A method and system for monitoring the source of a debris flow

Abstract

This application provides a method and system for monitoring debris flow sources. Based on the matching data between monitoring features corresponding to each debris flow attribute data, the system integrates these monitoring features to obtain global monitoring features corresponding to the debris flow source data set to be identified. Analyzing these global monitoring features yields the analysis results for the debris flow source data set to be identified. In the task of analyzing and processing the source data set, combining the monitoring features corresponding to multiple debris flow attribute data in the source data set to obtain more representative global monitoring features can accurately obtain the source hazard analysis results, thus improving the accuracy of source data set analysis and processing.

Description

Technical Field

[0001] This application relates to the field of debris flow source monitoring technology, and more specifically, to a debris flow source monitoring method and system. Background Technology

[0002] Debris flows are a frequent geological disaster in mountainous areas. Their formation is closely related to the occurrence, stability, and activation conditions of the debris source. The type, reserves, distribution, and dynamic evolution characteristics of the debris source directly determine the probability, scale, and severity of debris flows. Therefore, accurate and efficient monitoring and analysis of debris flow debris sources is the core prerequisite for achieving early warning and scientific prevention and control of debris flow disasters.

[0003] Currently, the field of debris flow source monitoring has formed a monitoring system based on various technologies such as field survey, remote sensing interpretation, and sensor monitoring. It can collect raw data from multiple dimensions, including source type, spatial distribution, physical and mechanical properties, and dynamic evolution. However, there are still many technical shortcomings in the data processing and analysis stage, making it difficult to meet the needs of refined and intelligent source monitoring. The specific problems are reflected in the following aspects:

[0004] The lack of a standardized system for data processing: The raw data of debris flow sources obtained from monitoring are diverse, heterogeneous in format, and complex in dimensions, and contain a large amount of invalid and interfering data. Existing technologies have not established a unified filtering and standardized processing procedure for multi-attribute data of debris sources, which easily leads to low data validity and poor consistency, and cannot provide a reliable data foundation for subsequent analysis. At the same time, there is a lack of reasonable supplementation strategies for missing debris source attribute data, which further affects the comprehensiveness of monitoring and analysis.

[0005] Feature extraction methods are simplistic and lack specificity: The inherent characteristic patterns of different attribute data of material sources differ significantly, while existing technologies mostly use general feature extraction methods to process material source data. They fail to design differentiated feature extraction strategies based on the characteristics of different types of data such as key material source type data, historical evolution data, and hidden danger standard data. This makes it difficult to uncover the core monitoring features contained in the data, resulting in weak feature representation capabilities and an inability to accurately reflect the actual hidden danger status of the material source.

[0006] Feature integration lacks consideration of correlation: Existing technologies often integrate extracted source features by simply stacking or splicing them together, without analyzing the inherent correlation between different source attribute features. This easily leads to feature redundancy and information overlap, making it impossible to construct a global feature that can comprehensively reflect the overall characteristics of the source. As a result, subsequent analysis can only reflect the state of a single dimension of the source, making it difficult to achieve an overall judgment of potential risks to the source.

[0007] The analysis models suffer from low accuracy and poor generalization ability: existing debris flow source analysis relies heavily on manual interpretation or simple machine learning models, which involve high human intervention, low efficiency, and strong subjectivity. Traditional models are mostly trained based on single-dimensional features and do not fully integrate multi-attribute source features. Furthermore, the model training lacks a dedicated optimization process for debris flow source scenarios, resulting in low accuracy in identifying source hazards and weak generalization ability, making it difficult to adapt to the monitoring and analysis needs of debris flow sources in different watersheds and of different types.

[0008] The monitoring and analysis are not intelligent or automated enough: the existing technology has not yet built a fully intelligent technology system from raw data acquisition, feature extraction, feature integration to analysis result output. There is a lack of effective connection between each link, and a large amount of manual intervention is needed to complete data screening, feature processing and result interpretation. This not only increases the monitoring and operation costs, but also leads to low analysis efficiency and makes it impossible to achieve normalized and real-time monitoring of debris flow sources.

