A method and system for evaluating the risk of slope geological disasters

Abstract

The application provides a kind of side slope geological disaster risk assessment method and system, by obtaining non-natural vibration data, the feature information of side slope object can be accurately evaluated in combination with the hidden danger description factor information of side slope object, then the adjacent side slope object around target side slope object is considered in side slope object set, and the comprehensive analysis of target side slope object is formed through the side slope hidden danger data characteristics in side slope object set, then the description vector of target side slope object set is obtained by extracting the characteristics of important hidden danger of target side slope object set, and whether the side slope object attribute of target side slope object affects the side slope geological disaster risk assessment result is regressed according to the description vector of target side slope object set, the side slope object attribute is used to represent the attribute of side slope object hidden danger, improve the regression analysis accuracy of side slope object attribute change, ensure the accuracy and confidence of evaluation.

Description

Technical Field

[0001] This application relates to the field of slope geological hazard risk assessment technology, and more specifically, to a slope geological hazard risk assessment method and system. Background Technology

[0002] Slopes, as important geomorphic units in engineering construction and the natural geographical environment, are widely distributed along highways and railways, in mining areas, in reservoir areas of water conservancy projects, and on urban building slopes. With the continuous advancement of infrastructure construction, human engineering activities such as slope excavation and load disturbance are becoming increasingly frequent. Coupled with the impact of natural factors such as extreme rainfall and earthquakes, the risk of slope geological disasters (such as landslides, collapses, and debris flows) continues to rise. These disasters are characterized by their suddenness, wide destructive range, and chain reaction of secondary disasters. They can not only cause significant casualties and property losses but also damage important infrastructure such as transportation arteries and water conservancy facilities, leading to ecological degradation and posing a serious threat to the stable economic and social development of the region.

[0003] Therefore, slope geological hazard risk assessment, as a core component of disaster prevention and control, directly impacts the scientific validity of disaster prevention and mitigation decisions in terms of accuracy and timeliness. Traditional slope geological hazard risk assessment methods are mostly based on geological survey data of a single slope (such as physical and mechanical parameters of soil and rock, slope gradient, slope height, groundwater level, etc.), using numerical simulation or empirical formulas for stability analysis and risk assessment. However, these methods have significant limitations: on the one hand, they focus only on the inherent properties of a single slope, neglecting the spatial and geological connections and risk transmission effects between adjacent slopes and slopes in the same area, resulting in a lack of comprehensiveness and systematicity in the assessment results; on the other hand, the data collection dimensions are singular, failing to fully consider the inducing effects of non-natural vibration factors such as blasting construction, traffic vibration, and mechanical operations on slope stability, while such external disturbances are often key triggers for slope instability; furthermore, traditional methods have weak anti-interference capabilities, are sensitive to noise in monitoring data, and lack dynamically updated assessment models, making it difficult to adapt to real-time changes in slope conditions, leading to discrepancies between assessment results and actual risks, and failing to meet the needs of accurate early warning and dynamic prevention and control.

[0004] Meanwhile, existing assessment technologies lack standardized processes for data processing and feature extraction, making it difficult to effectively integrate data from different sources and of different types (such as geological survey data, monitoring data, vibration data, etc.), resulting in incomplete characterization of slope hazards. Furthermore, the model training process does not fully utilize historical data and related information from similar slopes, leading to insufficient generalization ability and difficulty in adapting to complex and changing geological environments and engineering conditions. Therefore, there is an urgent need for a slope geological hazard risk assessment method that can integrate multi-dimensional data, consider the correlation between slopes, has strong anti-interference capabilities, and possesses dynamic adaptive capabilities. This would improve the accuracy, comprehensiveness, and real-time nature of risk assessment, providing reliable technical support for precise early warning and scientific management of slope geological hazards. Summary of the Invention

[0005] To address the technical problems existing in related technologies, this application provides a method and system for assessing geological hazards on slopes.

[0006] Firstly, a method for assessing geological hazard risks on slopes is provided, the method comprising:

[0007] Obtain real-time slope objects similar to the target slope object, and construct a set of slope objects based on the target slope object according to the association between the target slope object and the real-time slope objects;

[0008] The slope objects in the slope object set are identified as slope hazard data, and the hazard description factor information and non-natural vibration data of each slope hazard data are obtained.

[0009] Based on the hazard description factor information and non-natural vibration data of each slope hazard data, the slope hazard data features in the slope object set are generated to obtain the target slope object set;

[0010] The feature extraction of important hidden dangers of the target slope object set is used to obtain the description vector of the target slope object set, and the slope geological disaster risk assessment result is obtained based on the description vector of the target slope object set. The slope object attributes are used to characterize the attributes that affect the hidden dangers of the slope object.

[0011] In this application, the step of generating slope hazard data features in the slope object set based on the hazard descriptor information and non-natural vibration data of each slope hazard data to obtain the target slope object set includes:

[0012] The non-natural vibration data of the slope hazard data is subjected to feature extraction processing, and the hazard point weight features of each segment of the data obtained after feature extraction processing are extracted to obtain the hazard feature weight of the slope hazard data.

[0013] The hazard descriptive factor information of the slope hazard data is extracted to obtain the hazard descriptive factor features, and the hazard feature weights of the slope hazard data and the hazard descriptive factor features are integrated to obtain the slope hazard data features of the slope hazard data;

[0014] The slope object set is optimized based on the slope hazard data characteristics of each slope hazard data point to obtain the target slope object set.

[0015] In this application, the step of performing feature extraction processing on the non-natural vibration data of the slope hazard data, and extracting the hazard point weight features of each segment of the data obtained after feature extraction processing to obtain the hazard feature weight of the slope hazard data, includes:

[0016] The non-natural vibration data of the slope hazard data are divided into different data types to obtain several data scales of non-natural local vibration data.

[0017] Feature extraction processing is performed on non-natural local vibration data of various data scales, and the hidden danger features of each segment in the data obtained after feature extraction processing are extracted to obtain the sub-hidden danger feature weights at different data scales.

[0018] The sub-hazard feature weights of the slope hazard data are obtained by integrating and processing the sub-hazard feature weights of the different data scales.

[0019] In this application, the step of extracting the features of the target slope object set to obtain the description vector of the target slope object set, and determining whether the slope object attributes of the target slope object affect the slope geological hazard risk assessment result based on the description vector of the target slope object set, includes:

[0020] The target number of analyses for the analysis thread is determined based on the association between the target slope object and the real-time slope object.

[0021] The target slope object set is loaded into the analysis thread of the target analysis number to extract the features of important hidden dangers and obtain the anti-noise features of the target slope object set;

[0022] The noise resistance features are loaded into the dimensionality reduction unit for dimensionality reduction processing to obtain the description vector of the target slope object set;

[0023] The description vector of the target slope object set is loaded into the decision unit to obtain the regression analysis vector. Based on the regression analysis vector being switched to the slope object attribute influence probability value of the target slope object, the slope geological hazard risk assessment result is obtained.

[0024] In this application, the step of loading the target slope object set into the analysis thread of the target analysis number to extract the noise resistance features of the target slope object set by extracting features of important hidden dangers includes:

[0025] A feature queue is constructed based on the slope hazard data features of each slope hazard data in the target slope object set, and an adjacency queue is constructed based on the connection relationship between each slope hazard data.

[0026] The feature queue and the adjacency queue are loaded into the first-level analysis thread so that the features of the adjacent slope hazard data of each slope hazard data are integrated through the first-level analysis thread to obtain the optimized slope hazard data features.

[0027] The optimized slope hazard data features and the adjacent queue are loaded into the next-level analysis thread so that the features of the adjacent slope hazard data of each slope hazard data can be integrated through the next-level analysis thread to obtain the optimized slope hazard data features.

[0028] Repeatedly load the optimized slope hazard data features and the adjacency queue into the next layer of analysis thread until the last layer of analysis thread outputs the noise resistance features of the target slope object set.

