Point-shaped landslide and debris flow potential degree self-adaptive monitoring and early warning method

Abstract

The invention relates to the technical field of geological disaster monitoring, and discloses a point-like landslide and debris flow potential degree adaptive monitoring and early warning method. According to the method, a dot matrix monitoring equipment array is arranged, and displacement data and environmental parameters of multiple monitoring points are collected; analyzing the displacement data timestamp through a timeline analysis module, generating a dynamic timeline, and marking geological event nodes; a geographic information system is combined to carry out spatial mapping on displacement data, and environmental parameters are fused to obtain a preliminary potential degree evaluation result. The pattern recognition engine classifies displacement speed and deformation characteristics in the preliminary result to generate a deformation characteristic spectrum with a space-time label; calculating a dynamic potential degree evaluation map according to the multi-factor weighted evaluation model; and adjusting an early warning threshold according to a user instruction, generating a personalized early warning rule set, and scanning the atlas to output a multi-stage early warning response scheme. On the basis of an early warning feedback iteration updating model and an engine, monitoring data are newly added to trigger and re-analyze in real time, a related scheme is updated in a linkage mode, and different scene requirements are met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geological disaster monitoring, in particular to a point landslide and debris flow potential degree self-adaptive monitoring and early warning method. BACKGROUND

[0002] In the mountainous geological disaster prevention and control work, point landslide and debris flow has always been a difficult problem in the field of monitoring and early warning because of its dispersed occurrence position and complex triggering conditions. The traditional monitoring method mainly adopts single-point equipment layout, which can only obtain displacement or environmental data in a local area, and it is difficult to form a comprehensive perception of the overall geological body changes. When the landslide and debris flow hidden danger points are widely distributed, the limitations of single-point monitoring data will lead to deviation in the judgment of the disaster development process, and it is difficult to accurately capture the spatial correlation between different monitoring points. In the data processing link, the traditional method often ignores the dynamic change characteristics of the data in the time dimension, lacks systematic analysis of the timestamp information of the displacement data, and is difficult to construct a complete geological event development time sequence, which leads to the inability to accurately mark key geological event nodes such as crack extension and soil creep, and further affects the timeliness of the disaster potential degree evaluation. At the same time, the existing evaluation model mainly depends on fixed environmental parameter weights, and does not consider the differences in the influence degree of different regional geological conditions (such as rock and soil properties, terrain slope) on the parameters. When the environmental factors change dynamically, the fixed weight model cannot adjust the evaluation logic in time, and it is easy to cause the evaluation results of the potential degree to be inconsistent with the actual situation. In the early warning threshold setting aspect, the traditional method usually adopts a unified threshold standard, and does not combine the actual needs of different regions such as population distribution and infrastructure type for individualized adjustment, which leads to insufficient pertinence of the early warning information. Some areas may miss the early warning opportunity due to too high threshold, while another part of the area may produce false alarms due to too low threshold, which not only affects the efficiency of disaster prevention decision-making, but also may cause waste of social resources. In addition, the existing system lacks effective model iteration mechanism, and cannot optimize the evaluation model and recognition algorithm according to the actual early warning execution feedback data. With the passage of time, the model accuracy will gradually decrease, and it is difficult to adapt to the needs of long-term monitoring and early warning work. In the disaster response stage, the traditional method mainly relies on artificial experience to formulate the response scheme, and does not quantitatively evaluate the resource consumption of different schemes, which cannot scientifically select the optimal scheme. When multiple hidden danger points are warned at the same time, it is easy to cause unreasonable allocation of resources, which affects the effect of disaster prevention and control. At the same time, the existing system has a lag in processing new monitoring data, and cannot real-time trigger the data reanalysis and evaluation atlas updating process, which leads to the early warning scheme cannot keep up with the latest changes of the geological body, further reducing the reliability of the monitoring and early warning. SUMMARY

[0003] The point landslide debris flow potential degree adaptive monitoring and early warning method aims to solve the problems in the background art.

[0004] To achieve the above-mentioned purpose, the point landslide debris flow potential degree adaptive monitoring and early warning method comprises the following steps: An array of dot array monitoring devices is arranged to collect displacement data and environmental parameters of multiple monitoring points; The timestamp information of the displacement data is analyzed by a time axis analysis module to generate a dynamic time axis and mark geological event nodes on the dynamic time axis; Based on a geographic information system and the dynamic time axis, the displacement data is subjected to spatial mapping processing to generate a preliminary potential degree evaluation result in combination with environmental parameters; The displacement speed and deformation characteristics in the preliminary potential degree evaluation result are classified by a pattern recognition engine to generate a deformation characteristic map with spatiotemporal labels; Based on the deformation characteristic map, a dynamic potential degree evaluation map is calculated by a multi-factor weighted evaluation model; According to user interaction instructions, the early warning threshold parameters are dynamically adjusted to generate a personalized early warning rule set; The dynamic potential degree evaluation map is subjected to real-time scanning processing by the personalized early warning rule set to output a multi-level early warning response scheme; Based on early warning execution feedback data, the pattern recognition engine and the multi-factor weighted evaluation model are subjected to iterative update processing; An integrated collaborative optimization algorithm is used to calculate resource consumption evaluation values under different early warning response schemes and select an optimal response scheme; According to new monitoring data, the time axis analysis module is triggered to re-analyze the flow process, and the dynamic potential degree evaluation map and the multi-level early warning response scheme are updated.

[0005] Preferably, the timestamp information of the displacement data is analyzed by a time axis analysis module to generate a dynamic time axis and mark geological event nodes on the dynamic time axis, which comprises the following steps: For each displacement data sequence of a monitoring point, a sub-time axis is independently constructed and time distribution characteristics are extracted; According to a historical geological disaster database, rainfall duration and intensity thresholds are matched to mark potential trigger event nodes on the sub-time axis; All sub-time axes of the monitoring points are fused to generate a global dynamic time axis and mark a set of associated geological events; The set of associated geological events is input to the spatial mapping processing flow.