[0009] In summary, given the shortcomings of existing debris flow source monitoring technologies in data processing, feature extraction, integration, and analysis, there is an urgent need to develop a standardized, differentiated, and intelligent debris flow source monitoring method and system. This system would enable efficient processing and in-depth analysis of multi-dimensional source data, improve the accuracy and comprehensiveness of source hazard analysis, and provide reliable technical support for the scientific prevention and control of debris flow disasters. Summary of the Invention

[0010] To address the technical problems existing in related technologies, this application provides a method and system for monitoring the source of debris flows.

[0011] Firstly, a method for monitoring the source of debris flows is provided, the method comprising:

[0012] Obtain X debris flow attribute data corresponding to the data set of debris flow sources to be identified;

[0013] Feature extraction is performed on each debris flow attribute data to obtain the data monitoring features corresponding to each debris flow attribute data, and debris flow source attribute matching data between the data monitoring features corresponding to each debris flow attribute data is obtained.

[0014] Based on the debris flow source attribute matching data, feature integration processing is performed on the data monitoring features corresponding to each debris flow attribute data to obtain the global data monitoring features corresponding to the debris flow source data set to be identified.

[0015] The global data monitoring features are analyzed and processed to obtain the analysis results corresponding to the debris flow source data set to be identified.

[0016] In this application, obtaining X debris flow attribute data corresponding to the debris flow source data set to be identified includes:

[0017] Obtain the range of debris flow source attribute types in the debris flow source monitoring command, and based on the debris flow source attribute types within the range of debris flow source attribute types, parse the debris flow source data set to be identified to obtain the original data that matches the debris flow source attribute types within the range of debris flow source attribute types.

[0018] The original data is filtered to obtain X debris flow attribute data corresponding to the debris flow source data set to be identified.

[0019] In this application, the step of performing data filtering on the original data to obtain X debris flow attribute data corresponding to the debris flow source data set to be identified includes:

[0020] The quality of the mountain-related information in the original data is verified to obtain the quality verification results of the mountain-related information.

[0021] If the quality verification result of the mountain matter indicates that the verification is qualified, the mountain matter information will be standardized to obtain standard data of debris flow source hazards;

[0022] Key source type data are obtained from the source type data covered by the original data, and the source historical data in the original data is subjected to interference removal processing to obtain interference-removed source historical data.

[0023] Based on the standard data of debris flow source hazards, the key source type data, and the historical data of the de-interference source, X debris flow attribute data corresponding to the data set of debris flow sources to be identified are obtained.

[0024] In this application, the X debris flow attribute data include key source type data, historical data of de-interferenced sources, and standard data on debris flow source hazards; the feature extraction for each debris flow attribute data to obtain the data monitoring features corresponding to each debris flow attribute data includes:

[0025] The key material source type data is transformed into material source weight features. The material source weight features are then convolved to obtain hidden danger weight features. The material source weight features and the hidden danger weight features are used as data monitoring features corresponding to the key material source type data.

[0026] The source historical data of the de-interference material is subjected to convolution processing to obtain the source factor features corresponding to the source historical data of the de-interference material;

[0027] The historical data of the interference removal source is analyzed to obtain the feature analysis results. The source factor features and the feature analysis results are used as the data monitoring features corresponding to the historical data of the interference removal source.

[0028] The confidence level integration information associated with the standard data of debris flow source hazards is obtained, and the confidence level integration information and the standard data of debris flow source hazards are convolved to obtain the data monitoring features corresponding to the standard data of debris flow source hazards.

[0029] In this application, the step of converting the key source type data into source weight features includes:

[0030] The key material source type data is converted into W hidden danger factors, and the hidden danger vectors corresponding to the W hidden danger factors are obtained;

[0031] Based on the important information of the W hidden danger factors in the key material source type data, the feature vectors corresponding to the W hidden danger factors are obtained.