[0029] In this application, a pre-trained slope object regression analysis thread generates slope hazard data features in the slope object set based on the hazard descriptor information and non-natural vibration data of each slope hazard data, and extracts the features of important hazards from the target slope object set to obtain the description vector of the target slope object set. The slope geological disaster risk assessment result is obtained based on the description vector of the target slope object set.

[0030] The training steps for the slope object regression analysis thread include:

[0031] Obtain the example adjacent slope object that is similar to the example test slope object, and the slope object attribute directory of the example test slope object;

[0032] Based on the association between the example test slope object and the example adjacent slope object, a set of example slope objects is constructed, and the example hidden danger descriptor information and example non-natural vibration data of each example slope object in the set of example slope objects are obtained;

[0033] The example slope object set, the example hazard description factor information of each example slope object, and the example non-natural vibration data are loaded into the training thread to generate the example slope object features of each example slope object, and the example slope geological hazard risk assessment results are obtained based on the example slope object set containing the example slope object features.

[0034] Based on the geological hazard risk assessment results of the example slope and the slope object attribute catalog of the example test slope object, a prediction reliability index is constructed, and the thread coefficients of the thread to be trained are updated according to the prediction reliability index to obtain the slope object regression analysis thread.

[0035] In this application, the set of exemplary slope objects includes several exemplary adjacent slope objects;

[0036] The acquisition of exemplary hazard descriptor information and exemplary non-natural vibration data for each exemplary slope object in the exemplary slope object set includes:

[0037] Obtain the risk level and impact data volume of each adjacent slope object in the example slope object set; clean the target example adjacent slope objects from the example slope object set if the slope object risk level is lower than a preset risk level specified value or the slope object impact data volume is greater than a preset impact data volume specified value, and obtain the target example slope object set.

[0038] Obtain at least one of the following in the target example slope object set: risk level, amount of impact data, scale of interference data, and slope object type for each example slope object. Then, standardize the obtained data to obtain example hazard descriptor information for each example slope object.

[0039] Obtain the cumulative non-natural vibration data of each example slope object in the target example slope object set within a preset period;

[0040] If there are null values ​​in the cumulative unnatural vibration data, the cumulative unnatural vibration data is repaired to obtain the example unnatural vibration data for each example slope object.

[0041] In this application, the step of loading the set of exemplary slope objects, the exemplary hazard descriptor information of each exemplary slope object, and the exemplary non-natural vibration data into the thread to be trained to generate the exemplary slope object features of each exemplary slope object includes:

[0042] The example integrated feature is obtained by integrating the example hazard feature weights corresponding to the example non-natural vibration data and the example hazard descriptive factor features corresponding to the example hazard descriptive factor information through the thread to be trained.

[0043] Obtain the historical changes in the weights of potential hazard points of the example slope object and the real-time geological characteristics of the example;

[0044] Based on the historical changes and geological characteristics of the examples, the integrated features of the example slope objects are weighted to obtain the example slope object features.

[0045] In this application, the step of loading the set of exemplary slope objects, the exemplary hazard descriptor information of each slope object, and the exemplary non-natural vibration data into the thread to be trained to generate exemplary slope object features for each slope object includes:

[0046] The training thread integrates the example hazard feature weights corresponding to the example non-natural vibration data and the example hazard descriptor features corresponding to the example hazard descriptor information in the target direction to obtain example integrated features.

[0047] The example integration features are loaded into the saliency analysis unit of the thread to be trained to obtain the saliency weights corresponding to the example integration features;

[0048] The feature vector of the example is weighted according to the significance weight to obtain the feature of the example slope object.

[0049] In this application, the integration of the example hazard feature weights corresponding to the example non-natural vibration data and the example hazard descriptor factor features corresponding to the example hazard descriptor factor information in the target direction through the training thread to obtain example integrated features includes:

[0050] If the example slope object has unique information within the corresponding period of the example non-natural vibration data, feature extraction is performed on the unique information to obtain the feature;

[0051] The training thread integrates the example hazard feature weights corresponding to the example non-natural vibration data, the example hazard descriptor features corresponding to the example hazard descriptor information, and the event features in the target direction to obtain example integrated features.

[0052] In this application, after determining whether the slope object attributes of the target slope object affect the slope geological hazard risk assessment result based on the description vector of the target slope object set, the process includes:

[0053] If the slope geological hazard risk assessment result indicates that the probability value of the slope object attribute impact of the target slope object is higher than the preset probability value, then the target slope object is sent to the data acquisition device to collect the changed slope object attributes of the target slope object through the data acquisition device;

[0054] The collected changes in slope object attributes are associated with the target slope object and entered into the landslide database.

[0055] Secondly, a slope geological hazard risk assessment system is provided, including a processor and a memory that communicate with each other, wherein the processor is used to read a computer program from the memory and execute it to implement the above-mentioned method.

[0056] The slope geological hazard risk assessment method and system provided in this application embodiment obtains real-time slope objects similar to the target slope object, and constructs a slope object set based on the target slope object according to the correlation between the target slope object and the real-time slope objects; identifies the slope objects in the slope object set as slope hazard data, and obtains hazard descriptive factor information and non-natural vibration data for each slope hazard data; generates slope hazard data features in the slope object set based on the hazard descriptive factor information and non-natural vibration data for each slope hazard data to obtain the target slope object set; and obtains the target slope object set by combining the hazard descriptive factor information of the slope objects with the non-natural vibration data. Sub-information can accurately assess the characteristic information of slope objects, and then consider adjacent slope objects around the target slope object in the slope object set. By using the slope hazard data characteristics in the slope object set, a comprehensive analysis of the target slope object is formed. Then, the characteristic of important hazards is extracted from the target slope object set to obtain the description vector of the target slope object set. Based on the description vector of the target slope object set, regression analysis is performed to determine whether the slope object attributes of the target slope object affect the slope geological hazard risk assessment results. The slope object attributes are used to characterize the attributes that affect the slope object hazards, improve the accuracy of the regression analysis of slope object attribute changes, and ensure the accuracy and confidence of the assessment. Attached Figure Description

[0057] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This is a flowchart of a slope geological hazard risk assessment method provided in an embodiment of this application. Detailed Implementation

[0059] To better understand the above technical solutions, the technical solutions of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of this application and the specific features in the embodiments are detailed descriptions of the technical solutions of this application, rather than limitations on the technical solutions of this application. In the absence of conflict, the embodiments of this application and the technical features in the embodiments can be combined with each other.

[0060] Please see Figure 1 This paper presents a method for assessing geological hazards on slopes, which may include the technical solutions described in steps S210-S240.

[0061] S210. Obtain real-time slope objects similar to the target slope object, and construct a set of slope objects based on the target slope object according to the relationship between the target slope object and the real-time slope objects.

[0062] For example, the target slope object: In technical scenarios such as slope monitoring, analysis and early warning, the slope entity object selected as the benchmark reference is the core basis for subsequent slope object selection, assembly and analysis decisions.

[0063] Core attributes: It has key characteristics such as clear spatial location, geological conditions (rock / soil, structural features), engineering conditions (support method, construction stage), monitoring parameters (displacement, stress, moisture content, rainfall, etc.) and risk level, which can be determined through historical data, on-site investigation or preset threads.

[0064] Real-time slope objects: Slope entities that reflect the current instantaneous state of the slope, acquired through real-time acquisition methods (such as slope monitoring sensors, remote sensing images, drone inspections, and real-time on-site surveys).

[0065] Key features: The data is timely (dynamically updated), containing real-time characteristics such as the current geometry of the slope, deformation, stress changes, and environmental parameters (rainfall, temperature), which distinguishes it from static historical slope objects or benchmark target slope objects.

[0066] Similar real-time slope objects: Real-time slope objects that are highly similar to the target slope object in core feature dimensions, based on preset similarity judgment rules (such as feature matching algorithms, similarity thresholds, and key indicator comparisons).