[0006] Preferably, based on the geographic information system and the dynamic time axis, the displacement data is processed for spatial mapping, combined with environmental parameters to generate a preliminary potential degree evaluation result, including: mapping the displacement data in the dynamic time axis to the spatial coordinate nodes of the three-dimensional geological model; extracting slope factors, rock-soil permeability coefficients, and groundwater level variables to construct a spatial weight matrix; using the spatial weight matrix to perform environmental correction processing on the displacement mapping result to generate a spatial distribution map with geological feature labels; superimposing the spatial distribution map with real-time rainfall intensity data to output the preliminary potential degree evaluation result.

[0007] Preferably, the displacement velocity and deformation characteristics in the preliminary potential degree evaluation result are classified using a pattern recognition engine to generate a deformation characteristic map with spatiotemporal labels, including: establishing displacement velocity section classification rule libraries and deformation mode classification rule libraries; using an incremental learning mechanism to update the displacement velocity section classification rule libraries in real time; matching the displacement acceleration trend in the preliminary potential degree evaluation result with typical instability modes in the deformation mode classification rule libraries; spatiotemporally correlating the matching result with geological event nodes of corresponding monitoring points to generate the deformation characteristic map with spatiotemporal labels.

[0008] Preferably, based on the deformation characteristic map, a dynamic potential degree evaluation map is calculated through a multi-factor weighted evaluation model, including: extracting displacement velocity section labels and deformation mode labels in the deformation characteristic map as core evaluation factors; fusing soil water content saturation and rock fissure development degree as auxiliary evaluation factors; assigning dynamic weight coefficients to the core evaluation factors and auxiliary evaluation factors according to a historical disaster case library; using the dynamic weight coefficients for weighted fusion calculation to generate the dynamic potential degree evaluation map.

[0009] Preferably, the warning threshold parameters are dynamically adjusted according to user interaction instructions to generate a personalized warning rule set, including: receiving user-set risk level division instructions through a visual operation interface; using semantic parsing technology to perform intent conversion processing on unstructured adjustment requirements; based on the distribution characteristics of the dynamic potential degree evaluation map, intelligently recommending a threshold parameter interval; The threshold parameters confirmed by the user are integrated to generate the personalized early warning rule set.

[0010] Preferably, the dynamic potentiality evaluation graph is scanned in real time using the personalized early warning rule set, and a multi-level early warning response scheme is output, including: The dynamic potentiality evaluation graph is divided into an emergency response area, a key attention area, and a safety monitoring area; The evacuation path planning and monitoring equipment reinforcement scheme is started for the emergency response area; The encrypted monitoring frequency scheme and engineering reinforcement suggestions are generated for the key attention area; All regional response measures are integrated to generate the multi-level early warning response scheme.

[0011] Preferably, the pattern recognition engine and the multi-factor weighted evaluation model are iteratively updated based on early warning execution feedback data, including: The spatial coincidence degree data of the actual disaster occurrence position and the emergency response area are collected; The early warning response delay duration and resource scheduling error value are calculated; The matching weight of the deformation pattern classification rule library is corrected according to the spatial coincidence degree data and the resource scheduling error value; The dynamic weight coefficient distribution strategy in the multi-factor weighted evaluation model is adjusted in linkage.

[0012] Preferably, a cooperative optimization algorithm is integrated, resource consumption evaluation values under different early warning response schemes are calculated, and an optimal response scheme is selected, including: The personnel transfer time cost and equipment scheduling cost in the evacuation path planning scheme are quantified; The material consumption and construction period of the engineering reinforcement suggestion scheme are evaluated; The resource consumption evaluation values are generated through multi-dimensional cost superposition calculation; The response scheme with the smallest resource consumption evaluation value is selected as the optimal response scheme.

[0013] Preferably, the time axis analysis module is triggered to re-analyze in real time according to the new monitoring data, and the dynamic potentiality evaluation graph and the multi-level early warning response scheme are updated in linkage, including: When the monitoring point displacement mutation value exceeds the preset threshold value, the incremental analysis mechanism of the time axis analysis module is automatically triggered; The new geological event node is synchronously updated to the global dynamic time axis; The spatial mapping processing flow and the multi-factor weighted evaluation model calculation flow are re-executed; The multi-level early warning response scheme is regenerated based on the updated dynamic potentiality evaluation graph.