[0032] Based on the location of the hidden danger areas in the key material source type data using the W hidden danger factors, the location vectors corresponding to the W hidden danger factors are obtained.

[0033] The hazard vector, the feature vector, and the location vector are combined to obtain the material source weight feature corresponding to the key material source type data.

[0034] In this application, the step of performing convolution processing on the de-interference source history data to obtain the source factor features corresponding to the de-interference source history data includes:

[0035] The historical data of the de-interference material source is loaded into the material source data convolutional unit to obtain the input feature of the m-th feature extraction module in the material source data convolutional unit; when m is 1, the input feature of the m-th feature extraction module is the historical data of the de-interference material source.

[0036] Based on one or more feature extraction layers in the m-th feature extraction module, the input features of the m-th feature extraction module are processed to obtain the processing result;

[0037] Based on the weight coefficients corresponding to the simplification layer in the m-th feature extraction module, the processing result is simplified to obtain simplified features.

[0038] The simplified features and the input features of the m-th feature extraction module are combined to obtain the output features of the m-th feature extraction module. The output features of the Y-th feature extraction module in the source data convolution unit are used as the source factor features corresponding to the de-interference source historical data.

[0039] In this application, the analysis of the historical data of the interference source to obtain feature analysis results includes:

[0040] Data detection is performed on the historical data of the interference removal source to obtain data detection results. Based on the data detection results, the historical data of the interference removal source is cleaned to obtain cleaning results.

[0041] The cleaning results are classified to obtain classification results. The classification results are then analyzed to obtain feature analysis results corresponding to the historical data of the interference source.

[0042] In this application, the step of performing feature integration processing on the data monitoring features corresponding to each debris flow attribute data based on the debris flow source attribute matching data to obtain the global data monitoring features corresponding to the debris flow source data set to be identified includes:

[0043] If the debris flow source attribute matching data indicates that there is a correlation between the data monitoring feature corresponding to the a-th debris flow attribute data and the data monitoring feature corresponding to the b-th debris flow attribute data, then the debris flow attribute data corresponding to the a-th debris flow attribute data and the data monitoring feature corresponding to the b-th debris flow attribute data are fused to obtain a debris flow source attribute combination feature.

[0044] The combined features of debris flow source attributes and the data monitoring features corresponding to each debris flow attribute data are processed to obtain the global data monitoring features corresponding to the set of debris flow source data to be identified.

[0045] In this application, the step of analyzing and processing the global data monitoring features to obtain the analysis results corresponding to the debris flow source data set to be identified includes:

[0046] The global data monitoring features are loaded into the first data analysis network, and the first regression analysis results corresponding to the debris flow source data set to be identified are output through the first data analysis network.

[0047] The global data monitoring features are loaded into the second data analysis network, and the second regression analysis results corresponding to the data set of debris flow sources to be identified are output through the second data analysis network.

[0048] The results of the first regression analysis and the second regression analysis are processed by a function to obtain the analysis results corresponding to the data set of debris flow sources to be identified.

[0049] In this application, the method further includes:

[0050] Obtain the attribute data of X typical debris flows corresponding to the typical material source data set;

[0051] Feature extraction is performed on the debris flow attribute data of each example to obtain the example data monitoring features corresponding to each example debris flow attribute data, and feature matching data between the example data monitoring features corresponding to each example debris flow attribute data is obtained.

[0052] Based on the feature matching data, feature integration processing is performed on the example data monitoring features corresponding to each example debris flow attribute data to obtain the global debris flow source attribute example features corresponding to the example source data set.

[0053] The global debris flow source attribute example features are loaded into a pre-trained network, and the global debris flow source attribute example features are analyzed and processed through the pre-trained network to obtain the example category corresponding to the example source data set.

[0054] Based on the anomalies between the example category and the example directory corresponding to the example source data set, the network coefficients of the pre-trained network are optimized, and the pre-trained network covering the optimized network coefficients is used as the source data set data analysis network; the source data set data analysis network includes at least one of a first data analysis network and a second data analysis network.