[0067] Core similarity dimensions include, but are not limited to, geological type (same rock / soil slope), slope morphology (slope, height, slope type), structural characteristics (joints, fissure distribution), monitoring indicators (deformation rate, stress level), engineering environment (same region, same rainfall conditions, same support scheme), instability mode (potential landslide / collapse type), etc.; similarity must meet the preset quantitative threshold (e.g., similarity ≥ 80%) or qualitative matching conditions.

[0068] Relationships: The logical or physical connection between the target slope object and similar real-time slope objects is the core basis for constructing a slope object set (obtaining real-time slope objects similar to the target slope object, and constructing a slope object set based on the target slope object based on the spatial, geological, monitoring, engineering, or feature relationships between the two. Similar real-time slope objects must meet the requirement of a core feature dimension similarity ≥ 80% (such as consistent geological type, slope morphology, monitoring indicators, etc.), and may include first-order, second-order, or third-order adjacent slope objects of the target slope). It is categorized into the following types:

[0069] Spatial relationships: such as adjacent slopes, slopes in the same area, and subordinate slopes (main slope - secondary slope).

[0070] Geological correlation: slopes belonging to the same geological unit, tectonic zone, and rock and soil type;

[0071] Monitoring correlation: slopes covered by the same monitoring network, the same monitoring indicator system, and the same monitoring equipment;

[0072] Project related: slopes in the same project section, construction stage, and with the same support / reinforcement scheme;

[0073] Feature correlation: such as similar deformation patterns, similar risk evolution trends, and slopes with the same instability causes.

[0074] The set of slope objects based on the target slope object: The slope object group is formed by integrating similar real-time slope objects selected based on the target slope object and in accordance with the above-mentioned association rules.

[0075] Key features: The set uses the target slope object as a reference benchmark, and all member objects have similarity and clear association with the target slope object; the objects in the set have a unified analysis dimension (such as monitoring indicators, risk level, evolution law), which can be used for comparative analysis of slopes, trend prediction, risk warning, thread training, engineering decision-making and other applications; the set is dynamic and can dynamically adjust the composition of members as the slope object is updated in real time and the similarity and association relationships change.

[0076] In this embodiment, the target slope object can be any slope object. Real-time slope objects similar to the target slope object include the adjacent slope hazard data upstream and downstream of the target slope object. For example, upstream and downstream slope objects directly similar to the target slope object are first-order adjacent slope objects of the target slope object. Alternatively, upstream and downstream slope objects indirectly similar to the target slope object, i.e., given the first-order adjacent slope objects, upstream and downstream slope objects directly similar to the first-order adjacent slope objects are second-order adjacent slope objects of the target slope object. In other embodiments, real-time slope objects similar to the target slope object also include third-order adjacent slope objects of the target slope object.

[0077] S220. Identify the slope objects in the slope object set as slope hazard data, and obtain the hazard descriptor information and non-natural vibration data for each slope hazard data.

[0078] The data includes slope hazard information and key details. Slopes meeting certain criteria (high / medium risk), historical instability records, proximity to sensitive targets (residential areas / main roads within 500m), or structural anomalies) are identified as slope hazard data and assigned a unique identifier (HID). Two types of core information are also collected: first, hazard description factors (including qualitative and quantitative indicators such as soil / rock type, slope, and crack width); second, non-natural vibration data (including vibration source parameters from blasting, construction, and traffic, as well as peak velocity and dominant frequency response data). These data are preprocessed through filtering and calibration to form a standardized dataset.

[0079] The definition and preprocessing of the slope object set includes: constructing a slope object set using methods such as remote sensing image interpretation, on-site survey, geographic information system (GIS) vector data, and slope monitoring equipment access. Each slope object in the set contains basic attributes: unique slope identifier ID, geographic coordinates (latitude and longitude / UTM coordinates), slope type (soil slope / rock slope / mixed soil and rock slope), slope range, and surrounding sensitive targets (residential areas, roads, pipelines, structures, etc.).

[0080] Preprocess the slope objects within the set: remove duplicate data, complete missing basic geographic attributes, and verify coordinate accuracy to form a standardized set of slope objects.

[0081] Slope Hazard Data Identification and Marking: Based on preset slope hazard identification rules, objects in the standardized slope object set that meet the conditions are identified as slope hazard data. The identification rules include (but are not limited to): Risk level rules: The slope instability risk level is high / medium risk (based on the "Slope Engineering Safety Level Standard" GB50330, combined with slope gradient, slope height, soil and rock strength, and support status for comprehensive judgment); Historical event rules: There are records of historical instability / hazard events such as slope collapse, landslide, and crack expansion.

[0082] Sensitive target rule: If there are sensitive targets such as residential areas, main roads, and important pipelines within 500m around the slope, instability may easily cause casualties or property damage;

[0083] Structural anomaly rules: The slope has obvious abnormal features such as cracks, bulging, slip surfaces, and damage / failure of the support structure.

[0084] For slope objects identified as potential hazards, a unique hazard data identifier (HID) is assigned and the "slope hazard data" attribute is marked to form a slope hazard dataset.

[0085] Obtain hazard description factor information for each slope hazard data: For each object marked as slope hazard data, extract hazard description factor information. The factors are divided into qualitative description factors and quantitative description factors.

[0086] Information output requirements: The hazard description factor information for each slope hazard data is stored in structured fields. Qualitative factors use standardized terms (such as "strongly weathered granite" and "local failure of support structure"), and quantitative factors are labeled with units (such as "slope 68°" and "crack width 12cm"), and are bound one-to-one with the hazard data HID.

[0087] Obtaining non-natural vibration data for each slope hazard: Non-natural vibration is a significant contributing factor to slope instability. For each slope hazard, relevant data on non-natural vibration sources and vibration response data are collected. The specific collection content and methods are as follows: Non-natural vibration source data collection: Identify the types of non-natural vibration sources around the slope hazard point and collect basic source term data:

[0088] Vibration source types: blasting vibration (mining / engineering blasting), mechanical construction vibration (excavators, road rollers, drilling rigs), traffic vibration (heavy vehicles, railway trains, subways), and industrial equipment vibration (fans, water pumps, crushers).

[0089] Vibration source parameters: explosive charge / detonation method, construction machinery model / operation time, traffic flow / vehicle load, industrial equipment operating power / frequency;

[0090] Data sources: blasting monitoring reports from construction units, traffic flow data from traffic management departments, equipment operation logs from industrial enterprises, and on-site vibration source investigation records.

[0091] Slope vibration response data acquisition: Vibration monitoring sensors (velocity sensors / accelerometers, sampling frequency 100Hz~1kHz) are deployed at potential hazard points and key locations on the slope (slope top, slope shoulder, slope toe, and cracks) to collect real-time data on the slope's response to non-natural vibrations. Key parameters include:

[0092] Time-domain parameters: peak velocity (PPV), peak acceleration (PGA), vibration duration, and vibration amplitude time history curve;

[0093] Frequency domain parameters: dominant vibration frequency, spectral distribution, and concentrated vibration energy frequency band;

[0094] Spatial parameters: vibration propagation attenuation coefficient along the slope, vibration phase difference at different monitoring points.

[0095] Data association and preprocessing: The collected non-natural vibration source data, vibration response data and slope hazard data HID are bound together to clarify the collection time, monitoring location and vibration source type of the data;

[0096] Vibration data preprocessing: noise is removed by filtering (low-pass filtering / band-pass filtering), outliers (such as sensor fault data) are removed, and sensor accuracy is calibrated to form a standardized non-natural vibration dataset.

[0097] This method achieves accurate transformation of slope objects into slope hazard data through standardized hazard identification, factor extraction, and vibration data acquisition processes. At the same time, it obtains complete correlation data covering "hazard characteristics-inducing factors," providing core data support for subsequent slope hazard risk assessment, early warning model construction, and treatment plan formulation, which meets the requirements of completeness and practicality of patented technical solutions.

[0098] S230. Generate slope hazard data features in the slope object set based on the hazard description factor information and non-natural vibration data of each slope hazard data, so as to obtain the target slope object set.