[0014] Compared with the prior art, the present application has the beneficial effects that: By setting up the dot matrix monitoring device array, the limitations of traditional single-point monitoring are changed, the displacement data and environmental parameters of multiple monitoring points can be synchronously collected, the all-around perception of the geological body changes in the monitoring area is realized, the staff can simultaneously master the geological dynamics at different positions, the spatial correlation between the monitoring points is clearly identified, and the judgment deviation caused by the local data loss is avoided. The application of the time axis analysis module can systematically analyze the timestamp information of the displacement data, construct a complete dynamic time axis and accurately label the geological event nodes, so that the development process of the geological event is presented in the form of time sequence, the time nodes of key changes such as crack expansion and soil creep are tracked, and the dynamic process of disaster development is clearly displayed, thereby providing a time dimension basis with more timeliness for potential degree assessment. Combined with the spatial mapping processing of the geographic information system, the displacement data and the geographic spatial position are accurately associated, and the environmental parameters are integrated to generate preliminary assessment results, so that the assessment process considers both the time dynamic change and the spatial distribution characteristics, and the preliminary assessment results are more in line with the actual geological conditions. The mode recognition engine classifies and processes the displacement speed and deformation characteristics in the preliminary assessment results to generate a deformation characteristic map with time and space labels, which can quickly identify different types of deformation modes, distinguish normal geological changes from abnormal deformation, reduce the subjectivity and errors of manual recognition, and make the analysis of deformation characteristics more objective and systematic. The multi-factor weighted assessment model calculates a dynamic potential degree assessment graph based on the deformation characteristic graph, breaks through the limitations of traditional fixed weight models, dynamically adjusts the weights of various factors according to different regional geological conditions, adapts to the geological differences of different monitoring areas, makes the potential degree assessment results more in line with the actual situation, and improves the assessment accuracy. According to the user interaction instruction, the warning threshold parameters are dynamically adjusted and the personalized warning rule set is generated, which can meet the actual needs of different regions, set different warning thresholds for different scenarios such as densely populated areas and the periphery of important infrastructure, avoid the problem of inaccurate warning caused by uniform threshold, make the warning information more targeted, and help relevant departments to develop more practical disaster prevention strategies according to their own needs. The dynamic potential degree assessment graph is scanned in real time using the personalized warning rule set, and a multi-level warning response scheme is output, which can quickly match the corresponding measures according to the potential degree level, make the warning response more hierarchical and operable, and facilitate the staff to take different intensity of prevention and control actions according to the warning level. The iterative updating of the pattern recognition engine and the multi-factor weighted evaluation model based on the feedback data of early warning execution can keep the model and the engine in synchronization with the actual monitoring situation at all times, continuously optimize the identification algorithm and the evaluation logic as the monitoring data accumulates and the feedback information is supplemented, avoid the decline of the model accuracy over time, and ensure the reliability of long-term monitoring and early warning work. The integrated cooperative optimization algorithm calculates the resource consumption evaluation values of different early warning response schemes and selects the optimal scheme, which can quantitatively analyze the input of manpower, materials and other resources, select the scheme with reasonable resource consumption and good prevention and control effect from multiple response schemes, avoid the problem of unreasonable resource allocation, and improve the utilization efficiency of disaster prevention and control resources. According to the real-time triggering of the timeline analysis module based on the newly added monitoring data, the dynamic potential degree evaluation atlas and the multi-level early warning response scheme are updated, which can realize the real-time linkage of monitoring data and evaluation, early warning scheme, ensure that the evaluation result and the early warning scheme are always based on the latest geological dynamic data, timely follow up the change of the geological body, avoid the problem of untimely early warning caused by lagging data processing, and further improve the reliability and timeliness of monitoring and early warning. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 A working principle diagram of the point landslide debris flow potential degree adaptive monitoring and early warning method is provided. Figure 2 A flowchart of timeline analysis and geological event labeling is provided. Figure 3 A flowchart of pattern recognition and deformation feature atlas generation is provided. Figure 4 A flowchart of personalized early warning rule set generation is provided. Figure 5 A flowchart of iterative updating of the pattern recognition and evaluation model is provided. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0017] Please refer to Figure 1 The present application provides a point landslide debris flow potential degree adaptive monitoring and early warning method, which comprises: In the target area, a dot matrix monitoring device array is laid out, which continuously collects ground displacement data such as GNSS displacement, crack meter data and key environmental parameters at each monitoring point. The raw displacement data collected is sent to the time axis analysis module for processing. This module analyzes the timestamp information in the displacement data, constructs a dynamic time axis reflecting the displacement evolution process of each monitoring point, and labels key geological event nodes on the time axis based on historical disaster data and real-time environmental parameters. Combined with the geographic information system platform, the displacement data with time axis information is mapped to the spatial coordinate nodes of the three-dimensional geological model. At the same time, environmental parameters such as slope, rock-soil permeability coefficient and groundwater level are extracted and fused, and the environmental factors are corrected through a spatial weight matrix to generate a preliminary potential degree assessment result, which reflects the basic risk state of different spatial positions under specific environmental conditions. The preliminary assessment result is input into the pattern recognition engine, which uses pre-set displacement velocity section classification rules and deformation mode classification rules, combined with an incremental learning mechanism, to intelligently identify and classify displacement velocity, acceleration trend and deformation characteristics, and output a deformation feature map with accurate spatio-temporal location and event node labels. Based on this map, the multi-factor weighted assessment model starts to work, which not only considers displacement velocity and deformation mode as two core factors, but also fuses soil water content saturation, rock fracture development degree and other auxiliary factors, and dynamically adjusts the weight coefficients of each factor according to the historical disaster case library, finally calculates and generates a dynamic potential degree assessment map covering the entire monitoring area, which directly shows the real-time risk level of different regions.

[0018] The system provides a user interaction interface, allowing professionals to dynamically adjust warning threshold parameters according to field experience or specific needs through visual operations or semantic instructions, forming a personalized warning rule set. Using this rule set, the system performs real-time scanning and partitioning on the dynamic potential degree assessment map, and automatically generates corresponding multi-level warning response schemes. Feedback data during the warning execution process is collected by the system for iterative updating of the classification rule library of the pattern recognition engine and the weight distribution strategy of the multi-factor weighted assessment model, improving the accuracy of subsequent assessment. In addition, the system integrates a collaborative optimization algorithm to evaluate the resource consumption of different warning response schemes generated, automatically selecting the optimal response scheme with the minimum resource consumption. The entire system has dynamic response capability, and when new monitoring data is triggered, it will automatically start the re-analysis process of the time axis analysis module, and update the dynamic time axis, potential degree assessment map and multi-level warning response scheme in linkage, ensuring the real-time and accuracy of the warning information.

[0019] Example 1: see Figure 2The deployment of a dot-matrix array of monitoring equipment forms the data collection foundation for the entire monitoring and early warning system. Within the target monitoring area, based on a detailed analysis of geological structural characteristics and historical disaster distribution patterns, monitoring equipment is systematically selected at geologically representative locations. These devices primarily include high-precision GNSS displacement monitoring stations, crack meters, inclinometers, and more. Each monitoring point operates independently, continuously generating a three-dimensional displacement data sequence with millisecond-level timestamps, and simultaneously collecting key environmental parameters such as rainfall, soil moisture content, and groundwater levels. Timestamp information serves as the timing reference for the data stream, accurately recording the moment each data point was collected, providing the fundamental basis for subsequently establishing a logical chain of event timing. The raw monitoring data is transmitted back to the data processing center in real time via a low-power wide-area IoT or fiber optic network, awaiting in-depth analysis by the core module.