[0055] Secondly, a debris flow source monitoring system is provided, comprising 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-described method.

[0056] The debris flow source monitoring method and system provided in this application involves obtaining X debris flow attribute data corresponding to each debris flow source data set to be identified. Feature extraction is performed on each debris flow attribute data to obtain a corresponding data monitoring feature. Furthermore, based on the matching data between the data monitoring features corresponding to each debris flow attribute data, feature integration processing is performed on the data monitoring features corresponding to each debris flow attribute data to obtain a global data monitoring feature corresponding to the debris flow source data set to be identified. Analysis of this global data monitoring feature yields the analysis results corresponding to the debris flow source data set to be identified. In the source data set analysis and processing task, combining the data monitoring features corresponding to multiple debris flow attribute data in the debris flow source data set to be identified yields a more representative global data monitoring feature. This global data monitoring feature allows for accurate analysis of source hazard results, improving the accuracy of source data set analysis and processing. 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 debris flow source monitoring 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 illustrates a method for monitoring the source of debris flows, which may include the technical solutions described in steps S101-S104.

[0061] Step S101: Obtain X debris flow attribute data corresponding to the debris flow source data set to be identified.

[0062] Among them, the X debris flow attribute data include six categories: source type, spatial distribution, geometric morphology, physical mechanics, reserves and recharge, and dynamic evolution.

[0063] Data on source types: Based on their formation and distribution, debris flow sources can be divided into the following categories, which need to be statistically analyzed one by one;

[0064] Material sources for landslides: landslides, collapses, and debris flow deposits (the most important material source in strong earthquake zones);

[0065] Sources of material on slopes: slope residual deposits, weathered debris, and slope gully deposits;

[0066] Source of sediment in the gully: gully bed deposits, terrace deposits, and historical debris flow deposits;

[0067] Artificial material sources: engineering waste, tailings, excavated stockpiles, and excavated soil and rocks;

[0068] Other sources: glacial till, volcanic pyroclastic material, material from fault fracture zones.

[0069] Spatial distribution data:

[0070] Source location: coordinates (latitude and longitude / projected coordinates), gully area, slope (upper / middle / lower slope), gully segment (upper / middle / lower).

[0071] Distribution range: source boundary, distribution area, planar morphology (strips / sheets / dots);

[0072] Distribution density: number of sources / area (sources / km²), ratio of source area to watershed area;

[0073] Spatial relationships: distance from the ditch, distance from roads / residential areas, and relationship with fault / seismic intensity zones;

[0074] Geometric morphological data:

[0075] Planarity factors: major axis, minor axis, shape factor, perimeter, and area of ​​the source material;

[0076] Profile parameters: source thickness (average / maximum / minimum), slope, aspect, and height;

[0077] Volume factor: single source volume, total source volume, source volume (m³).

[0078] Channel coefficients: longitudinal slope of the channel bed, channel width, location of the choke / drop, and morphology of the accretion fan;

[0079] Physical and mechanical data:

[0080] Material composition: particle size distribution (content of clay / silt / sand / gravel / rubble), lithology, degree of weathering (full / strong / medium / weak weathering);

[0081] Physical properties: natural density, dry density, moisture content, void ratio, degree of saturation;

[0082] Mechanical properties: internal friction angle, cohesion, shear strength, angle of repose, permeability coefficient;

[0083] Structural characteristics: looseness, degree of cementation, and development of bedding / joints;

[0084] Reserves and supply data:

[0085] Total reserves: The total amount of loose material (m³) within the watershed that can participate in debris flows.