[0099] The slope hazard data features and target set are as follows: Non-natural vibration data are categorized by data type and multi-scale feature extraction is performed, integrating these to obtain hazard feature weights. Hazard descriptive factor information is also extracted and integrated with the hazard feature weights in a unified feature direction to generate slope hazard data features. Based on these features, the original slope object set is optimized, redundant and invalid data are removed, and the target slope object set is formed.

[0100] Among them, the slope object set refers to a standardized dataset containing multiple slope objects, constructed through remote sensing image interpretation, on-site investigation, integration of geographic information system (GIS) vector data, and access to slope monitoring equipment data. It is the original data basis for slope hazard data screening and feature generation.

[0101] Collection composition: Each slope object is an independent data unit, containing basic attributes (unique identifier ID of the slope, geographic coordinates, slope type, slope range, surrounding sensitive targets, etc.).

[0102] Data source: Fusion of diverse and heterogeneous data (spatial remote sensing data, engineering survey data, on-site inspection data, monitoring equipment data, etc.);

[0103] Technical function: As the original input set for slope hazard identification, it provides a data carrier for subsequent hazard determination and feature extraction.

[0104] Slope hazard data: refers to slope object data that is selected from the set of slope objects according to the preset slope hazard judgment rules (risk level, historical instability events, proximity of sensitive targets, structural anomalies, etc.), which have the risk of slope instability and may cause disasters. It is the core processing object for feature generation.

[0105] Judgment criteria: High / medium risk slopes that meet the standards of "Safety Level Standard for Slope Engineering" (GB50330), slopes with historical hidden dangers such as landslides / collapses / crack expansion, slopes with sensitive targets such as residential areas / main roads within 500m, and slopes with obvious structural anomalies (cracks, bulging, support failure, etc.).

[0106] Identification attribute: Assign a unique identifier (HID) to the hazard data, mark the "slope hazard data" attribute, and distinguish it from ordinary slope objects;

[0107] Technical function: It is the core intermediate data connecting the "original slope object" and the "target slope object set", providing a targeted target for the extraction of hazard description factors and the collection of non-natural vibration data.

[0108] Hazard description factor information: refers to the set of parameters used to qualitatively or quantitatively describe the inherent characteristics, geological environment, engineering structure, hydrological and meteorological and risk-related attributes of slope hazard data. It is the core data reflecting the vulnerability of slope hazards themselves.

[0109] Factor classification: divided into qualitative descriptive factors (rock and soil type, support status, potential instability mode, etc.) and quantitative descriptive factors (slope, slope height, crack width, groundwater depth, stability coefficient, etc.).

[0110] Dimensional coverage: five major dimensions including geological environmental factors, engineering disturbance factors, hydrological and meteorological factors, structural characteristic factors, and risk-related factors;

[0111] Technical role: To characterize the inherent properties of slope hazards and provide "internal" data support for subsequent feature generation by fusing with non-natural vibration data.

[0112] Non-natural vibration data: refers to the vibration source data caused by non-natural factors around the slope hazard point, as well as the monitoring data of the slope hazard point's response to the vibration. It is the core data reflecting the external inducing factors of slope hazards.

[0113] Data type:

[0114] ① Vibration source data: blasting vibration (charge / detonation method), construction machinery vibration (model / operation time), traffic vibration (flow rate / load), industrial equipment vibration (power / frequency), etc.;

[0115] ② Vibration response data: peak velocity (PPV), peak acceleration (PGA), dominant frequency, spectral distribution, vibration attenuation coefficient, and other time-domain / frequency-domain / spatial parameters;

[0116] Data acquisition method: Real-time data is collected via vibration sensors (velocity / accelerometers) and integrated with construction logs, traffic data, etc.

[0117] Technical function: To characterize the external inducing factors of slope hazards, providing "external factor" data support for the subsequent fusion with hazard descriptive factors to generate features.

[0118] Slope hazard data characteristics: refers to the structured numerical / vector set that can characterize the essence of slope hazard risk after integrating, transforming and deriving hazard description factor information (internal factors) and non-natural vibration data (external factors) through data preprocessing, multi-dimensional fusion, mathematical coupling and other methods.

[0119] Feature hierarchy: divided into basic feature layer (basic attributes of slope), hidden danger description factor feature layer (quantification of inherent attributes), non-natural vibration feature layer (quantification of inducing factors), and coupling and fusion feature layer (quantification of interaction between internal and external factors).

[0120] Generation methods: qualitative factor coding (isothermal / tag coding), quantitative factor normalization (Min-Max / Z-score), vibration parameter calculation, factor-vibration coupling operation (product / ratio / correlation), etc.

[0121] Technical role: It transforms raw heterogeneous data into standardized, computable feature vectors, which are the core data units for constructing the target slope object set and provide input for subsequent risk assessment and early warning model training.

[0122] Target slope object set: refers to the standardized slope hazard dataset formed by feature generation, feature filtering and optimization based on slope hazard data. It contains complete core features, no redundant / invalid data, and is the ultimate target data carrier for intelligent analysis and application of slope hazards.

[0123] Set characteristics: no duplicate data, no missing core features, consistent feature dimensions, strong correlation between features and potential risks, and removal of invalid potential risk objects;

[0124] Data composition: Each element is an integrated data unit consisting of "Hazard Identifier (HID) + Basic Attributes + Core Feature Vector + Data Source Information";

[0125] Technical role: It provides high-quality, standardized datasets for slope hazard risk assessment, early warning model construction, and treatment plan formulation, and is the final output of the entire technical process.

[0126] Slope object set (original input) → Slope hazard data obtained by filtering (core processing object) → Hazard description factor information extracted (internal factors) + non-natural vibration data collected (external factors) → Slope hazard data features generated by fusion (core transformation) → Target slope object set obtained after filtering and optimization (final output).

[0127] In this embodiment of the application, for each slope hazard data in the slope object set, the slope hazard data features are obtained by combining the hazard description factor information and non-natural vibration data of the slope hazard data. Based on the slope hazard data features of each slope hazard data, the slope object set is optimized to obtain the target slope object set.

[0128] In one example, the features of slope hazard data include features corresponding to hazard descriptive factor information and features corresponding to non-natural vibration data. Specifically, features are extracted from the hazard descriptive factor information and non-natural vibration data respectively, and the features corresponding to the hazard descriptive factor information and non-natural vibration data are integrated or weighted and summed to obtain the slope hazard data features.

[0129] S240. Extract the features of important hidden dangers from the target slope object set to obtain the description vector of the target slope object set, and determine whether the slope object attributes of the target slope object affect the slope geological disaster risk assessment results based on the description vector of the target slope object set. The slope object attributes are used to characterize the attributes that affect the hidden dangers of the slope object.

[0130] Among them: Description vector generation and risk assessment: Using global averaging or global maximum dimensionality reduction methods, high-dimensional noise-resistant features are transformed into standardized description vectors; regression analysis is completed through decision units to accurately determine the degree of influence of slope object attributes on disaster risk.

[0131] Among them, the feature extraction of important hidden dangers refers to the technical process of selecting and extracting the core features that play a key role in the characterization of slope hidden dangers from the full set of hidden danger features of the target slope object set, based on quantitative indicators such as the correlation between features and slope geological disaster risks, feature importance, and redundancy, through a preset algorithm.

[0132] Extraction criteria: feature importance score (random forest / gradient boosting tree), mutual information value (correlation with risk label), variance threshold (discrimination), correlation coefficient (redundancy), etc.

[0133] Extraction methods: L1 regularization (Lasso), variance thresholding, Pearson correlation analysis, recursive feature elimination (RFE), etc.

[0134] Technical role: Eliminate invalid features with low contribution and high redundancy, retain the core features that can accurately characterize the essential risks of slope hazards, reduce the dimensional complexity of subsequent description vector construction, and improve the accuracy of analysis.