[0020] The spatial configuration of the equipment array follows a geological risk-based approach, with array density differentiated based on the extent of potential slip surfaces, terrain complexity, and historical deformation intensity. Intensive deployment is implemented in key locations such as the leading edge of known slip zones, the tension zone at the trailing edge, and the lateral boundary shear zone; a moderately sparse deployment is adopted in secondary risk areas; and reference points are established in stable bedrock areas. This gradient deployment strategy ensures effective capture of key deformation precursors such as cracking of locked sections and creep acceleration. The synchronous environmental parameter acquisition system and displacement monitoring are strictly co-located, employing integrated sensors to eliminate temporal and spatial errors. Specifically for rainfall parameters, arrays of rain gauges are deployed at different slope elevations to capture spatially varying rainfall patterns. All data records are associated with unique geographic coordinates, forming the basic index unit for spatial analysis. The data acquisition frequency is adaptively adjustable: the base frequency is used during routine monitoring; when the displacement rate at a single point exceeds a specific multiple of the baseline value or when the accumulated rainfall in the area reaches a preset threshold, the system automatically switches to high-frequency mode; and when a displacement acceleration inflection point occurs, overclocking is triggered. The raw data stream undergoes triple preprocessing of outlier filtering, signal denoising and missing value interpolation to ensure that the data quality input into the core analysis module meets the requirements of geomechanical inversion.

[0021] The core function of the time axis analysis module is to convert discrete displacement observations into a time series model with geodynamic significance, and to construct an independent sub-time axis for each monitoring point. The essence is to extract multi-scale features from the displacement-time series: the sliding window method is used to calculate the displacement rate time series, and the rate mutation points are identified by statistical control chart; the mathematical analysis method is used to detect the displacement acceleration change inflection point; combined with signal processing technology to detect periodic deformation fluctuations. At the same time of feature extraction, the system accesses the historical geological disaster case knowledge base in real time: the base integrates the historical landslide and debris flow event precursor data in the region, and establishes a rainfall trigger threshold model for different rock groups through machine learning. The system matches the real-time rainfall sequence with the dynamic threshold in the knowledge base: when the rainfall intensity of a certain monitoring point in the rock group exceeds its specific threshold or the cumulative rainfall breaks the set value, it is marked as a specific type of node in the corresponding time period of the sub-time axis; when the displacement acceleration continuously exceeds the baseline value by a certain standard deviation, it is marked as a deformation acceleration node. All sub-time axes are fused through spatio-temporal correlation analysis: a spatial topological network between monitoring points is constructed, and the spatio-temporal synchronicity of adjacent point event nodes is detected based on the network structure; the clustering algorithm is used to identify node clusters with spatio-temporal continuity. The generated global dynamic time axis contains three element layers: the basic time scale line, the geological event node layer, and the associated event cluster layer, which completely records the time evolution law of regional geological activity.

[0022] The spatial mapping process relies on the three-dimensional geological model platform to integrate engineering geological maps, drilling data, and geophysical profiles to construct a refined three-dimensional grid. Displacement data is mapped to grid nodes through spatio-temporal interpolation to generate a displacement field. Environmental parameters are processed in layers: slope factors are calculated from DEM to distribute slope directions, permeability coefficients are assigned to matrices according to geological units, and groundwater levels are interpolated to generate a surface from monitoring wells. A spatial weight matrix is constructed to integrate slope, permeability coefficient, and groundwater level amplitude, each assigned a weight coefficient based on historical disaster inversion. The slope factor reflects the steepness of the terrain, the permeability coefficient represents the infiltration capacity of the rock-soil body, and the groundwater level amplitude affects the pore water pressure. After standardization, the parameters are weighted and fused to form a matrix, which is used to correct the displacement field: displacement risk values are strengthened in high-permeability steep slope areas and weakened in low-permeability gentle slope areas.

[0023] The corrected results output a raster layer with three-level labels, which is overlaid with real-time rainfall raster for analysis: strong rain areas automatically upgrade the risk level, and when strong rain areas overlap with high displacement and high sensitivity areas, they are marked as the highest risk warning unit. The final output is a displacement correction raster, a risk classification vector, and a warning distribution Figure Three The spatial data product provides structured input of spatio-temporal geological features for subsequent analysis.

[0024] Example 2: see Figure 3The preliminary potentiality assessment results provide an overview of the spatial distribution of risk levels at different locations within the region, but the details of displacement changes and evolution patterns contained therein require more in-depth interpretation to accurately determine the likelihood and type of disaster occurrence. The pattern recognition engine plays a core role in this stage, and its task is to identify displacement feature patterns of warning significance from the preliminary assessment results. The engine relies on two pre-constructed and continuously maintained knowledge bases: a displacement velocity segment classification rule base and a deformation pattern classification rule base.

[0025] The displacement velocity segment classification rule base defines the risk state levels corresponding to different displacement velocity ranges, and the deformation pattern classification rule base is more complex, defining typical slope instability or failure patterns based on displacement acceleration trend and spatial deformation characteristics. These rule bases are not fixed, and the system uses an incremental learning mechanism, especially for real-time updating of the displacement velocity segment classification rule base. This means that as monitoring data continues to flow in, the system can automatically learn new displacement velocity characteristics or distribution patterns, dynamically adjusting or supplementing the original velocity interval division criteria to better fit the actual deformation behavior of the current monitoring area. The pattern recognition engine focuses on the displacement velocity data of each assessment unit when processing the preliminary potentiality assessment results.

[0026] The engine determines which predefined interval the unit's current displacement velocity belongs to according to the current version of the displacement velocity segment classification rule base, and labels it with the corresponding velocity segment tag. The engine deeply analyzes the displacement acceleration trend of the unit, observing whether its velocity is steadily increasing, accelerating and then stabilizing, or showing fluctuations. At the same time, the engine examines whether the deformation of the unit is related to the deformation of its spatially adjacent units, combining the acceleration trend and spatial correlation characteristics, and searching for the most typical instability pattern in the deformation pattern classification rule base that matches the current deformation characteristics. After classification, the engine accurately binds the recognition results, i.e., the velocity segment tag and the deformation pattern tag of the unit, with the geological event nodes associated with it on the global dynamic timeline generated previously. This binding records the temporal association and spatial location.

[0027] The output of the engine is a deformation feature map covering the entire monitoring area, clearly marking the time point, specific location, current displacement velocity state, and typical slope deformation or potential instability pattern. This map provides key structured information for understanding the incubation mechanism, evolution stage, and spatial distribution of geological disasters in the region, and is the basis for more accurate comprehensive risk assessment.

