the consistent prediction region; and acquiring the risk prediction result based on the risk index of the consistent prediction region, the average stability ratio, and a second preset threshold. This invention has the advantages of adaptability, collaborative correlation, and prediction sensitivity.

Description

Technical Field

[0001] This invention relates to the field of predictive data processing technology, specifically to a method and system for dynamic prediction of underground subsidence risk based on time-series remote sensing. Background Technology

[0002] The problem of land subsidence induced by excessive groundwater extraction, large-scale engineering construction, or geological activities is becoming increasingly prominent, and its potential risks include major disasters such as underground pipeline rupture, building structural instability, and ground collapse.

[0003] Currently, the prediction of land subsidence risk mainly relies on surface height change data collected by time-series remote sensing technology. However, on the one hand, existing methods can usually only analyze the deformation trend of a single monitoring area independently, ignoring the fact that subsidence often has correlations and synergies in spatial distribution. This isolated analysis mode is difficult to effectively capture the linkage deformation caused by the continuity of geological structure or load transfer effects in adjacent areas, which may lead to the omission of key risk information or misjudgment of abnormal areas. On the other hand, existing risk assessment models are insufficient in adaptability. Deformation sequences themselves are often accompanied by fluctuations due to measurement noise, environmental interference (such as cloud cover), or seasonal changes. Furthermore, the stability of data varies in different regions due to differences in geological conditions (e.g., large fluctuations in soft soil areas and small fluctuations in bedrock areas). If a fixed threshold is used to uniformly judge subsidence risk or divide areas, areas with low stability are prone to frequent false alarms due to noise interference, while areas with high stability may mask slow but continuous accelerating subsidence signals, seriously affecting the accuracy and reliability of prediction. Furthermore, existing technologies for quantifying risk often rely on cumulative settlement or a single rate value, making it difficult to sensitively identify potential dangerous acceleration trends during deformation (such as a sudden change from uniform settlement to accelerated settlement), resulting in delayed early warnings of critical risks. In summary, there is an urgent need for a dynamic prediction method that can effectively integrate spatiotemporal correlation information, adaptively handle data fluctuations, and sensitively capture signs of settlement acceleration, in order to improve the accuracy and timeliness of early warnings of regional settlement risks in complex geological environments. Summary of the Invention

[0004] In view of the technical problems described in the background art, the present invention provides a method and system for dynamic prediction of underground subsidence risk based on time-series remote sensing.

[0005] A dynamic prediction method for subsidence risk based on time-series remote sensing includes: acquiring a target ground area and dividing it into multiple sub-regions; collecting surface remote sensing subsidence data for each sub-region over multiple detection periods prior to the current time to form an initial deformation sequence; obtaining the stability ratio of each sub-region based on the initial deformation sequence of each sub-region, and obtaining the difference index of each adjacent sub-region based on the initial deformation sequences of each adjacent sub-region; obtaining the average stability ratio between the i-th and j-th sub-regions; if the difference index between the i-th and j-th sub-regions is lower than the product of the average stability ratio and a first preset threshold, then merging the i-th and j-th sub-regions to form a consistent prediction region; merging the two initial deformation sequences corresponding to the consistent prediction region to form the target deformation sequence of the consistent prediction region; obtaining the risk index of each consistent prediction region based on the target deformation sequence of each consistent prediction region; and obtaining the risk prediction result based on the risk index, average stability ratio, and second preset threshold of each consistent prediction region.

[0006] Optionally, obtaining risk indicators for each consistent prediction area based on the target deformation sequence of each consistent prediction area includes: subtracting the previous surface remote sensing subsidence from the next surface remote sensing subsidence in the adjacent data of the target deformation sequence of each consistent prediction area, and obtaining the subsidence difference; summing all subsidence differences in the target deformation sequence of each consistent prediction area, and obtaining the sum of differences; dividing the sum of differences in the target deformation sequences of each consistent prediction area by the average surface remote sensing subsidence of the target deformation sequence, and obtaining the risk indicator for each consistent prediction area.

[0007] Optionally, obtaining risk prediction results based on the risk indicators, average stability ratio, and second preset threshold of each consensus prediction area includes: multiplying the average stability ratio by the second preset threshold to obtain a correction threshold; if the risk indicator of the kth consensus prediction area exceeds the correction threshold, a risk prediction result indicating abnormal settlement is generated for the kth consensus prediction area, otherwise a risk prediction result indicating normal settlement is generated for the kth consensus prediction area.

[0008] Optionally, obtaining the stability ratio of each sub-region based on the initial deformation sequence of each sub-region includes: calculating the absolute value of the difference between the remotely sensed subsidence of adjacent surfaces in the initial deformation sequence of each sub-region and obtaining the absolute difference; obtaining the proportion of the absolute difference exceeding the difference threshold in the initial deformation sequence of each sub-region and using it as the stability ratio of each sub-region.

[0009] Optionally, obtaining the difference index of each adjacent sub-region based on the initial deformation sequence of each adjacent sub-region includes: obtaining the difference between the surface remote sensing subsidence in the initial deformation sequence of the i-th sub-region and the initial deformation sequence of the j-th sub-region in the nth detection period and calculating the absolute value to obtain the absolute difference in the nth detection period; obtaining the difference index between the i-th sub-region and the j-th sub-region based on the absolute difference in the initial deformation sequence of the i-th sub-region and the j-th sub-region in each detection period.

[0010] Optionally, fusing the two initial deformation sequences corresponding to the consistent prediction area to form the target deformation sequence of the consistent prediction area includes: taking the average of the sum of the surface remote sensing subsidence amounts in the nth detection period of the two initial deformation sequences corresponding to the consistent prediction area as the data in the nth detection period of the target deformation sequence of the consistent prediction area.

[0011] A dynamic prediction system for subsidence risk based on time-series remote sensing is also provided. The system includes: an acquisition module, used to acquire a target ground area and divide it into multiple sub-regions, collect the surface remote sensing subsidence of each sub-region within multiple detection cycles before the current time, and form an initial deformation sequence; a first data processing module, used to obtain the stability ratio of each sub-region based on the initial deformation sequence of each sub-region, and obtain the difference index of each adjacent sub-region based on the initial deformation sequence of each adjacent sub-region; a second data processing module, used to obtain the average stability ratio between the i-th sub-region and the j-th sub-region, if the difference index between the i-th sub-region and the j-th sub-region is lower than the product of the average stability ratio and a first preset threshold, then the i-th sub-region and the j-th sub-region are merged to form a consistent prediction region, and the two initial deformation sequences corresponding to the consistent prediction region are merged to form the target deformation sequence of the consistent prediction region; and a risk prediction module, used to obtain the risk index of each consistent prediction region based on the target deformation sequence of each consistent prediction region, and obtain the risk prediction result based on the risk index, average stability ratio, and second preset threshold of each consistent prediction region.