[0135] The descriptive vector of the target slope object set: refers to the structured numerical vector that represents the overall hazard characteristics of the target slope object set by transforming the core hazard characteristics extracted from the important hazard characteristics through numerical coding, normalization, dimensional fusion and other methods. It is a quantitative representation of the hazard characteristics at the set level.

[0136] Vector composition: It consists of extracted key hazard feature dimensions (such as slope, PPV, crack width, vibration dominant frequency, soil and rock type coding value, etc.), with each dimension corresponding to a standardized value;

[0137] Vector form: usually a one-dimensional / multi-dimensional floating-point numerical vector (such as [0.68, 0.32, 0.85, 0.12, ...]), with the dimension consistent with the number of important hidden danger features;

[0138] Technical function: It transforms unstructured / semi-structured hazard characteristics into calculable and comparable mathematical vectors, which are the core inputs for analyzing the impact of slope object attributes on risk assessment results, and realize the quantitative representation of cascade-level hazard characteristics.

[0139] Slope object attributes: refers to the set of parameters used to qualitatively or quantitatively characterize the various inherent attributes, external disturbance attributes, and related attributes that affect the formation, development, and instability risk of slope objects. It is the core input variable for slope geological hazard risk assessment.

[0140] Attribute Classification:

[0141] ①Inherent properties: The slope's own geological and structural properties (rock and soil type, slope, slope height, crack parameters, support status, stability coefficient, etc., and corresponding hazard description factor information);

[0142] ② External disturbance attributes: Slope external inducing factors attributes (non-natural vibration source type, peak velocity / acceleration, rainfall intensity, groundwater level, etc., corresponding non-natural vibration data and hydro-meteorological factors);

[0143] ③ Associated attributes: Risk associated attributes around the slope (sensitive target type / distance, potential instability mode, scope of impact of hidden dangers, etc.);

[0144] Technical role: It directly determines the vulnerability and induced risk of slope hazards, and is the core influencing factor in slope geological disaster risk assessment. It is also the analysis object that needs to be verified in this technical process to determine "whether it affects the assessment results".

[0145] Slope geological hazard risk assessment results: These are quantitative / qualitative conclusions that characterize the risk level, probability of occurrence, potential losses, and risk controllability of slope geological hazards, such as landslides and collapses, calculated based on core data such as slope object attributes and slope hazard characteristics, using qualitative / quantitative risk assessment models (such as analytic hierarchy process, fuzzy comprehensive evaluation method, machine learning model, numerical simulation model, etc.).

[0146] Result type:

[0147] ① Qualitative results: Risk levels such as high risk / medium risk / low risk / no risk;

[0148] ② Quantitative results: probability of slope instability (%), potential economic loss (ten thousand yuan), casualty risk index, risk value (e.g., R = probability × loss), etc.;

[0149] Generation basis: slope object attributes (internal factors + external factors), hazard characteristics, assessment model algorithm and preset thresholds;

[0150] Technical role: It provides core basis for the formulation of slope hazard management plan, the setting of early warning threshold and disaster prevention and mitigation decision-making. It is the output object that needs to be verified in this technical process to determine whether the slope object attributes have an impact.

[0151] In this embodiment, real-time slope objects similar to the target slope object are obtained, and a slope object set based on the target slope object is constructed according to the correlation between the target slope object and the real-time slope objects. The slope objects in the slope object set are identified as slope hazard data, and hazard descriptor information and non-natural vibration data for each slope hazard data are obtained. Slope hazard data features in the slope object set are generated based on the hazard descriptor information and non-natural vibration data for each slope hazard data to obtain the target slope object set. By obtaining non-natural vibration data and combining it with the hazard descriptor information of the slope objects, the slope hazard data can be accurately assessed. The characteristic information of the slope object is used to consider the adjacent slope objects around the target slope object in the slope object set. By using the slope hazard data characteristics in the slope object set, a comprehensive analysis of the target slope object is formed. Then, the characteristic of important hazards is extracted from the target slope object set to obtain the description vector of the target slope object set. Based on the description vector of the target slope object set, regression analysis is performed to determine whether the slope object attributes of the target slope object affect the slope geological hazard risk assessment results. The slope object attributes are used to characterize the attributes that affect the slope object hazards, improve the accuracy of the regression analysis of slope object attribute changes, and ensure the accuracy and confidence of the assessment.

[0152] In one embodiment of this application, another method for assessing slope geological hazards is provided, and details S410 to S430 are described below:

[0153] S410. Perform feature extraction processing on the non-natural vibration data of slope hazard data, and obtain the hazard feature weights of the slope hazard data by extracting the hazard point weights of each segment of the data after feature extraction processing.

[0154] Among them: Hazard feature weight calculation: Non-natural vibration data is divided into local vibration data of different scales, and the hazard features of each segment are extracted to obtain the sub-hazard feature weights. The sub-hazard feature weights are then integrated into the final hazard feature weights through weighted averaging and other methods to improve the comprehensiveness of data representation.

[0155] Optionally, the non-natural vibration data of slope hazard data can be divided into different data types to obtain several data scales of non-natural local vibration data; feature extraction processing can be performed on the non-natural local vibration data of each data scale, and the hazard features of each segment in the data obtained after feature extraction processing can be extracted to obtain the sub-hazard feature weights at different data scales; the sub-hazard feature weights at different data scales can be integrated to obtain the hazard feature weights of slope hazard data.

[0156] Integrating features from different data scales can be achieved, such as integrating the sub-hazard feature weights from different time periods to obtain the hazard feature weights for slope hazard data, or by weighting the features from different data scales and assigning weights according to the importance of each data scale to obtain the hazard feature weights for slope hazard data.

[0157] S420. Extract the features of the hazard description factor information from the slope hazard data to obtain the hazard description factor features, and integrate the hazard feature weights and hazard description factor features of the slope hazard data to obtain the slope hazard data features.

[0158] In this embodiment of the application, the hazard descriptive factor information of the slope hazard data includes several types of attribute information. Then, feature extraction is performed on each type of attribute information, and then feature integration is performed to obtain the hazard descriptive factor features.

[0159] In one example, the hazard feature weights and hazard descriptive factor features of the slope hazard data can be integrated to obtain the slope hazard data features. Among them, the hazard feature weights and hazard descriptive factor features of the slope hazard data have the same feature direction. If the hazard feature weights and hazard descriptive factor features of the slope hazard data have different feature directions, after obtaining the hazard feature weights of the slope hazard data, the hazard feature weights of the slope hazard data can be converted into a feature vector with a fixed direction, which is the feature direction corresponding to the hazard descriptive factor features.

[0160] In one example, the features of slope hazard data can be obtained by weighting and summing the hazard feature weights and hazard descriptive factor features. The weight values ​​of the hazard feature weights and the hazard descriptive factor features can be flexibly updated according to the actual situation.

[0161] S430. Optimize the set of slope objects based on the characteristics of each slope hazard data to obtain the target set of slope objects.

[0162] The slope hazard data features are added to the slope object set to obtain the target slope object set.

[0163] In this embodiment, non-natural vibration data is processed by feature extraction, and the hazard point weight features of each segment are extracted to obtain the hazard feature weight. When obtaining the hazard feature weight, a multi-scale feature extraction mechanism can be introduced to extract features at different data scales. The hazard feature weight and the hazard descriptive factor features of the slope hazard data are integrated to obtain the slope hazard data features, which can accurately evaluate the feature information of the slope object and comprehensively understand the slope object.

[0164] This application provides another method for assessing slope geological hazard risk, details of which are as follows: S510-S540 are described in detail below:

[0165] S510. Determine the target number of analyses for the analysis thread based on the association between the target slope object and the real-time slope object.

[0166] S520: Load the target slope object set into the analysis thread of the target analysis number to extract the noise resistance features of the target slope object set for important hidden dangers.

[0167] Among them: Multi-threaded noise-resistant feature extraction: Based on the features of slope hazard data, a feature queue is constructed, and an adjacency queue is constructed in combination with the connection relationship between slopes. The features are optimized layer by layer through the analysis thread from the first layer to the last layer, effectively filtering data noise and strengthening the core hazard features.