[0028] While deformation signature maps provide core information about deformation states and patterns, forming a final dynamic potential assessment map requires comprehensive consideration of other influencing factors and their quantitative integration. A multi-factor weighted assessment model undertakes this task. This model extracts two factors from the deformation signature map that most directly reflect the current deformation risk as core assessment factors: displacement velocity segment labels and deformation pattern labels. Velocity labels directly indicate the severity of deformation, while pattern labels reveal potential failure mechanisms and the likely type of disaster. Together, these factors are crucial for assessing disaster potential. However, geological hazards often result from the interaction of multiple factors, and deformation signatures alone are insufficient for a comprehensive risk assessment. Therefore, the model incorporates auxiliary assessment factors to enhance the comprehensiveness and accuracy of the assessment. These auxiliary factors typically include soil moisture saturation, which reflects the degree of saturation of the rock mass after rainfall infiltration and directly affects its shear strength. Higher saturation generally results in a greater decrease in strength. Another common auxiliary factor is the degree of fracture development in the rock formation, which reflects the degree of fragmentation and integrity of the rock mass. A higher degree of fracture development indicates the presence of more potential sliding surfaces or seepage pathways. The data for these auxiliary factors comes from environmental monitoring sensors, geological survey reports, remote sensing image interpretation or geophysical exploration results. The historical disaster case database is a valuable knowledge base that stores various monitoring parameters and geological environment data recorded before the occurrence of real disaster events such as landslides and mudslides in the past, including displacement velocity, deformation characteristics, soil moisture content, crack development, etc. By analyzing these historical cases, the model uses statistical or machine learning methods to learn the correlation between different factor combinations and their relative importance and the actual probability of disaster occurrence. Based on this law learned from historical experience, the model dynamically assigns weight coefficients to each core evaluation factor and auxiliary evaluation factor in the current evaluation. The weight coefficient is not static. It will be adjusted according to the specific factor combination of the current evaluated unit and with reference to cases in similar situations in the historical database. The model uses these dynamically determined weight coefficients to perform weighted fusion calculations on all evaluation factors. This process can be expressed by a mathematical formula: , Where: symbol P represents the calculated comprehensive potential value; symbol Represents the quantitative value of the displacement velocity segment factor; symbol represent Corresponding dynamic weight coefficient; symbol Represents the quantitative value of the deformation mode factor; symbol represent Corresponding dynamic weight coefficient; symbol Represents the quantitative value of the soil moisture saturation factor; symbol represent Corresponding dynamic weight coefficient; symbol quantitative value representing the fracture development degree of the rock stratum; symbol representing corresponding dynamic weight coefficient. The result P of the weighted calculation is a quantitative value representing the comprehensive potential degree of the current geological disaster in the evaluation unit. By applying this calculation process to each evaluation unit in the monitoring area and mapping the potential degree values obtained to the corresponding spatial positions, a dynamic potential degree evaluation map covering the entire area is generated. This map usually visually displays the real-time comprehensive risk level of each position using different colors or levels. This map is the most direct and comprehensive basis for early warning decision-making, integrating real-time deformation characteristics, key environmental state information, and valuable historical experience and lessons. Its evaluation results are more refined and reliable than preliminary evaluations.

[0029] The generation of the dynamic potential degree evaluation map marks the completion of the transformation from raw monitoring data to comprehensive risk assessment, condensing complex displacement changes, environmental conditions, and historical experience into an intuitive spatial risk distribution map. The color or level of each point or regional unit on the map represents the possibility of a landslide or debris flow occurring at that location at the current time, as calculated by the multi-factor dynamic weighting process. The generation of the map relies on continuous data input and model calculation. The accurate classification of deformation characteristics by the pattern recognition engine ensures that the core evaluation factors truly reflect the deformation state and potential instability mechanism of the slope.

[0030] The dynamic weight allocation of the multi-factor weighting evaluation model is crucial, as it allows the evaluation results to incorporate historical disaster experience and flexibly adjust the contribution of each factor based on current specific conditions. The introduction of auxiliary evaluation factors compensates for the shortcomings of relying solely on displacement information for risk assessment. Soil moisture saturation is directly related to the degree of rock-soil strength decay, especially under rainfall conditions, where high saturation is often an important factor in inducing sliding. The fracture development degree of the rock stratum reflects the structural integrity of the rock mass, and areas with developed fractures are more likely to form sliding surfaces or experience collapse. By integrating all this information through a quantitative model, a comprehensive potential degree value is ultimately output, allowing risk levels in different regions to be compared and providing a unified, objective quantitative standard for subsequent early warning classification and response decision-making. The dynamic updating feature ensures that the map can reflect the latest risk changes. When new monitoring data shows accelerated displacement or deteriorating environmental conditions, the newly generated map will immediately reflect these changes. This map is the core link connecting risk monitoring and early warning action, and its accuracy and real-time nature directly determine the effectiveness of subsequent early warning response schemes.

[0031] Example 3: see Figure 4, the dynamic potential degree evaluation atlas provides a comprehensive quantitative distribution of regional risk, and the system realizes user interaction through a visual operation interface, which is based on a geographic information system platform and visually displays the atlas in the form of a map. Users can directly set or modify the potential degree threshold parameters corresponding to different risk levels according to the geological background, historical experience and emergency resource status of the monitoring area. The system supports user input of numerical threshold parameters and can process unstructured adjustment instructions in natural language form. Semantic parsing technology is used to analyze the structure and keywords of user input sentences, understand the core intent and convert it into an executable threshold parameter adjustment scheme.

[0032] The system also has an intelligent recommendation function, which analyzes the spatial distribution characteristics of the current atlas and recommends a reasonable threshold parameter interval range to the user based on the built-in algorithm. Users can refer to the recommended values for confirmation or fine-tuning to ultimately form a personalized early warning rule set for the current evaluation period.