[0012] Optionally, the risk prediction module is also used to: subtract the previous land surface remote sensing subsidence from the next land surface remote sensing subsidence in the target deformation sequence of each consistent prediction area, and obtain the subsidence difference; add up all the subsidence differences in the target deformation sequence of each consistent prediction area, and obtain the sum of the differences; divide the sum of the differences in the target deformation sequence of each consistent prediction area by the average land surface remote sensing subsidence of the target deformation sequence, and obtain the risk index of each consistent prediction area.

[0013] Optionally, the risk prediction module is also used to: multiply the average stability ratio by a second preset threshold to obtain a correction threshold; if the risk index of the kth consistent prediction area exceeds the correction threshold, then generate a risk prediction result for the kth consistent prediction area with abnormal settlement, otherwise generate a risk prediction result for the kth consistent prediction area with normal settlement.

[0014] Optionally, the first data processing module is further configured to: calculate the absolute value of the difference between the remotely sensed subsidence amounts of adjacent surfaces in the initial deformation sequence of each sub-region, and obtain the absolute difference; obtain the proportion of absolute differences exceeding the difference threshold in the initial deformation sequence of each sub-region, and use it as the stable proportion of each sub-region.

[0015] The beneficial effects of this invention are reflected in:

[0016] In the entire dynamic prediction method for underground subsidence risk based on time-series remote sensing, firstly, the consistent prediction area is dynamically divided based on the difference index and the stability ratio, breaking through the limitations of isolated analysis by traditional fixed grids. This avoids information fragmentation in geologically continuous areas while preserving the signal strength of anomalous independent areas, thus improving spatial risk coverage. Furthermore, in the regional fusion stage, the lower the average stability ratio, the more stringent the threshold of the difference index (e.g., backfilled soil areas require strict coordination for merging), avoiding contamination of analysis data by low-reliability areas. Furthermore, in the risk assessment stage, the correction threshold is automatically scaled with the average stability ratio (threshold compression in soft soil areas triggers sensitive early warnings, while threshold relaxation in bedrock areas resists interference and false alarms), achieving tolerance adaptation driven by geological characteristics. Furthermore, risk indicators are calculated through nonlinear coupling of the sum of differences and the sequence mean (e.g., the expansion of cavities in industrial areas causes a sharp increase in the difference index jump), which, compared to cumulative subsidence or average rate, more sensitively identifies dangerous inflection points from uniform to accelerating rates (even if the total subsidence has not exceeded the threshold), providing an intervention window several cycles ahead of the development of underground diseases. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0018] Figure 1 This is a schematic diagram of part S1 to S4 of the dynamic prediction method for underground subsidence risk based on time-series remote sensing in this invention.

[0019] Figure 2 This is a schematic diagram of part S4 in the dynamic prediction method for underground subsidence risk based on time-series remote sensing of the present invention;

[0020] Figure 3This is a schematic diagram illustrating the steps of the dynamic prediction method for underground subsidence risk based on time-series remote sensing according to the present invention.

[0021] Figure 4 This is a schematic diagram of part of step S2 in the dynamic prediction method for underground subsidence risk based on time-series remote sensing of the present invention;

[0022] Figure 5 This is a schematic diagram of another part of step S2 in the dynamic prediction method for underground subsidence risk based on time-series remote sensing of the present invention.

[0023] Figure 6 This is a schematic diagram of part of step S4 in the dynamic prediction method for underground subsidence risk based on time-series remote sensing of the present invention;

[0024] Figure 7 This is a schematic diagram of another part of step S4 in the dynamic prediction method for underground subsidence risk based on time-series remote sensing of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0026] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0027] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0028] like Figures 1 to 3 As shown, a method for dynamic prediction of underground subsidence risk based on time-series remote sensing is provided. In one embodiment, the method includes:

[0029] S1. Obtain the target ground area and divide it into multiple sub-regions. Collect the surface remote sensing subsidence of each sub-region within multiple detection cycles before the current time and form an initial deformation sequence.

[0030] S2. Obtain the stable proportion of each sub-region based on the initial deformation sequence of each sub-region, and obtain the difference index of each adjacent sub-region based on the initial deformation sequence of each adjacent sub-region.

[0031] S3. Obtain the average stability ratio between the i-th sub-region and the j-th sub-region. If the difference index between the i-th sub-region and the j-th sub-region is lower than the product of the average stability ratio and the first preset threshold, then merge the i-th sub-region and the j-th sub-region to form a consistent prediction region. Merge the two initial deformation sequences corresponding to the consistent prediction region to form the target deformation sequence of the consistent prediction region.

[0032] S4. Obtain the risk indicators of each consistent prediction region based on the target deformation sequence of each consistent prediction region, and obtain the risk prediction results based on the risk indicators, average stability ratio and second preset threshold of each consistent prediction region.

[0033] In this embodiment, it should be noted that the key to acquiring the target ground area and dividing it into multiple sub-regions in S1 lies in adapting to the spatial heterogeneity of the geological structure. For example, when monitoring the subsidence risk of a new urban area, it is necessary to comprehensively consider the gradient of groundwater level changes, the distribution of soil compressibility, and differences in building loads. If the entire new area is regarded as a single region, it may obscure local accelerated subsidence points (such as soft soil layers along subway tunnels). Therefore, the sub-region division needs to take into account both geological exploration results and infrastructure layout. For example, using a 500m×500m grid as the basic unit, the grid around important facilities (bridge pile foundations, underground pipe corridors) can be further densified to 100m×100m to ensure that high-risk areas are monitored in detail. This division can be achieved through simple geometric cutting.