[0168] In one example, the process of obtaining the noise resistance features of the target slope object set by the analysis thread based on the number of target analyses includes: constructing a feature queue based on the slope hazard data features of each slope hazard data in the target slope object set, and constructing an adjacency queue based on the connection relationship between each slope hazard data; loading the feature queue and adjacency queue into the first-level analysis thread, so as to integrate the features of the adjacent slope hazard data of each slope hazard data through the first-level analysis thread to obtain the optimized slope hazard data features; loading the optimized slope hazard data features and adjacency queue into the next-level analysis thread, so as to integrate the features of the adjacent slope hazard data of each slope hazard data through the next-level analysis thread to obtain the optimized slope hazard data features; repeatedly executing the process of loading the optimized slope hazard data features and adjacency queue into the next-level analysis thread until the last-level analysis thread outputs the noise resistance features of the target slope object set.

[0169] S530. Load the noise resistance features into the dimensionality reduction unit for dimensionality reduction processing to obtain the description vector of the target slope object set.

[0170] In this embodiment, the noise resistance features of the target slope object set include the slope hazard data features of all slope hazard data in the slope object set and the correlation features between the slope hazard data. The noise resistance features are loaded into the dimensionality reduction unit, which is used to integrate the feature representations of the slope hazard data of the entire target slope object set into a full-map representation. The dimensionality reduction unit can use global average dimensionality reduction, that is, calculate the average value of all slope hazard data features in the target slope object set to obtain the overall description vector of the target slope object set; the dimensionality reduction unit can also use global maximum dimensionality reduction, calculate the maximum value of all slope hazard data features in the sub-graph to obtain the overall description vector of the target slope object set.

[0171] S540. Load the description vector of the target slope object set into the decision unit to obtain the regression analysis vector, and based on the regression analysis vector being switched to the slope object attribute influence probability value of the target slope object, obtain the slope geological hazard risk assessment result.

[0172] This application also provides another method for assessing the geological hazard risk of slopes. The training steps of the regression analysis thread for the slope object include S610 to S640, which are described in detail below:

[0173] S610. Obtain the example adjacent slope object that is similar to the example test slope object, and the slope object property directory of the example test slope object.

[0174] In this embodiment, the exemplary adjacent slope object includes at least a first-order similar real-time slope object and a second-order similar real-time slope object of the exemplary test slope object. The slope object attribute directory of the exemplary test slope object is used to indicate whether the slope object attributes of the exemplary test slope object have changed.

[0175] S620. Based on the relationship between the example test slope object and the example adjacent slope object, construct a set of example slope objects, and obtain the example hidden danger descriptor information and example non-natural vibration data of each example slope object in the example slope object set.

[0176] Obtain the example hazard descriptor information and example non-natural vibration data of each example slope object in the example slope object set. Each example slope object includes example test slope objects and each example adjacent slope object. In order to facilitate thread training, the obtained data also needs to be preprocessed when obtaining the example hazard descriptor information and example non-natural vibration data.

[0177] In one example, the collection of exemplary slope objects includes several adjacent exemplary slope objects. Exemplary hazard descriptor information and exemplary unnatural vibration data are obtained for each exemplary slope object, including:

[0178] Obtain the risk level and impact data volume of each adjacent slope object in the example slope object set. Clean the target example adjacent slope objects whose risk level is lower than a preset risk level value or whose impact data volume is greater than a preset impact data volume value from the example slope object set to obtain the target example slope object set. Obtain at least one of the following for each example slope object in the target example slope object set: risk level, impact data volume, interference data scale, and slope object type. Standardize the obtained data to obtain the example hazard descriptor information for each example slope object. Obtain the cumulative non-natural vibration data of each example slope object in the target example slope object set within a preset period. If there are null values ​​in the cumulative non-natural vibration data, repair the cumulative non-natural vibration data to obtain the example non-natural vibration data for each example slope object.

[0179] S630. Load the sample slope object set, sample hazard description factor information of each sample slope object, and sample non-natural vibration data into the thread to be trained, generate sample slope object features for each sample slope object, and obtain sample slope geological hazard risk assessment results based on the sample slope object set containing sample slope object features.

[0180] In this embodiment, the set of example slope objects, the example hazard descriptor information of each example slope object, and the example non-natural vibration data are loaded into the training thread. The training thread first generates slope object features of each example slope object based on the example hazard descriptor information and the example non-natural vibration data. Then, the set of example slope objects is optimized. The set of example slope objects containing slope object features is used to extract the features of important hazards to obtain the example description vector of the set of example slope objects. Finally, the geological hazard risk assessment result of the example slope is obtained based on the example description vector.

[0181] S640. Based on the geological hazard risk assessment results of the example slope and the slope object attribute catalog of the example test slope object, construct a prediction reliability index, and update the thread coefficients of the thread to be trained according to the prediction reliability index to obtain the slope object regression analysis thread.

[0182] Among them: slope object regression analysis thread training

[0183] Construct the training dataset: Obtain example test slope objects, similar example adjacent slope objects, and slope object attribute directories to form an example slope object set. Clean up invalid samples with low-risk and high-impact data volumes, and collect and standardize example hazard descriptor information and non-natural vibration data (null values ​​are repaired).

[0184] Generate example features: By integrating the example hazard feature weights and hazard descriptor features through the thread to be trained, and combining the historical changes of hazard point weights, real-time geological features, and unique event features (such as special construction and extreme events), the example slope object features are obtained through weighted processing or significance analysis.

[0185] Thread optimization and update: Based on the example features, the predicted risk assessment results are obtained. The cross-entropy prediction reliability index is constructed by comparing it with the actual attribute catalog. The thread coefficients are iteratively updated until convergence, forming the optimized slope object regression analysis thread.

[0186] In this embodiment, the geological hazard risk assessment result of the exemplary slope includes the probability of the exemplary regression analysis corresponding to the change of the slope object attributes of the exemplary test slope object. A prediction confidence index is constructed based on the difference between the exemplary regression analysis probability and the actual probability corresponding to the slope object attribute catalog. For example, the prediction confidence index is the cross-entropy prediction confidence index. Then, the thread coefficients of the thread to be trained are updated according to the prediction confidence index until the thread converges. If the difference between the exemplary regression analysis probability and the actual probability output by the thread is less than a preset difference value, the slope object regression analysis thread is obtained.

[0187] In this embodiment, the hazard feature weights of the hazard points and the topological information of the slope object are combined for modeling, and whether the slope object has undergone attribute changes is determined as a directory. Threads are trained to find whether the slope object has undergone attribute changes within this time series period and determine the output. The question of whether the slope object has undergone attribute changes is transformed into a classification question through the thread. The thread can accurately perform regression analysis on whether the slope object's attributes have changed.

[0188] In one embodiment of this application, another method for assessing slope geological hazards is also provided, details of which are described in S710 to S730 below:

[0189] S710. The example integrated features are obtained by integrating the example hazard feature weights corresponding to the example non-natural vibration data and the example hazard descriptor feature information corresponding to the example hazard descriptor information through the thread to be trained.

[0190] In this embodiment, the thread to be trained is used to perform feature extraction processing on the example non-natural vibration data. After feature extraction processing, the hidden danger features of each segment in the data are extracted to obtain the example hidden danger feature weights. Then, the example hidden danger feature weights and example hidden danger descriptive factor features are integrated or weighted to obtain the example integrated features.

[0191] S720. Obtain the historical changes in the weights of potential hazards on the example slope and the real-time geological characteristics of the example slope.

[0192] In this embodiment of the application, a dynamic weight update mechanism is introduced during the thread training process. The weights of the slope hazard data features are updated according to historical changes and real-time geology. This results in weighted processing of the example integration features, improving the thread's ability to adapt to changing geology, helping the thread to more accurately capture the hazard feature weights and noise resistance features of the slope hazard data, and improving the accuracy of regression analysis.