[0086] Available reserves: the amount of material that can be easily activated in the near term (1–5 years) and the critical activation conditions;

[0087] Recharge rate: Annual increase in material source (collapse / erosion / artificial waste), recharge method (gravity / hydraulic / human-made);

[0088] Connectivity: the degree of connection between the material source and the channel, and the path and distance to the channel after startup;

[0089] Dynamic evolution data:

[0090] Temporal variations: Changes in the area, volume, and density of source material at different times (before / after / year by year);

[0091] Evolutionary trends: activation / stabilization / decay rate of biomass, and the impact of vegetation restoration on biomass;

[0092] Triggering conditions: critical rainfall intensity, rainfall duration, previous effective rainfall, and ground motion coefficient;

[0093] Engineering impact: Changes in material source quantity and stability caused by retaining / drainage / slope stabilization projects;

[0094] In the source data set analysis and processing task involved in the embodiments of this application, the range of debris flow source attribute types in the source data set analysis and processing task can be preset; the range of debris flow source attribute types can be used to limit the types of debris flow attribute data that the source data set analysis and processing task can identify, and the range of debris flow source attribute types can include two or more types of debris flow source attributes.

[0095] Based on the six debris flow source attribute types mentioned above (within the range of debris flow source attribute types), the relevant information of the debris flow source dataset to be identified can be analyzed to obtain raw data that matches the debris flow source attribute types within the range of debris flow source attribute types. After data filtering of the raw data, X debris flow attribute data corresponding to the debris flow source dataset to be identified can be obtained; X can be an integer greater than 1, specifically the number of debris flow source attribute types covered within the range of debris flow source attribute types, where X is 6.

[0096] Optionally, if the debris flow source data set to be identified only covers data corresponding to some of the debris flow source attribute types within the aforementioned range of debris flow source attribute types, then the original data corresponding to some debris flow source attribute types can be obtained from the debris flow source data set to be identified; since the debris flow source data set to be identified does not cover data corresponding to the remaining debris flow source attribute types, a pre-set default feature can be used as the original data corresponding to the remaining debris flow source attribute types.

[0097] After obtaining the raw data corresponding to each debris flow source attribute type from the debris flow source dataset to be identified, preprocessing can be performed on the raw data corresponding to each debris flow source attribute type (e.g., data filtering). Different preprocessing methods can be used for the raw data of different debris flow source attribute types, or the same preprocessing method can be used; this application does not limit this. For example, computer equipment can perform quality verification on the mountain event information in the raw data to obtain the mountain event quality verification result corresponding to the mountain event information. If the mountain event quality verification result indicates that the verification is qualified, the mountain event information can be standardized to obtain standard data for debris flow source hazards. Key source type data is obtained from the source type data covered by the raw data, and interference removal processing is performed on the source historical data in the raw data to obtain interference-free source historical data. Based on the standard data for debris flow source hazards, the key source type data, and the interference-free source historical data, X debris flow attribute data corresponding to the debris flow source dataset to be identified are obtained.

[0098] Step S102: Extract features from each debris flow attribute data to obtain the data monitoring features corresponding to each debris flow attribute data, and obtain debris flow source attribute matching data between the data monitoring features corresponding to each debris flow attribute data.

[0099] For example, different feature extraction methods can be used for different debris flow attribute data. For instance, key source type data, historical data of interference removal sources, and standard data of debris flow source hazards can be extracted using different feature extraction methods. Alternatively, different feature extraction methods can be used to extract features from the same debris flow attribute data. For instance, different feature extraction methods can be used to extract features from historical data of interference removal sources to obtain different types of features in the historical data of interference removal sources. This application does not limit the feature extraction method for each debris flow attribute data.

[0100] The data monitoring features corresponding to the standard data on debris flow source hazards can include one or more features such as the description features of the mountain event and the basic features of the mountain event. This application does not limit the types of data monitoring features corresponding to the standard data on debris flow source hazards, nor the feature extraction method of the standard data on debris flow source hazards.

[0101] For the standard data on received debris flow source hazards from X debris flow attribute data, the same feature extraction method as that used for sending debris flow source hazard standard data can be employed to extract features from the received debris flow source hazard standard data, thereby obtaining the data monitoring features corresponding to the received debris flow source hazard standard data. These data monitoring features may include one or more features such as features describing the received mountain event and basic features of the received mountain event. This application does not limit the types of data monitoring features corresponding to the received debris flow source hazard standard data, nor the feature extraction method for the received debris flow source hazard standard data.