[0034] Subsequently, assuming the use of satellite interferometric synthetic aperture radar (InSAR) technology to acquire elevation data every 12 days, each detection cycle is a 12-day interval. For a certain sub-region of fill area, the average height change of each cycle within a continuous 24 months (about 60 cycles) is extracted (e.g., the ground surface decreases by 1.2 mm from cycle 1 to cycle 2, and by 0.8 mm from cycle 2 to cycle 3, etc.). These height changes arranged in chronological order constitute the "initial deformation sequence".

[0035] Furthermore, the initial deformation sequence is essentially a dynamic fingerprint reflecting the geological response of a subregion, and its construction process implicitly incorporates adaptive treatment of the subsidence mechanism. For example, in coastal soft soil areas, due to the lag effect of pore water pressure dissipation, periodic fluctuations may occur in the sequence (accelerated subsidence during the rainy season and slowed down during the dry season). Such phenomena need to be distinguished from actual geological risks. Therefore, the surface remote sensing subsidence in the sequence is not the original observation value, but a relative change after calibration by ground control points.

[0036] Taking the clay proton area of ​​a certain industrial park as an example, the sequence from January 2023 to January 2024 shows that the settlement was stable at 0.5-0.7 mm per month for the first 6 months (during the factory's uniform loading period), and then suddenly increased to 1.5-2.0 mm per month for the last 6 months (due to the compression of the aquifer induced by the rupture of the underground water pipe). This kind of sequence, which includes an acceleration inflection point, can reveal potential dangers better than simply the cumulative settlement.

[0037] For example, in a highway subgrade sub-region, if the InSAR sequence shows that the settlement exceeds the threshold for three consecutive cycles (e.g., >3mm / cycle), it is necessary to simultaneously check whether the tiltmeter data for that section is abnormal (excessive horizontal displacement may cause slope instability) to confirm the engineering significance of the remote sensing data. It is worth emphasizing that sub-regional scale sequences overcome the drawbacks of averaging over a large area: a residential sub-region experiences a rebound of 0.3mm / cycle due to the closure of a deep pumping well, while an adjacent commercial area experiences a settlement of 1.1mm / cycle due to new foundation excavation. If the two are combined for analysis, it will lead to misjudgment of risk; however, independent sequences can capture the stable signal of the rebound area and the accelerated signal of the excavation area separately, providing a data basis for the stability assessment of S2 and the regional fusion of S3, ultimately achieving the goal of dynamic prediction and identifying the inflection point of accelerated settlement before visible cracks on the surface.

[0038] In S2, the stability ratio is derived from the temporal volatility analysis of the initial deformation sequence of each sub-region: for any sub-region, the calculation result of the absolute value of the settlement difference between adjacent detection periods in its sequence (S21) directly reflects the intensity of short-term deformation oscillations. Taking the coastal backfill area as an example, if the groundwater level fluctuates frequently due to tides, the settlement difference between adjacent periods in the sequence may exceed the preset difference threshold (e.g., 1.0 mm) multiple times. At this time, the proportion of differences exceeding the threshold is extremely high, and the stability ratio approaches 0. Conversely, in the rock foundation distribution area, due to the low compressibility of the soil, the settlement difference between adjacent periods is extremely small (e.g., the absolute value of the difference is generally less than 0.2 mm), and the proportion of differences exceeding the threshold is close to 0, while the stability ratio approaches 1. This parameter not only indicates the reliability of the data of the region itself (low noise interference in high-proportion areas), but also deeply reflects the inherent properties of the geological medium—the volatility caused by the lag in the dissipation of pore water pressure in soft soil areas is quantified as a low stability ratio, while the mechanical rigidity of rock foundation areas is manifested as a high stability ratio.

[0039] Furthermore, the difference index focuses on the spatial dimension, revealing the spatial correlation of geological responses by comparing the settlement behavior of adjacent sub-regions at the same time point (S23). A typical example is two sub-regions spanning the same aquifer: if pumping water from the aquifer causes uniform settlement, the difference in settlement between the two regions in each period is minimal, and the proportion of the statistical absolute difference exceeding the preset difference (such as the regional allowable fluctuation tolerance) is extremely low, with the difference index approaching 0; however, if a concealed fault exists beneath a sub-region, causing its settlement rate to be significantly higher than that of adjacent regions, the absolute difference in each period continuously exceeds the tolerance, the proportion of statistical exceedance increases, and the difference index increases significantly. The essence of the difference index is to measure the degree of failure of spatial coordinated deformation—an index approaching 0 indicates that load transfer or geological continuity has not been disrupted, while an increase in the index warns of local abnormal disturbances or boundary effects.

[0040] Furthermore, the stability ratio is calculated using the over-threshold percentage method. Its physical significance lies in distinguishing between normal fluctuations and abnormal jumps: In permafrost regions, the settlement difference between adjacent periods during the thawing period may periodically exceed the threshold due to frost heave. However, if the percentage of over-threshold points does not exceed 60% (the difference threshold is set according to regional characteristics), it is still considered a controllable fluctuation, and the stability ratio remains at 0.4. In contrast, in a mining area, due to goaf collapse, the single-period settlement difference suddenly increases three times the threshold and remains high thereafter. When the over-threshold percentage reaches 90%, the stability ratio drops below 0.1. This mechanism avoids the shortcomings of fixed difference standards that cannot adapt to regional characteristics, providing an adaptive basis for the S3 fusion judgment.

[0041] Furthermore, the acquisition of the difference index relies on cumulative statistical analysis of differences across cycles, emphasizing the verification of spatiotemporal consistency. Taking adjacent areas crossing a subway tunnel as an example, the difference index between the two areas was stable at 0.1 before tunnel construction (indicating stratum homogeneity). During the shield tunneling period, the sub-region directly above the tunnel experienced accelerated settlement, while the deformation of the lateral region lagged behind, resulting in the absolute difference exceeding the preset value for five consecutive periods (e.g., the difference increased from the normal 1mm to 4mm). The surge in the proportion of periods exceeding the standard caused the difference index to rise to 0.8. This parameter breaks through the existing static neighborhood analysis method, coupling temporal evolution with spatial comparison, providing a quantitative basis for identifying local disconnection caused by engineering activities.