[0193] S730. Based on the historical changes and geological characteristics of the examples, the integrated features of the examples of the example slope objects are weighted to obtain the characteristics of the example slope objects. Based on the set of example slope objects containing the characteristics of the example slope objects, the geological hazard risk assessment results of the example slopes are obtained.

[0194] In this embodiment, feature weights are calculated based on the historical changes in the weights of potential hazard points and the degree of change in the geological features of the examples. Then, the integrated features of the example slope objects are weighted to obtain the features of the example slope objects. Finally, the geological hazard risk assessment results of the example slopes are obtained based on the set of example slope objects containing the features of the example slope objects.

[0195] In one embodiment of this application, another method for assessing slope geological hazards is also provided, and details S810 to S830 are described below:

[0196] S810. By integrating the example hazard feature weights corresponding to the example non-natural vibration data and the example hazard descriptor feature corresponding to the example hazard descriptor information in the target direction through the thread to be trained, the example integrated feature is obtained.

[0197] In this embodiment, the thread to be trained is used to perform feature extraction processing on the example non-natural vibration data. After feature extraction processing, the hidden danger features of each segment in the data are extracted to obtain the example hidden danger feature weights. If the feature directions of the example hidden danger feature weights and the example hidden danger descriptive factor features are the same, feature integration is performed directly. If the feature directions of the example hidden danger feature weights and the example hidden danger descriptive factor features are different, the example hidden danger feature weights and the example hidden danger descriptive factor features are switched to the features of the target direction, and then feature integration is performed to obtain the example integrated features. The switching of the example hidden danger feature weights and the example hidden danger descriptive factor features to the features of the target direction can be done using deep learning technology.

[0198] In one example, to improve the accuracy of thread regression analysis, in addition to considering the example non-natural vibration data and example hazard descriptor information, unique event information can also be considered during thread training. Therefore, when obtaining the example integrated features, the event features corresponding to the unique event information can also be considered. For example, if the example slope object has unique event information within the period corresponding to the example non-natural vibration data, the event features are obtained by feature extraction of the unique event information. The example integrated features are obtained by integrating the example hazard feature weights corresponding to the example non-natural vibration data, the example hazard descriptor features corresponding to the example hazard descriptor information, and the event features in the target direction through the thread to be trained.

[0199] Among them, the non-natural vibration data of the example is the weight of the example hidden danger points in the previous month. Then it is determined whether there is unique event information for the example slope object in the previous month. If there is unique event information, the unique event information may affect the weight of the hidden danger points. Therefore, it is necessary to consider the unique event information and extract the event features from the unique event information.

[0200] First, feature extraction is performed on the thread to be trained to obtain the example hazard feature weights and example hazard descriptor features. In one example, the example hazard feature weights, example hazard descriptor information, and event features are directly integrated in the target direction to obtain the example integrated features. In another example, the example hazard feature weights and event features are first associated, for example, the event features and example hazard feature weights are time-aligned to ensure that the event features match the corresponding time series data points. Then, the event features and hazard feature weights can be interactively processed, such as using feature multiplication, feature summation, etc. After obtaining the associated features, they can be integrated with the example hazard descriptor features to obtain the example integrated features.

[0201] S820. Load the example integration features into the saliency analysis unit of the thread to be trained to obtain the saliency weights corresponding to the example integration features.

[0202] S830. The feature vector of the example integration is weighted according to the significance weight to obtain the feature of the example slope object, and the geological hazard risk assessment result of the example slope is obtained based on the set of example slope objects containing the feature of the example slope object.

[0203] The integrated feature vector is weighted using a saliency weight vector, which involves element-wise multiplication of the weight vector and the feature vector to obtain the features of the example slope object. Then, the geological hazard risk assessment result of the example slope is obtained based on the set of example slope objects containing the features of the example slope object.

[0204] It should be noted that, in this embodiment, the example slope object features can be obtained simultaneously by using the example historical change request, example geological features, and attention mechanism. For example, the example hazard feature weights corresponding to the example non-natural vibration data and the example hazard descriptor features corresponding to the example hazard descriptor information are integrated in the target direction through the thread to be trained to obtain example integrated features. Then, based on the example historical change and example geological features, the example integrated features of the example slope object are weighted to obtain the first example slope object features of the example slope object. The example integrated feature vector is weighted according to the significance weight to obtain the second example slope object features. The average value of the first example slope object features and the second example slope object features can be determined as the final example slope object features, or a feature with a larger feature value can be selected from the first example slope object features and the second example slope object features to be determined as the final example slope object features.

[0205] In one embodiment of this application, another method for assessing slope geological hazards is also provided, and details S910-S920 are described below:

[0206] S910. If the slope geological hazard risk assessment result indicates that the probability value of the slope object attribute of the target slope object is higher than the preset probability value, then the target slope object will be sent to the data acquisition device to collect the changed slope object attributes of the target slope object through the data acquisition device.

[0207] S920. Associate the collected changed slope object attributes with the target slope object and enter them into the landslide database.

[0208] In this embodiment of the application, if the slope geological disaster risk assessment result indicates that the probability value of the slope object attribute of the target slope object is higher than the preset probability value, it indicates that the target slope object is very likely to undergo attribute changes within a certain time period. At this time, it is necessary to send the target slope object to the data acquisition device. The data acquisition device will collect the slope object attribute of the target slope object and associate the collected changed slope object attribute with the target slope object.

[0209] In this embodiment of the application, when the probability value of the slope object attribute of the target slope object is higher than the preset probability value, data is collected by a data acquisition device and associated with the target slope object to improve the accuracy and reliability of landslide identification.

[0210] Based on the above, a slope geological hazard risk assessment system is shown, including a processor and a memory that communicate with each other. The processor is used to read a computer program from the memory and execute it to implement the above method.

[0211] Based on the above, a computer-readable storage medium is also provided, on which a computer program stored implements the above method during runtime.

[0212] In summary, based on the above scheme, real-time slope objects similar to the target slope object are obtained, and a slope object set based on the target slope object is constructed according to the correlation between the target slope object and the real-time slope objects. The slope objects in the slope object set are identified as slope hazard data, and hazard descriptor information and non-natural vibration data are obtained for each hazard data point. Based on the hazard descriptor information and non-natural vibration data of each hazard data point, slope hazard data features in the slope object set are generated to obtain the target slope object set. By obtaining non-natural vibration data and combining it with the hazard descriptor information of the slope objects, the slope hazard data can be accurately assessed. The characteristic information of the slope object is used to consider the adjacent slope objects around the target slope object in the slope object set. By using the slope hazard data characteristics in the slope object set, a comprehensive analysis of the target slope object is formed. Then, the characteristic of important hazards is extracted from the target slope object set to obtain the description vector of the target slope object set. Based on the description vector of the target slope object set, regression analysis is performed to determine whether the slope object attributes of the target slope object affect the slope geological hazard risk assessment results. The slope object attributes are used to characterize the attributes that affect the slope object hazards, improve the accuracy of the regression analysis of slope object attribute changes, and ensure the accuracy and confidence of the assessment.

[0213] It should be understood that the systems and modules described above can be implemented in various ways. For example, in some embodiments, the systems and modules can be implemented by hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the methods and systems described above can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The systems and modules of this application can be implemented not only by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., but also by software executed by various types of processors, or by a combination of the aforementioned hardware circuits and software (e.g., firmware).

[0214] It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects may be any one or a combination of the above, or any other possible beneficial effects.