[0033] The personalized early warning rule set is the direct basis for the system to conduct early warning zoning, and the system uses the rule set to scan and process the dynamic potential degree evaluation atlas, classifying the potential degree values of each location into pre-set risk levels. It is typically divided into three areas: emergency response area, key attention area and safety monitoring area. For different level areas, the system automatically generates differentiated response schemes. For the emergency response area, the evacuation path planning scheme is started, based on the geographic information system network analysis function, the latest road network data, population distribution data and shelter information are called. The optimal path is calculated using a path optimization model, which is represented as: , where symbol T represents the estimated total evacuation time, symbol N represents the total number of evacuation paths to be planned, symbol i represents the number of a specific evacuation path, symbol represents the length of the evacuation path numbered i, i.e. the actual road distance from the evacuation starting point to the shelter, symbol represents the average travel speed of vehicles or personnel on path numbered i, symbol represents the road condition influence coefficient of path numbered i, which reflects the road capacity, and symbol represents the weather influence coefficient. At the same time, the monitoring device reinforcement scheme is generated, including instructions to move monitoring devices to strengthen deployment or unmanned aerial vehicle emergency patrol. For the key attention area, the encrypted monitoring frequency scheme is developed to increase the data collection and reporting frequency and closely track the deformation trend. According to the geological characteristics, topographic conditions and main contributing factors of potential degree, engineering reinforcement suggestions are generated, including slope toe counterpressure, slope waterproofing, drainage dredging and other measures.

[0034] The system integrates regional response measures to form a complete multi-level early warning response plan, clearly defining space, time, and operational deployment, providing a guide for emergency command. The plan output contains multi-dimensional information, with spatial partition information as the core, clearly indicating the specific geographic range of each response area. For emergency response areas, detailed evacuation route descriptions, estimated times, personnel lists, and responsible parties are listed; monitoring equipment deployment instructions are included. For key focus areas, encrypted monitoring instructions and engineering reinforcement suggestions are included, with clear execution priorities and time requirements. The plan also summarizes resource demand information in a standardized format to ensure completeness and executability. The plan is based on objective risk classification and individualized rule sets, combined with geographic information system analysis and pre-set measure knowledge base, to ensure scientificity and pertinence, completing the full-process automated processing from risk monitoring to response decision-making.

[0035] Example 4: Referring to Figure 5 The actual execution process of the early warning response plan generates a large amount of feedback information, which is an important basis for evaluating the accuracy of the system's early warning and the effectiveness of the response, and is also a key data source for driving the system's self-optimization and upgrading. The system actively collects multi-dimensional feedback data during the early warning execution process. One type of core feedback data is actual disaster event information, including the precise time of disaster occurrence, specific geographic coordinates, disaster impact range, and disaster scale grade. The system performs spatial overlay analysis on the location coordinates of actual disaster occurrence and the geographic range of the emergency response area previously defined based on the dynamic potential degree evaluation map and individualized early warning rule set. Through geographic information system tools, the spatial coincidence degree data between the two is calculated, with high coincidence degree indicating accurate regional delineation of the system's early warning, and low coincidence degree indicating the existence of false negatives or false positives.

[0036] Another important feedback data is timeliness data, which records the time delay from the system's formal issuance of early warning instructions to the actual start and completion of key response measures. This includes the time for early warning information to reach various response departments, on-site command decision-making time, and the execution time of specific actions. At the same time, the system collects error information that occurs during resource scheduling, which reflects the efficiency of the response process and the accuracy of resource allocation. In addition, the system also collects the execution effect data of early warning response measures, as shown in Table 1.

[0037] Table 1: Feedback data type and purpose of early warning execution.

[0038]

[0039] The early warning execution feedback data is structured and stored in a special database to support model iteration and update, and the spatial coincidence degree data is used to evaluate the accuracy of the early warning area delineation: when the actual disaster point and the emergency response area have a low coincidence degree below the preset threshold, the system traces back to the deformation feature map before the disaster occurs, and analyzes the weight settings of the displacement velocity section label and the deformation mode label. If it is found that the high-risk deformation mode has a high disaster risk in historical cases but the current rule library weight is low, the weight coefficient of the related matching rule in the deformation mode classification rule library is increased.

[0040] The resource scheduling error value is used to optimize the response efficiency model parameters, and the dynamic weight distribution strategy of the multi-factor weighted evaluation model is adjusted synchronously: the system adds new disaster cases to the historical case library, and recalculates the correlation strength between each evaluation factor and the disaster occurrence probability by statistical methods. The weight coefficient of the auxiliary factor that plays a key role in this disaster is increased; the weight of the outstanding factor in multiple false alarms is reduced. This closed-loop optimization mechanism continuously absorbs the latest experience and improves the accuracy of subsequent potential degree evaluation.

[0041] In order to optimize resource allocation and improve response efficiency under the premise of meeting early warning needs, the system integrates a collaborative optimization algorithm for evaluating resource consumption under different early warning response schemes and selecting the optimal scheme. The collaborative optimization algorithm needs to quantitatively evaluate the resource consumption of multiple alternative early warning response schemes generated based on the personalized early warning rule set. The evaluation needs to convert various resources required for scheme execution into calculable cost indicators. For schemes that include evacuation actions, the algorithm quantifies the personnel transfer time cost, which includes the estimated organization mobilization time, personnel assembly time, and total transportation time on the evacuation path. At the same time, the algorithm calculates the equipment scheduling cost, including the estimated number of vehicles needed, vehicle mileage, usage cost of on-board equipment, and hourly cost of command and support personnel involved in evacuation work. For schemes that include engineering reinforcement measures, the algorithm evaluates the material consumption, including the types, quantities, procurement unit prices, and transportation distances of the required materials. At the same time, the algorithm estimates the construction period, including material preparation time, equipment access time, and actual construction time, and converts the time into labor cost and equipment rental cost. For schemes that involve increasing the number of monitoring devices or increasing monitoring, the algorithm evaluates the device transportation cost, device usage and wear cost, and the incremental cost of data transmission and storage due to increased monitoring frequency. The algorithm calculates the total resource consumption evaluation value of the early warning response scheme by superimposing all quantifiable time cost, labor cost, material cost, and equipment usage cost through multi-dimensional cost superposition. This evaluation value reflects the economic cost, time efficiency, and resource demand intensity of the scheme. In calculating the total resource consumption evaluation value, the system uses the following formula for quantification: , wherein: symbol C represents the total resource consumption evaluation value of the early warning response scheme; symbol represents the total time consumption estimated for the evacuation action plan, including personnel organization, assembly and transportation time; symbol represents the cost coefficient per unit time, which is the comprehensive cost converted from human resources, management, delay, etc. per unit time; symbol represents the total construction time consumption estimated for the engineering reinforcement measure plan; symbol M represents the total consumption of various materials required for the engineering reinforcement measures, which is a comprehensive quantity and may be standardized and converted according to different materials; symbol represents the cost coefficient per unit material consumption, which is the procurement, transportation and loss cost per unit of material; symbol E represents the total consumption related to equipment scheduling and use, including vehicle mileage, equipment use time, data transmission volume and other comprehensive conversion quantities; symbol represents the cost coefficient per unit equipment consumption, which is the fuel, rental, maintenance or communication cost corresponding to each unit of equipment consumption. The system calculates and compares the resource consumption evaluation values C of all feasible early warning response plans generated, and selects the plan with the smallest evaluation value as the optimal response plan. This means that under the premise of meeting the early warning response target, this plan consumes the least human, material, financial and time resources, thereby achieving the optimization of resource utilization. The selection process of the optimal response plan relies on accurate cost quantification models and comprehensive cost item coverage to ensure that the selection result has practical guiding significance. The selected optimal plan will be output and executed, and its execution effect will be fed back into the system as new data to drive subsequent continuous optimization.