[0042] In S3, based on the stability ratio and difference index quantified in S2, prediction units with geological consistency are dynamically divided. A prerequisite for regional fusion is that the settlement behavior of adjacent sub-regions exhibits spatial synergy—that is, the difference index must be lower than the product threshold formed by the average stability ratio of the two regions and a first preset threshold. The first preset threshold, determined through a limited number of historical engineering experiments, is between 0.4 and 0.6. A first preset threshold of 0.5 (default value) is suitable for most homogeneous soil layers (such as alluvial plains); a first preset threshold of 0.4 (strict mode) is used for high-risk areas (such as along subway lines) to prevent backfill soil disturbance from contaminating bedrock data; and a first preset threshold of 0.6 (relaxed mode) is suitable for homogeneous soft soil areas, enhancing noise suppression capabilities.

[0043] For example, two sub-regions on an alluvial plain (sub-region A has a stability ratio of 0.9 and sub-region B has a stability ratio of 0.8), with an average stability ratio of 0.85. If the two sub-regions have synchronous subsidence trends due to sharing an aquifer, and the difference index is only 0.1 (e.g., the subsidence deviation in each cycle is consistently less than the tolerance), and the product threshold is 0.425 when the first preset threshold is 0.5, and the difference index of 0.1 is significantly lower than 0.425, then the fusion mechanism is triggered.

[0044] Furthermore, the resulting consistent prediction region after fusion essentially represents a geomechanical response unit whose boundary is no longer constrained by the initial geometric grid, but is naturally defined by the continuity of load transfer or the homogeneity of the soil and rock layers. Sequence fusion employs periodic-level averaging: the settlement amounts at the same time points of the two original sequences are averaged (e.g., in period 1, region A settles 1.0 mm and region B settles 1.2 mm, the fused sequence is recorded as 1.1 mm). This operation suppresses local noise in the spatial dimension. However, it should be noted that if a region experiences strong isolated disturbances (e.g., a sudden acceleration of settlement in a local location of sub-region B due to pipeline leakage), the difference index increases, causing the fusion conditions to be unmet. In such cases, independent analysis is retained to avoid signal dilution.

[0045] Furthermore, the lower the average stability ratio (indicating large data fluctuations in at least one region), the smaller the product threshold and the stricter the fusion conditions. For example, the average stability ratio of a backfill soil sub-region (stability ratio 0.3) and the adjacent bedrock sub-region (stability ratio 0.9) in a mining area is only 0.6. Even if the difference index between the two is 0.25 (due to some periodic exceedances caused by backfill soil settlement fluctuations), the product threshold is 0.3 when the first preset threshold of 0.5 is set. Although the difference index of 0.25 is lower than 0.3, it is close to the critical value. After fusion, the abnormal fluctuations of backfill soil in the target deformation sequence will be partially masked by the bedrock data. Therefore, low-reliability regions must be merged with highly coordinated regions (e.g., a backfill soil region can only be merged with a region of the same type of soil and with extremely small differences).

[0046] For example, in cases involving cross-faults, the difference index between the two sub-regions on either side of the fault reaches 0.7 due to rock mass displacement (e.g., continuous subsidence on the north side and stability on the south side). Even if both sides have a relatively high stability ratio (0.8), the product threshold of the average stability ratio (0.8 × preset threshold 0.5) = 0.4 is still far below 0.7, thus merging is rejected. This measure ensures that the function of the fault as a geological hazard boundary is not blurred, maintaining the independent early warning capability of high-risk areas.

[0047] In S4, the calculation of the risk index reveals the intensity of the inherent trend change in the sequence: the continuous accumulation of the difference between adjacent settlements in the target deformation sequence (the sum of the differences) essentially captures the cumulative fluctuation energy, and then the difference is divided by the sequence average to eliminate the interference of the absolute settlement scale.

[0048] For example, in the initial stage of a target sequence after the integration of an industrial zone, uniform settlement is observed (the difference between adjacent values ​​remains stable at 0.5 mm). Later, due to the expansion of underground cavities, the difference increases sharply to 2.0 mm, and the total difference rises significantly. However, the sequence mean, affected by the earlier stage, only increases slowly from 0.6 mm to 0.9 mm. At this point, the risk index still jumps due to the surge in the number of molecules, making it more sensitive to identify acceleration inflection points compared to existing rate indices. This design effectively distinguishes between dangerous acceleration (sudden increase in the total difference) and low-risk uniform settlement (stable total difference), and can provide early warning even if the total settlement does not reach the warning value.

[0049] Furthermore, the risk assessment introduces an average stability ratio to dynamically adjust the second threshold: for areas with a low stability ratio (such as a stability ratio of 0.3 in the backfill soil fusion zone of the mining area), the adjustment threshold is compressed to 30% of the original threshold (assuming the second threshold is 0.5 → adjusted to 0.15), and an alarm is triggered even with a slight acceleration (risk index 0.18); while for areas with a high stability ratio (such as a stability ratio of 0.9 in the bedrock fusion zone), the adjustment threshold is expanded to 0.45, and an alarm is only triggered when the risk index exceeds 0.45, in order to avoid false alarms caused by slight vibrations of the rock strata.

[0050] Furthermore, the dynamic scaling of the correction threshold is essentially an adaptation of the risk tolerance of the reliability of the fusion region: the sequence in the region with a low stability ratio fluctuates greatly (such as the natural high sum of differences due to precipitation disturbance in the soft soil region). By lowering the correction threshold, the monitoring sensitivity is forcibly increased, which is equivalent to amplifying the danger signal in the background of noise.

[0051] For example, if the risk index of a target sequence for land reclamation is 0.25, and the regional stability ratio is only 0.4 (affected by tides), the correction threshold is adjusted to 0.2 (0.4 × 0.5). Then, 0.25 > 0.2 triggers an early warning, allowing for timely response to potential dam leakage hazards. Conversely, if the same risk index appears in a clay fusion zone with a stability ratio of 0.8 (correction threshold 0.4), it is considered normal settlement. For areas with strong disturbances and a stability ratio of 0.2, the correction threshold is lowered to 0.1, requiring the risk index to exceed 0.1 before an alarm is triggered, avoiding frequent false alarms due to normal fluctuations. For areas with a stability ratio of 0.9, the correction threshold is raised to 0.45, requiring the risk index to exceed a higher threshold before an alarm is triggered, preventing the omission of slowly accelerating deep landslide risks. The final output is a binary decision—abnormal settlement or normal settlement. Its physical meaning lies in identifying fusion zones where the accelerating energy exceeds the geologically reliable boundary (correction threshold), providing a priority basis for engineering treatment.