Claims

1. A method for assessing geological hazard risks on slopes, characterized in that, include: Obtain real-time slope objects similar to the target slope object, and construct a set of slope objects based on the target slope object according to the association between the target slope object and the real-time slope objects; The slope objects in the slope object set are identified as slope hazard data, and the hazard descriptor information and non-natural vibration data of each slope hazard data are obtained. Based on the hazard description factor information and non-natural vibration data of each slope hazard data, the slope hazard data features in the slope object set are generated to obtain the target slope object set; The feature extraction of important hidden dangers of the target slope object set is used to obtain the description vector of the target slope object set, and the slope geological disaster risk assessment result is obtained based on the description vector of the target slope object set; The process of extracting features of significant hidden dangers from the target slope object set to obtain a description vector for the target slope object set, and determining whether the slope object attributes of the target slope objects affect the slope geological hazard risk assessment results based on the description vector of the target slope object set, includes: The target number of analyses for the analysis thread is determined based on the association between the target slope object and the real-time slope object. The target slope object set is loaded into the analysis thread of the target analysis number to extract the features of important hidden dangers and obtain the anti-noise features of the target slope object set; The noise resistance features are loaded into the dimensionality reduction unit for dimensionality reduction processing to obtain the description vector of the target slope object set; The description vector of the target slope object set is loaded into the decision unit to obtain the regression analysis vector. Based on the regression analysis vector being switched to the slope object attribute influence probability value of the target slope object, the slope geological hazard risk assessment result is obtained. The slope object attribute is used to characterize the attribute that affects the hidden danger of the slope object.

2. The method as described in claim 1, characterized in that, The process of generating slope hazard data features in the slope object set based on the hazard descriptor information and non-natural vibration data of each slope hazard data to obtain the target slope object set includes: The non-natural vibration data of the slope hazard data is subjected to feature extraction processing, and the hazard point weight features of each segment of the data obtained after feature extraction processing are extracted to obtain the hazard feature weight of the slope hazard data. The hazard descriptive factor information of the slope hazard data is extracted to obtain the hazard descriptive factor features, and the hazard feature weights of the slope hazard data and the hazard descriptive factor features are integrated to obtain the slope hazard data features of the slope hazard data; The slope object set is optimized based on the slope hazard data characteristics of each slope hazard data point to obtain the target slope object set.

3. The method as described in claim 2, characterized in that, The process of extracting features from the non-natural vibration data of the slope hazard data, and extracting the hazard point weight features of each segment of the data after feature extraction to obtain the hazard feature weights of the slope hazard data, includes: The non-natural vibration data of the slope hazard data are divided according to different data types to obtain several data scales of non-natural local vibration data. Feature extraction processing is performed on non-natural local vibration data of various data scales, and the hidden danger features of each segment in the data obtained after feature extraction processing are extracted to obtain the sub-hidden danger feature weights at different data scales. The sub-hazard feature weights of the slope hazard data are obtained by integrating and processing the sub-hazard feature weights of the different data scales.

4. The method as described in claim 1, characterized in that, The step of loading the target slope object set into the analysis thread of the target analysis number and extracting the features of important hidden dangers to obtain the noise resistance features of the target slope object set includes: A feature queue is constructed based on the slope hazard data features of each slope hazard data in the target slope object set, and an adjacency queue is constructed based on the connection relationship between each slope hazard data. The feature queue and the adjacency queue are loaded into the first-level analysis thread so that the features of the adjacent slope hazard data of each slope hazard data are integrated through the first-level analysis thread to obtain the optimized slope hazard data features. The optimized slope hazard data features and the adjacent queue are loaded into the next-level analysis thread so that the features of the adjacent slope hazard data of each slope hazard data can be integrated through the next-level analysis thread to obtain the optimized slope hazard data features. Repeatedly load the optimized slope hazard data features and the adjacency queue into the next layer of analysis thread until the last layer of analysis thread outputs the noise resistance features of the target slope object set.

5. The method according to any one of claims 1 to 4, characterized in that, The pre-trained slope object regression analysis thread generates slope hazard data features in the slope object set based on the hazard descriptor information and non-natural vibration data of each slope hazard data. It also extracts the features of important hazards in the target slope object set to obtain the description vector of the target slope object set. Based on the description vector of the target slope object set, the slope geological disaster risk assessment result is obtained. The training steps for the slope object regression analysis thread include: Obtain the example adjacent slope object that is similar to the example test slope object, and the slope object attribute directory of the example test slope object; Based on the association between the example test slope object and the example adjacent slope object, a set of example slope objects is constructed, and the example hidden danger descriptor information and example non-natural vibration data of each example slope object in the example slope object set are obtained; The example slope object set, the example hazard description factor information of each example slope object, and the example non-natural vibration data are loaded into the training thread to generate the example slope object features of each example slope object, and the example slope geological hazard risk assessment results are obtained based on the example slope object set containing the example slope object features. Based on the geological hazard risk assessment results of the example slope and the slope object attribute catalog of the example test slope object, a prediction reliability index is constructed, and the thread coefficients of the thread to be trained are updated according to the prediction reliability index to obtain the slope object regression analysis thread.

6. The method as described in claim 5, characterized in that, The collection of exemplary slope objects includes several exemplary adjacent slope objects; The acquisition of exemplary hazard descriptor information and exemplary non-natural vibration data for each exemplary slope object in the exemplary slope object set includes: Obtain the risk level and impact data volume of each adjacent slope object in the example slope object set; clean the target example adjacent slope objects from the example slope object set if the slope object risk level is lower than a preset risk level specified value or the slope object impact data volume is greater than a preset impact data volume specified value, and obtain the target example slope object set. Obtain at least one of the following in the target example slope object set: risk level, amount of impact data, scale of interference data, and slope object type for each example slope object. Then, standardize the obtained data to obtain example hazard descriptor information for each example slope object. Obtain the cumulative non-natural vibration data of each example slope object in the target example slope object set within a preset period; If there are null values ​​in the cumulative unnatural vibration data, the cumulative unnatural vibration data is repaired to obtain the example unnatural vibration data for each example slope object.

7. The method as described in claim 5, characterized in that, The process of loading the set of example slope objects, the example hazard descriptor information of each example slope object, and the example non-natural vibration data into the training thread to generate the example slope object features for each example slope object includes: The example integrated feature is obtained by integrating the example hazard feature weights corresponding to the example non-natural vibration data and the example hazard descriptive factor features corresponding to the example hazard descriptive factor information through the thread to be trained. Obtain the historical changes in the weights of potential hazard points of the example slope object and the real-time geological characteristics of the example; Based on the historical changes and geological characteristics of the examples, the integrated features of the example slope objects are weighted to obtain the example slope object features of the example slope objects; The step of loading the set of exemplary slope objects, the exemplary hazard descriptor information of each slope object, and the exemplary non-natural vibration data into the thread to be trained to generate the exemplary slope object features of each slope object includes: The training thread integrates the example hazard feature weights corresponding to the example non-natural vibration data and the example hazard descriptor features corresponding to the example hazard descriptor information in the target direction to obtain example integrated features. The example integration features are loaded into the saliency analysis unit of the thread to be trained to obtain the saliency weights corresponding to the example integration features; The feature vector of the example is weighted according to the significance weight to obtain the feature of the example slope object; The step of integrating the example hazard feature weights corresponding to the example non-natural vibration data and the example hazard descriptor factor features corresponding to the example hazard descriptor information in the target direction through the training thread to obtain example integrated features includes: If the example slope object has unique information within the corresponding period of the example non-natural vibration data, feature extraction is performed on the unique information to obtain the feature; The training thread integrates the example hazard feature weights corresponding to the example non-natural vibration data, the example hazard descriptor features corresponding to the example hazard descriptor information, and the event features in the target direction to obtain example integrated features.

8. The method as described in claim 1, characterized in that, After determining whether the slope object attributes of the target slope objects affect the slope geological hazard risk assessment results based on the description vector of the target slope object set, the process includes: If the slope geological hazard risk assessment result indicates that the probability value of the slope object attribute impact of the target slope object is higher than the preset probability value, then the target slope object is sent to the data acquisition device to collect the changed slope object attributes of the target slope object through the data acquisition device; The collected changes in slope object attributes are associated with the target slope object and entered into the landslide database.

9. A slope geological hazard risk assessment system, characterized in that, The method includes a processor and a memory that communicate with each other, the processor being configured to read a computer program from the memory and execute it to implement the method of any one of claims 1-8.

Images