[0102] After obtaining the data monitoring features corresponding to X debris flow attribute data, debris flow source attribute matching data can be obtained between the data monitoring features corresponding to each debris flow attribute data. This debris flow source attribute matching data can be used to characterize the correlation between different debris flow attribute data, such as indicating whether there is a correlation between each debris flow attribute data.

[0103] Step S103: Based on the debris flow source attribute matching data, perform feature integration processing on the data monitoring features corresponding to each debris flow attribute data to obtain the global data monitoring features corresponding to the debris flow source data set to be identified.

[0104] Among them, the core role of debris flow source attribute matching data is that this matching data is equivalent to a "screening standard"—telling us which monitoring features are strongly correlated with source attributes and are valuable (such as features related to source stability and susceptibility), and which are irrelevant and can be eliminated to avoid interference from useless features.

[0105] Operation: Perform feature integration processing on the "data monitoring features" corresponding to each debris flow attribute data;

[0106] Each debris flow attribute data point is the same as the original attribute data from the previous step (slope, water content, source volume, etc.) and its corresponding monitoring characteristics (e.g., data A corresponds to "looseness characteristics", data B corresponds to "seepage characteristics", and data C corresponds to "rainfall response characteristics").

[0107] Feature integration processing: It's not just about piling things up, but about doing three things (fitting the mudslide scenario):

[0108] ① Filtering: Based on the matching data, retain useful features related to the source attributes and delete irrelevant features (such as deleting "mudslide flow velocity features" that are irrelevant to the source).

[0109] ② Integration: Merge multiple scattered and repetitive features (for example, "rock mass fracture features" and "rock mass loosening features" are both related to source stability and can be integrated into "source rock mass stability comprehensive features").

[0110] ③ Normalization: Unify the standard of all features (for example, some features are in "meters" and some are in "percentages", and convert them into a comparable and integrated format).

[0111] Result: We obtained the "global data monitoring features corresponding to the dataset of debris flow sources to be identified".

[0112] The set of debris flow source data to be identified: not a single debris flow data, but a group or batch of debris flow data that needs to be identified (such as identifying the source type and risk level);

[0113] Global data monitoring features: The integrated "unified core features" represent the overall characteristics of this batch of sources to be identified, such as "overall stability features of sources", "comprehensive features of sources that are easy to start" and "comprehensive features of sources that are sensitive to seepage". This is equivalent to summing up the scattered individual features into "general features" that can reflect the "entire source set", which facilitates the subsequent identification, early warning and classification of the entire source set.

[0114] For example, after obtaining the data monitoring features corresponding to X debris flow attribute data, feature integration processing can be performed on the data monitoring features corresponding to each debris flow attribute data based on the debris flow source attribute matching data between the data monitoring features corresponding to each debris flow attribute data, to obtain the global data monitoring features corresponding to the set of debris flow source data to be identified; the global data monitoring features can be considered as the result of feature selection of the data monitoring features corresponding to X debris flow attribute data. The feature integration processing involved in the embodiments of this application may include, but is not limited to, one or more of the following methods: feature fusion, feature integration based on attention mechanism, multi-debris source attribute processing, feature integration based on kernel function, weighted integration, and feature integration based on deep learning, etc., and this application does not limit them.

[0115] For example, if the debris flow source attribute matching data indicates a correlation between the monitoring feature corresponding to the *a*th debris flow attribute data and the monitoring feature corresponding to the *b*th debris flow attribute data among X debris flow attribute data, then the monitoring features corresponding to the *a*th and *b*th debris flow attribute data can be fused to obtain a combined debris flow source attribute feature. The *a*th and *b*th debris flow attribute data can be any two debris flow attribute data from the X debris flow attribute data, where *a* and *b* are both positive integers not greater than X. Processing the combined debris flow source attribute feature and the monitoring feature corresponding to each debris flow attribute data yields the global monitoring feature corresponding to the set of debris flow source data to be identified. It is understandable that if the debris flow source attribute matching data indicates a correlation between the data monitoring features corresponding to the three debris flow attribute data, then the data monitoring features corresponding to these three debris flow attribute data can be fused to obtain the debris flow source attribute combination feature; this application does not limit the number of data monitoring features during feature fusion.