[0052] In summary, the entire time-series remote sensing-based dynamic prediction method for underground subsidence risk, by dynamically dividing the consistent prediction area based on difference indicators and stability ratios, breaks through the limitations of isolated analysis using traditional fixed grids. This avoids information fragmentation in geologically continuous areas while preserving the signal strength of anomalous independent areas, thus improving spatial risk coverage. Furthermore, in the regional fusion stage, the lower the average stability ratio, the more stringent the difference indicator threshold (e.g., backfilled soil areas require strict coordination for merging), avoiding contamination of analysis data by low-reliability areas. Further, in the risk assessment stage, the correction threshold automatically scales with the average stability ratio (threshold compression in soft soil areas triggers sensitive early warnings, while threshold relaxation in bedrock areas resists interference and false alarms), achieving tolerance adaptation driven by geological characteristics. Furthermore, by nonlinearly coupling the sum of differences with the sequence mean to calculate risk indicators (e.g., the expansion of cavities in industrial areas causes a sharp increase in the difference indicator jump), compared to cumulative subsidence or average rate, it more sensitively identifies dangerous inflection points from uniform to accelerating rates (even if the total subsidence does not exceed the threshold), providing an intervention window several cycles ahead of the development of underground diseases.

[0053] like Figure 6 As shown, in one embodiment, obtaining the risk index of each consistent prediction region in S4 based on the target deformation sequence of each consistent prediction region includes:

[0054] S41. Subtract the previous surface remote sensing subsidence from the next surface remote sensing subsidence in the target deformation sequence of each consistent prediction area, and obtain the subsidence difference.

[0055] S42. Add up all the settlement differences in the target deformation sequence of each consistent prediction area and get the sum of the differences;

[0056] S43. Divide the sum of the differences of the target deformation sequences in each consistent prediction area by the average value of the surface remote sensing subsidence of the target deformation sequence, and obtain the risk index of each consistent prediction area.

[0057] In this embodiment, it should be noted that in S41, the change gradient of adjacent time points is extracted from the target deformation sequence of the consistency prediction region. Specifically, for the target deformation sequence of each consistency prediction region (such as the industrial area fusion sequence), the difference between the settlement amount of the next detection cycle and the previous cycle is calculated sequentially in chronological order.

[0058] For example, if the settlement of a certain region's sequence is continuously [1.2mm, 1.5mm, 1.0mm, 2.3mm, 3.1mm] from period 1 to period 5, then subtracting period 1 from period 2 yields +0.3mm, subtracting period 2 from period 3 yields -0.5mm, subtracting period 3 from period 4 yields +1.3mm, and so on. These differences characterize short-term settlement acceleration; positive values ​​indicate accelerated settlement, while negative values ​​indicate rebound or deceleration. Their physical significance lies in capturing the intensity of changes in deformation kinetic energy within adjacent periods.

[0059] In S42, the settlement difference sequence generated in S41 is accumulated. Continuing the previous example, all differences in [+0.3mm, -0.5mm, +1.3mm, +0.8mm] are summed to obtain a total difference of 0.3-0.5+1.3+0.8=1.9mm.

[0060] For example, when a certain area settles at a constant speed, the sum of the differences approaches zero (e.g., the sum of the differences in the sequence [1.0, 1.1, 1.2, 1.3] is only 0.3), but when there is a cavity collapse, the sum increases significantly due to the sharp increase in the period difference (e.g., from +0.2mm to +1.5mm).

[0061] In S43, taking the aforementioned difference sum of 1.9 mm as an example, the mean of the target sequence [1.2, 1.5, 1.0, 2.3, 3.1] is calculated to be approximately 1.82 mm, and the risk index = 1.9 / 1.82 ≈ 1.04. Here, the numerator (sum of differences) represents the cumulative fluctuation intensity, and the denominator (sequence mean) eliminates the interference of absolute subsidence (e.g., the acceleration characteristics of a region with an annual subsidence of 5 mm are comparable to those of a region with an annual subsidence of 50 mm). For uniform subsidence sequences, the index approaches 0 (ideal state) due to the low sum of differences and moderate mean, while sequences containing acceleration inflection points (e.g., a sudden increase in differences in the later stages of a collapse zone) cause the numerator's increase to far exceed that of the denominator. For example, in a certain region, the mean in the early stage is 1.5 mm, and the difference sum is 1.0 (index 0.67). After collapse, the mean increases to 2.0 mm, while the difference sum rises to 4.0 (index 2.0), a significantly higher increase than that calculated using linear rates.

[0062] like Figure 7As shown, in one embodiment, S4, obtaining the risk prediction result based on the risk index, average stability ratio, and second preset threshold of each consistent prediction region includes:

[0063] S44. Multiply the average stability ratio by the second preset threshold and obtain the correction threshold;

[0064] S45. If the risk index of the kth consistent prediction area exceeds the correction threshold, a risk prediction result of abnormal settlement is generated for the kth consistent prediction area; otherwise, a risk prediction result of normal settlement is generated for the kth consistent prediction area.

[0065] In this embodiment, it should be noted that in S44, the second preset threshold is corrected based on the average stability ratio of the consistent prediction region (calculated during fusion in S3). The second preset threshold is determined based on historical data of a limited number of accelerated subsidence disasters, statistically analyzing the 95% acceleration inflection point of the risk index. For example, when the risk index is 0.6, the number of historical accelerated subsidence disasters increases significantly, then the second preset threshold is 0.6 * 0.95 = 0.57. The second preset threshold is calibrated using a limited number of historical acceleration events (such as data from the three periods before collapse), and is generally between 0.4 and 0.6.

[0066] For example, if the stability ratio of the backfill soil fusion zone is 0.3, and the second threshold is set to 0.5, then the correction threshold = 0.3 × 0.5 = 0.15; when the stability ratio of the bedrock zone is 0.9, the correction threshold is 0.45. In low-stability zones (e.g., 0.3), data fluctuations are large, so the threshold is compressed (0.15) to improve early warning sensitivity, forcing an alarm to be triggered even with slight acceleration; in high-stability zones (e.g., 0.9), noise interference is weak, so the threshold is relaxed (0.45) to avoid false alarms due to normal rock mass creep. The correction process essentially sets differentiated risk tolerance boundaries for different geological units—soft soil zones need to tolerate high-frequency fluctuations, so the alarm threshold is lowered; rock mass zones have small natural fluctuations, so the alarm requirements are increased.