[0042] Example 5: The geological conditions and deformation state of the monitoring area are dynamically changing, and new monitoring data may reveal rapid evolution of risks or emergence of new risk points. The system has the ability to respond to new data in real time, and its core mechanism is to preset a series of data trigger conditions. A typical trigger condition is when the displacement data of any monitoring point mutates and the mutation value exceeds the preset safety threshold. This safety threshold is not fixed and is usually set based on the statistical characteristics of the historical displacement data of the monitoring point, the stability analysis of the geological structure at that location, and the engineering experience judgment. When calculating this trigger threshold, the system uses the following formula: , wherein: represents the displacement mutation trigger threshold calculated for the monitoring point; symbol represents the moving average of the historical displacement data of the monitoring point. The moving average is a statistical method that calculates the average of displacement data in a recent period of time, reflecting the typical deformation level in the recent period; symbol The standard deviation of the historical displacement data of the monitoring point, which measures the fluctuation degree of the historical displacement value around the average value, the greater the fluctuation, the greater the standard deviation; the symbol k represents the statistical confidence coefficient, which is a magnification factor set according to engineering experience and risk tolerance, usually taking a value between 2 and 3, used to increase a certain safety margin on the basis of the average value; the symbol represents the geological sensitivity coefficient, which is determined by the geological conditions of the location, the symbol represents the slope correction factor, which is determined according to the slope value of the location where the monitoring point is located, the steeper the slope, the greater the factor value, reflecting the enhancement of gravity effect. This formula comprehensively considers the historical deformation law of the monitoring point itself, the inherent sensitivity of the geological environment and the influence of the topographic conditions, and dynamically calculates the scientific safety threshold suitable for the point.

[0043] The core of the incremental analysis mechanism is efficiency, which does not reprocess all historical data, but focuses on the newly added data segment, especially the trigger point and the data of the associated period. The mechanism updates the independent sub-time axis of the trigger point. The update process includes adding the newly added displacement data point to the displacement data sequence of the point, analyzing the time distribution characteristics of the new data point, and recalculating the displacement rate and acceleration change trend of the point. At the same time, the system will check whether the newly added displacement is accompanied by a new environmental trigger event, especially a rainfall event. It will match the latest rainfall data of the area where the trigger point is located with the critical threshold in the historical geological disaster database. If it is found that there is rainfall exceeding the critical threshold during the newly added displacement, the system will mark a new "potential trigger event node on the updated sub-time axis of the monitoring point, and the system will update this newly added geological event node to the global dynamic time axis. The global dynamic time axis is the fusion view of all monitoring point sub-time axes, which records the timing context of the entire regional geological activity. The addition of the new event node means that the global time axis has incorporated the latest geological activity information, reflecting the latest timing development dynamics.

[0044] After the global dynamic time axis is updated, the system immediately starts the spatial mapping processing flow, and uses the updated time axis data and the latest displacement information of all monitoring points to remap the displacement data to the spatial coordinate nodes of the three-dimensional geological model. The model is constructed based on geological exploration data and accurately reflects the topographic relief and stratigraphic structure characteristics. At the same time, the latest environmental parameters are extracted to construct an updated spatial weight matrix. The matrix quantifies the correction strength of different location environmental conditions on disaster risk, for example, a high-permeability steep slope area is given a higher risk weight. Using the newly constructed spatial weight matrix, the system performs environmental correction processing on the displacement data to generate updated preliminary potential degree assessment results. The results integrate the new displacement data and environmental change information, reflecting the latest risk state. The updated assessment results are immediately input into the pattern recognition engine, which analyzes the new displacement characteristics based on the current rule base: identifies changes in velocity section, deformation mode transition, and spatial correlation evolution. The analysis results are spatiotemporally bound with the new geological event nodes to output an updated deformation characteristic map reflecting the regional deformation characteristic changes caused by the new data.

[0045] Based on the new deformation characteristic map, the multi-factor weighted assessment model recalculates: the latest velocity section label and deformation mode label are extracted as core factors, combined with post-mutation soil moisture content, fracture development degree and other auxiliary factors, and weighted and fused according to the current weight strategy. The calculation result generates an updated dynamic potential degree assessment map to display the real-time risk distribution changes of the region in color classification. The system immediately scans the new map using the original early warning rule set, redraws the response area according to the threshold, and generates the corresponding response scheme: a new emergency area starts evacuation path planning and equipment reinforcement; an upgraded attention area implements intensive monitoring and engineering reinforcement; a risk downgraded area reduces the response level. This closed-loop process realizes minute-level response, ensuring that the system dynamically captures sudden danger and optimizes the early warning strategy in real time, significantly improving the timeliness and accuracy of early warning. The system continues to run, and new triggered data will activate the full process update again.

[0046] It should be noted that, in this text, relationship terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device.