[0067] In S45, the risk index of a certain reclamation area is 0.25. If its stability ratio is 0.4 and the correction threshold is 0.2 (0.4×0.5), then 0.25>0.2 will trigger an "abnormal settlement" alarm. If the same risk index of 0.25 appears in the clay area (stability ratio 0.8, correction threshold 0.4), then it is judged as "normal settlement".

[0068] Specifically, the threshold for low reliability areas (stability ratio 0.2) is lowered to 0.1 to avoid false alarms due to normal high-frequency fluctuations (an alarm is only triggered when the indicator is >0.1); the threshold for high reliability areas (stability ratio 0.9) is raised to 0.45 to prevent missed alarms for deep, slow landslides (an alarm is only triggered when there is a sharp acceleration).

[0069] like Figure 4As shown, in one embodiment, obtaining the stable proportion of each sub-region based on the initial deformation sequence of each sub-region in S2 includes:

[0070] S21. Calculate the absolute value of the difference between the remote sensing subsidence of adjacent surfaces in the initial deformation sequence of each sub-region, and obtain the absolute difference.

[0071] S22. Obtain the percentage of absolute differences exceeding the difference threshold in the initial deformation sequence of each sub-region, and use it as the stable proportion of each sub-region.

[0072] In this embodiment, it should be noted that in S21, the short-term fluctuation intensity of the deformation sequence is captured by calculating the absolute value of the difference in settlement between adjacent detection cycles. Taking a certain frozen soil sub-region as an example, its initial deformation sequence shows a sudden increase in settlement during the snowmelt period (e.g., cycle 2 is 1.8 mm more than cycle 1) and a slight rebound during the freezing period (cycle 3 is 0.3 mm less than cycle 2). Then, the absolute value sequence of the difference between adjacent cycles is [|1.8|,|-0.3|,...]=[1.8,0.3,...]. This operation removes the influence of the sign and purely quantifies the deformation oscillation amplitude during the cycle: high values ​​(e.g., 1.8 mm) indicate a strong geological response (frozen soil thawing and compression), and low values ​​(e.g., 0.3 mm) reflect a stable state.

[0073] In S22, the proportion exceeding a preset difference threshold is used as the stability proportion. The difference threshold can be referenced by the 80th percentile of the absolute value of adjacent differences in the previous 10% of detection periods (e.g., if the historical difference sequence for a certain area is [0.3, 0.7, 0.2, 0.6, 1.1] mm, the 80th percentile ≈ 0.85 mm). Example: In a sub-region of soft soil, due to seasonal fluctuations in groundwater level, the difference sequence shows 0.3 mm, 1.1 mm, and 0.9 mm. Calculating the threshold according to the 80th percentile rule, it is 0.95 mm. Therefore, the proportion exceeding the threshold is only 33% (1.1 mm in period 2 exceeds the standard), and the stability proportion = 1 - 33% = 0.67.

[0074] For example, a sub-regional sequence in a mining area contains 20 adjacent periodic differences. 15 of these differences exceed the soft soil threshold of 1.0 mm due to disturbance in the goaf, representing 75% of the differences. Therefore, the stability ratio is 1 - 75% = 0.25. This ratio reflects the region's resistance to disturbance—in the permafrost region, when the periodic exceedance ratio is 40% (due to freeze-thaw cycles), the stability ratio remains at 0.6, unlike the 0.1 ratio in the subsidence region where 90% exceed the threshold.

[0075] like Figure 5 As shown, in one embodiment, obtaining the difference index of each adjacent sub-region based on the initial deformation sequence of each adjacent sub-region in S2 includes:

[0076] S23. Obtain the difference between the initial deformation sequence of the i-th sub-region and the initial deformation sequence of the j-th sub-region under the n-th detection period, calculate the absolute value, and obtain the absolute difference under the n-th detection period.

[0077] S24. Obtain the difference index between the i-th sub-region and the j-th sub-region based on the absolute difference in the initial deformation sequence of the i-th sub-region and the j-th sub-region under each detection cycle.

[0078] In this embodiment, it should be noted that in S23, the synchronous deformation behavior is compared periodically for adjacent sub-regions (such as regions A and B sharing an aquifer). In undisturbed homogeneous soil layers, region A settles by 1.0 mm and region B settles by 1.05 mm in period 1, with an absolute difference of |1.0-1.05|=0.05 mm; when fault activity causes region B to suddenly drop to 1.8 mm in period 5 (region A remains at 1.0 mm), the difference jumps to 0.8 mm.

[0079] This operation essentially detects the instantaneous breakpoint of spatial coordination—low difference (<0.1mm) indicates uniform load transfer, while high difference (>0.5mm) warns of local anomalies (such as fault displacement or pipeline leakage), and its output is the difference under the period.

[0080] In S24, the percentage of absolute differences generated in each cycle from S23 that exceed preset differences (e.g., 0.5mm in the structural zone and 0.2mm in the soil zone) is calculated. For example, in the two sub-regions crossing the subway tunnel, the difference exceeded 0.5mm only once in the first 10 cycles (10% proportion, difference index 0.1); during the shield tunneling crossing period, the difference exceeded the threshold for 6 consecutive cycles (60% proportion, difference index 0.6). This index integrates the time and spatial dimensions: an index of 0.1 represents 10% of the cycles experiencing collaborative failure (negligible), and 0.6 represents 60% of the cycles experiencing disconnection (requiring isolation analysis).

[0081] It should also be noted that in S24, the difference index between the i-th and j-th sub-regions is obtained based on the absolute difference in the initial deformation sequence of the i-th and j-th sub-regions under each detection cycle, and is expressed as follows:

[0082] ;in,

[0083] Let i be the difference index between the i-th sub-region and the j-th sub-region. This represents the number of detection cycles in the initial deformation sequence. This represents the absolute difference between the i-th and j-th sub-regions in the initial deformation sequence during the n-th detection period. This represents the standard difference in the initial deformation sequences of adjacent sub-regions within the same detection period.

[0084] It should also be noted that throughout the entire expression, absolute difference Difference from standard The comparison results are binarized. Specifically, when... (Differences are within tolerable limits) Output 0 or -1, after The function processes the data and outputs 0; when... (Differences exceeding tolerance range) Output +1, after The function processes the data and outputs 1. When determining the values, if the geological scenario is a homogeneous aquifer / alluvial plain, the standard difference is taken as 10% of the average settlement (e.g., average 1.0mm → 0.1mm), because the spatial difference in settlement due to hydraulic gradient changes is ≤10%. If the geological scenario is a fault-affected zone, the standard difference is taken as 0.3mm for the rock mass area and 0.5mm for the soil area, based on the bedrock displacement limit / soil shear deformation critical value. If the geological scenario is an underground engineering disturbance zone (along the tunnel), the standard difference is taken as 15% of the average settlement (e.g., average 2.0mm → 0.3mm), because the upper limit of the differential settlement gradient caused by excavation unloading is applied. If the historical difference between the two areas during the undisturbed period is stable at a certain value (e.g., 0.15mm), then it is directly set to a certain value. .

[0085] Furthermore, existing methods struggle to distinguish between slight and significant exceedances. The aforementioned binarization method forces a focus on whether synergy is compromised, avoiding interference from minor deviations in the judgment (e.g., normal fluctuation of 1.1 mm vs. tolerance of 1.0 mm in soft soil areas). For boundary-sensitive areas such as faults and tunnels, once the difference exceeds the geologically permissible value (e.g., difference of 0.8 mm due to rock strata displacement > bedrock tolerance of 0.3 mm), it is immediately marked as 1, providing a clear basis for spatial isolation.

[0086] Furthermore, The system accumulates the number of collaborative failures across all M detection cycles. If any cycle's difference exceeds the limit, the accumulated value is incremented by 1; otherwise, it is incremented by 0. Consecutive cycles exceeding the limit (e.g., exceeding the limit in cycles 5 / 6 of the tunnel boring machine's crossing period) will significantly increase the accumulated value (0.83), while transient interference (e.g., equipment error) only exceeds the limit in a single cycle (1 / 6 ≈ 0.17). Short-term fluctuations (e.g., cloud cover causing single-cycle data anomalies) are automatically filtered out because they do not form a continuous failure signal (accounting for <20%).

[0087] Furthermore, The cumulative number of failures is converted into the proportion within the [0,1] interval. For homogeneous aquifers, the difference index approaches 0 (failure rate <5%), triggering fusion to form a large-area analysis; for fault zones, the index approaches 1 (failure rate >80%), forcibly isolating and preserving independent risk units. This provides a quantitative basis for fusion: when... If the value is less than the average stability ratio multiplied by the threshold (e.g., 0.1 < 0.8 × 0.5 = 0.4), it is determined that the geological continuity has not been disrupted and fusion is allowed; otherwise (e.g., 0.7 > 0.4), fusion is rejected.

[0088] In one implementation, S3, which merges the two initial deformation sequences corresponding to the consistent prediction region to form the target deformation sequence of the consistent prediction region, includes:

[0089] The average of the sum of the surface remote sensing subsidence amounts in the nth detection period of the two initial deformation sequences corresponding to the consistent prediction area is used as the data in the nth detection period of the target deformation sequence of the consistent prediction area.

[0090] In this embodiment, it should be noted that for two sub-regions that have been determined to have geological consistency (such as region A and region B sharing an aquifer), the target deformation sequence is generated by aligning data points periodically and taking the average value.

[0091] Specifically, for any detection period n, the surface remote sensing subsidence amounts of the two sub-regions in that period are added together and divided by 2 to obtain the subsidence amount for the nth period of the fused consistent prediction area. For example, when the subsidence amount recorded in area A in period 3 is 1.5 mm and in area B is 1.7 mm, the fused sequence is recorded as 1.6 mm ((1.5+1.7) / 2) in that period. This fusion mechanism achieves dual optimization in the spatial dimension—it suppresses local measurement noise through an averaging algorithm (e.g., a 0.3 mm anomalous jump caused by a single-point instrument error in area A will be diluted by normal data in area B), while preserving the common characteristics of trend deformation (e.g., the 0.2 mm acceleration signal that occurs synchronously in both areas due to aquifer compression is completely preserved).

[0092] A dynamic prediction system for underground subsidence risk based on time-series remote sensing is also provided, the system including:

[0093] The acquisition module is used to acquire the target ground area and divide it into multiple sub-regions, collect the surface remote sensing subsidence of each sub-region within multiple detection cycles before the current time, and form an initial deformation sequence.

[0094] The first data processing module is used to obtain the stable proportion of each sub-region based on the initial deformation sequence of each sub-region, and to obtain the difference index of each adjacent sub-region based on the initial deformation sequence of each adjacent sub-region.

[0095] The second data processing module is used to obtain the average stability ratio between the i-th sub-region and the j-th sub-region. If the difference index between the i-th sub-region and the j-th sub-region is lower than the product of the average stability ratio and the first preset threshold, the i-th sub-region and the j-th sub-region are merged to form a consistent prediction region. The two initial deformation sequences corresponding to the consistent prediction region are merged to form the target deformation sequence of the consistent prediction region.

[0096] The risk prediction module is used to obtain the risk indicators of each consistent prediction region based on the target deformation sequence of each consistent prediction region, and to obtain the risk prediction results based on the risk indicators, average stability ratio and second preset threshold of each consistent prediction region.

[0097] In one implementation, the risk prediction module is further configured to: subtract the previous surface remote sensing subsidence from the subsequent surface remote sensing subsidence in the target deformation sequence of each consistent prediction area, and obtain the subsidence difference; add up all the subsidence differences in the target deformation sequence of each consistent prediction area, and obtain the sum of the differences; divide the sum of the differences in the target deformation sequence of each consistent prediction area by the average surface remote sensing subsidence of the target deformation sequence, and obtain the risk index of each consistent prediction area.

[0098] In one implementation, the risk prediction module is further configured to: multiply the average stability ratio by a second preset threshold to obtain a correction threshold; if the risk index of the kth consistent prediction area exceeds the correction threshold, generate a risk prediction result for the kth consistent prediction area with abnormal settlement, otherwise generate a risk prediction result for the kth consistent prediction area with normal settlement.

[0099] In one embodiment, the first data processing module is further configured to: calculate the absolute value of the difference between the remotely sensed subsidence amounts of adjacent surfaces in the initial deformation sequence of each sub-region, and obtain the absolute difference; obtain the proportion of absolute differences exceeding the difference threshold in the initial deformation sequence of each sub-region, and use it as the stable proportion of each sub-region.