[0047] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. A method for adaptive monitoring and early warning of point-like landslide and debris flow potential, characterized in that: include: Set up a dot matrix monitoring device array to collect displacement data and environmental parameters at multiple monitoring points; The time stamp information of the displacement data is parsed by a time axis parsing module to generate a dynamic time axis and mark geological event nodes on the dynamic time axis; Based on the geographic information system and the dynamic time axis, the displacement data is spatially mapped and processed, and a preliminary potential assessment result is generated in combination with environmental parameters; Using a pattern recognition engine to classify the displacement velocity and deformation characteristics in the preliminary potential assessment results to generate a deformation feature map with spatiotemporal labels; Based on the deformation characteristic map, a dynamic potential evaluation map is calculated using a multi-factor weighted evaluation model; Dynamically adjust warning threshold parameters based on user interaction instructions to generate personalized warning rule sets; Using the personalized warning rule set to scan and process the dynamic potential assessment map in real time, and output a multi-level warning response plan; Iteratively updating the pattern recognition engine and the multi-factor weighted evaluation model based on the early warning execution feedback data; Integrate collaborative optimization algorithms to calculate resource consumption assessment values ​​under different early warning response plans and select the optimal response plan; The re-analysis process of the timeline analysis module is triggered in real time according to the newly added monitoring data, and the dynamic potential assessment map and the multi-level early warning response plan are updated in a linked manner.

2. The method according to claim 1, characterized in that The time stamp information of the displacement data is parsed and processed by a time axis parsing module to generate a dynamic time axis and mark geological event nodes on the dynamic time axis, including: For the displacement data sequence of each monitoring point, a sub-time axis is independently constructed and the time distribution characteristics are extracted; Matching rainfall duration and intensity thresholds based on a historical geological disaster database, and marking potential triggering event nodes on the sub-timeline; The sub-time axes of all monitoring points are integrated to generate a global dynamic time axis and annotate clusters of related geological events; The clusters of correlated geological events are input into the spatial mapping process.

3. The method according to claim 2, characterized in that Based on the geographic information system and the dynamic time axis, the displacement data is spatially mapped and processed, and a preliminary potential assessment result is generated in combination with environmental parameters, including: Mapping the displacement data in the dynamic time axis to the spatial coordinate nodes of the three-dimensional geological model; Extract slope factor, rock and soil permeability coefficient and groundwater level variables to construct spatial weight matrix; Performing environmental correction processing on the displacement mapping results using the spatial weight matrix to generate a spatial distribution map with geological feature labels; The spatial distribution map is overlaid and analyzed with real-time rainfall intensity data to output the preliminary potential assessment result.

4. The method according to claim 3, characterized in that The displacement velocity and deformation characteristics in the preliminary potential assessment results are classified using a pattern recognition engine to generate a deformation feature map with spatiotemporal labels, including: Establish a displacement velocity segment classification rule base and a deformation mode classification rule base; Adopting an incremental learning mechanism to update the displacement velocity segment classification rule base in real time; Matching typical instability modes in the deformation mode classification rule library according to the displacement acceleration change trend in the preliminary potential assessment result; The matching results are temporally and spatially associated with the geological event nodes of the corresponding monitoring points to generate the deformation feature map with temporal and spatial labels.

5. The method according to claim 4, characterized in that Based on the deformation characteristic map, a dynamic potential evaluation map is calculated using a multi-factor weighted evaluation model, including: Extracting the displacement velocity segment label and the deformation mode label in the deformation feature map as core evaluation factors; Integrate soil moisture saturation and rock fracture development as auxiliary evaluation factors; Allocating dynamic weight coefficients to the core assessment factors and auxiliary assessment factors based on a historical disaster case database; The dynamic weight coefficient is used to perform weighted fusion calculation to generate the dynamic potential evaluation map.

6. The method according to claim 5, characterized in that Dynamically adjust warning threshold parameters based on user interaction instructions to generate a personalized warning rule set, including: Receive risk level classification instructions set by the user through a visual operation interface; Use semantic parsing technology to convert unstructured adjustment requirements into intent; Intelligently recommending threshold parameter intervals based on the distribution characteristics of the dynamic potential evaluation map; The threshold parameters confirmed by the user are integrated to generate the personalized warning rule set.

7. The method according to claim 6, characterized in that The personalized warning rule set is used to perform real-time scanning and processing on the dynamic potential assessment map, and a multi-level warning response plan is output, including: Dividing the dynamic potential assessment map into an emergency response area, a key focus area, and a safety monitoring area; Initiate evacuation route planning and additional monitoring equipment deployment plans for the emergency response area; Generate an enhanced monitoring frequency plan and engineering reinforcement suggestions for the key areas of concern; Integrate all regional response measures to generate the multi-level early warning response plan.

8. The method according to claim 7, characterized in that Iteratively updating the pattern recognition engine and the multi-factor weighted evaluation model based on the early warning execution feedback data includes: Collecting spatial overlap data between the actual disaster location and the emergency response area; Calculate the warning response delay time and resource scheduling error value; Correcting the matching weight of the deformation pattern classification rule base according to the spatial overlap data and the resource scheduling error value; The dynamic weight coefficient allocation strategy in the multi-factor weighted evaluation model is adjusted in a coordinated manner.

9. The method according to claim 8, characterized in that Integrate collaborative optimization algorithms to calculate resource consumption assessment values ​​under different early warning response plans and select the optimal response plan, including: Quantify the personnel transfer time cost and equipment scheduling cost in the evacuation path planning scheme; Evaluate the material consumption and construction period of the proposed reinforcement scheme; Generate the resource consumption assessment value by multi-dimensional cost superposition calculation; The response plan with the smallest resource consumption evaluation value is selected as the optimal response plan.

10. The method according to claim 9, characterized in that The re-analysis process of the timeline analysis module is triggered in real time according to the newly added monitoring data, and the dynamic potential assessment map and the multi-level early warning response plan are updated in a linked manner, including: When the displacement mutation value of the monitoring point exceeds the preset threshold, the incremental analysis mechanism of the time axis analysis module is automatically triggered; Synchronously updating the newly added geological event nodes to the global dynamic timeline; Re-execute the spatial mapping process and the multi-factor weighted evaluation model calculation process; The multi-level early warning response plan is regenerated based on the updated dynamic potential assessment map.

Images