[0100] In this embodiment, it should be noted that the specific method of performing the above-mentioned dynamic prediction system for underground subsidence risk based on time-series remote sensing has been described in detail in the embodiments of the dynamic prediction method for underground subsidence risk based on time-series remote sensing, and will not be elaborated here.

[0101] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.

[0102] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0103] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for dynamic prediction of underground subsidence risk based on time-series remote sensing, characterized in that, include: The target ground area is acquired and divided into multiple sub-regions. The surface remote sensing subsidence of each sub-region is collected in multiple detection cycles before the current time and an initial deformation sequence is formed. The absolute value of the difference between the remotely sensed subsidence of adjacent surfaces in the initial deformation sequence of each sub-region is calculated, and the absolute difference is obtained; the proportion of absolute differences exceeding the difference threshold in the initial deformation sequence of each sub-region is obtained and used as the stable proportion of each sub-region. Obtain the difference between the initial deformation sequence of the i-th subregion and the initial deformation sequence of the j-th subregion under the n-th detection period, calculate the absolute value, and obtain the absolute difference under the n-th detection period; obtain the difference index between the i-th subregion and the j-th subregion based on the absolute difference under each detection period in the initial deformation sequence of the i-th subregion and the j-th subregion. Obtain the average stability ratio between the i-th sub-region and the j-th sub-region. If the difference index between the i-th sub-region and the j-th sub-region is lower than the product of the average stability ratio and the first preset threshold, then merge the i-th sub-region and the j-th sub-region to form a consistent prediction region. Merge the two initial deformation sequences corresponding to the consistent prediction region to form the target deformation sequence of the consistent prediction region. Risk indicators for each consistent prediction region are obtained based on the target deformation sequence of each consistent prediction region, and risk prediction results are obtained based on the risk indicators, average stability ratio, and second preset threshold of each consistent prediction region.

2. The method for dynamic prediction of underground subsidence risk based on time-series remote sensing according to claim 1, characterized in that, The process of obtaining risk indicators for each consistent prediction region based on the target deformation sequence of each consistent prediction region includes: Subtract the previous land surface remote sensing subsidence from the next land surface remote sensing subsidence in the target deformation sequence of each consistent prediction area, and obtain the subsidence difference. Add up all the settlement differences in the target deformation sequence of each consistent prediction area and get the sum of the differences; The risk index for each consistent prediction area is obtained by dividing the sum of the differences in the target deformation sequences of each consistent prediction area by the average surface remote sensing subsidence of the target deformation sequences.

3. The method for dynamic prediction of underground subsidence risk based on time-series remote sensing according to claim 1, characterized in that, The step of obtaining risk prediction results based on risk indicators, average stability ratios, and a second preset threshold for each consistent prediction region includes: Multiply the average stability ratio by the second preset threshold to obtain the correction threshold; If the risk index of the kth consistent prediction area exceeds the correction threshold, a risk prediction result of abnormal settlement is generated for the kth consistent prediction area; otherwise, a risk prediction result of normal settlement is generated for the kth consistent prediction area.

4. The method for dynamic prediction of underground subsidence risk based on time-series remote sensing according to claim 1, characterized in that, The process of fusing the two initial deformation sequences corresponding to the consistent prediction region to form the target deformation sequence of the consistent prediction region includes: The average of the sum of the surface remote sensing subsidence amounts in the nth detection period of the two initial deformation sequences corresponding to the consistent prediction area is used as the data in the nth detection period of the target deformation sequence of the consistent prediction area.

5. A dynamic prediction system for underground subsidence risk based on time-series remote sensing, characterized in that, The system is used to implement the dynamic prediction method for underground subsidence risk based on time-series remote sensing as described in any one of claims 1 to 4, the system comprising: The acquisition module is used to acquire the target ground area and divide it into multiple sub-regions, collect the surface remote sensing subsidence of each sub-region within multiple detection cycles before the current time, and form an initial deformation sequence. The first data processing module is used to obtain the stable proportion of each sub-region based on the initial deformation sequence of each sub-region, and to obtain the difference index of each adjacent sub-region based on the initial deformation sequence of each adjacent sub-region. The second data processing module is used to obtain the average stability ratio between the i-th sub-region and the j-th sub-region. If the difference index between the i-th sub-region and the j-th sub-region is lower than the product of the average stability ratio and the first preset threshold, the i-th sub-region and the j-th sub-region are merged to form a consistent prediction region. The two initial deformation sequences corresponding to the consistent prediction region are merged to form the target deformation sequence of the consistent prediction region. The risk prediction module is used to obtain the risk indicators of each consistent prediction region based on the target deformation sequence of each consistent prediction region, and to obtain the risk prediction results based on the risk indicators, average stability ratio and second preset threshold of each consistent prediction region.

6. The dynamic prediction system for underground subsidence risk based on time-series remote sensing according to claim 5, characterized in that, The risk prediction module is also used for: Subtract the previous land surface remote sensing subsidence from the next land surface remote sensing subsidence in the target deformation sequence of each consistent prediction area, and obtain the subsidence difference. Add up all the settlement differences in the target deformation sequence of each consistent prediction area and get the sum of the differences; The risk index for each consistent prediction area is obtained by dividing the sum of the differences in the target deformation sequences of each consistent prediction area by the average surface remote sensing subsidence of the target deformation sequences.

7. The dynamic prediction system for underground subsidence risk based on time-series remote sensing according to claim 5, characterized in that, The risk prediction module is also used for: Multiply the average stability ratio by the second preset threshold to obtain the correction threshold; If the risk index of the kth consistent prediction area exceeds the correction threshold, a risk prediction result of abnormal settlement is generated for the kth consistent prediction area; otherwise, a risk prediction result of normal settlement is generated for the kth consistent prediction area.

8. The dynamic prediction system for underground subsidence risk based on time-series remote sensing according to claim 5, characterized in that, The first data processing module is also used for: The absolute value of the difference in remotely sensed subsidence between adjacent surfaces in the initial deformation sequence of each sub-region is calculated, and the absolute difference is obtained. Obtain the percentage of absolute differences exceeding the difference threshold in the initial deformation sequence of each sub-region, and use this percentage as the stable proportion of each sub-region.

Images