[0116] Optionally, if the debris flow source attribute matching data indicates a correlation between the data monitoring feature corresponding to the a-th debris flow attribute data and the data monitoring feature corresponding to the b-th debris flow attribute data, then the data monitoring feature corresponding to the a-th debris flow attribute data and the data monitoring feature corresponding to the b-th debris flow attribute data can be combined into a multi-debris flow source attribute fusion feature. This multi-debris flow source attribute fusion feature is then loaded into a multi-debris flow source attribute convolutional unit, and bidirectional feature convolution processing is performed on the multi-debris flow source attribute fusion feature through this multi-debris flow source attribute convolutional unit to obtain a combined debris flow source attribute feature. Furthermore, the combined debris flow source attribute feature and the data monitoring feature corresponding to each debris flow attribute data can be processed to obtain the global data monitoring feature corresponding to the set of debris flow source data to be identified.

[0117] Step S104: Analyze and process the global data monitoring features to obtain the analysis results corresponding to the data set of debris flow sources to be identified.

[0118] The operation target: global data monitoring features are the "unified core features" that were integrated in the previous step and represent the entire batch of debris flow sources to be identified. They are not scattered individual features, but a collection that can reflect the overall characteristics of the entire batch of sources (such as: the average looseness of the entire batch of sources, overall seepage sensitivity, comprehensive stability, etc.).

[0119] Operational Actions: Analysis and Processing. Here, "analysis and processing" doesn't simply mean looking at the data, but rather performing three key processing steps (tailored to practical application scenarios) based on the need to identify debris flow sources:

[0120] Feature Interpretation: Determine the specific meaning of each global feature (e.g., "high global looseness feature value" can be interpreted as "the entire source rock mass is loose and easily stirred up by rainwater").

[0121] Correlation analysis: Examine the correlation between global features (for example, the simultaneous presence of "high overall seepage sensitivity" and "low overall stability" indicates that the entire batch of material sources is prone to seepage failure, which in turn triggers debris flow).

[0122] Threshold judgment / classification: Based on the safety standards and classification specifications of debris flow sources, determine whether the global characteristics meet the standards (e.g., "the global easy-to-start characteristics exceed the safety threshold", indicating that the entire batch of sources belongs to high-risk sources).

[0123] Simply put, it means translating abstract "feature data" into information that can be understood and applied.

[0124] Final result: Analysis results corresponding to the dataset of debris flow sources to be identified.

[0125] The "analysis results" here refer to concrete and practically valuable conclusions, not abstract data. The core is answering three key questions (in conjunction with the need for debris flow source identification):

[0126] ① Overall characteristics of the entire batch of sources to be identified: for example, "the entire batch of sources is mainly composed of loose rock, which is sensitive to seepage and has poor stability";

[0127] ② Risk level of the entire batch of materials: For example, "Global features show that this batch of materials is a high-risk material source, which is easily triggered by rainfall to form debris flows";

[0128] ③ Source classification / identification conclusion: For example, "Based on global feature comparison, this batch of sources belongs to 'loosely deposited source,' which is the main source of mudslide replenishment."

[0129] Simply put, it means processing the "raw material" of "overall characteristics" to arrive at the "final conclusion" of "what the entire batch of materials is and what risks it carries".

[0130] For example, after obtaining the global data monitoring features corresponding to the debris flow source data set to be identified, these global data monitoring features can be loaded into the source data set data analysis network. The source data set data analysis network then performs forward computation on the global data monitoring features, outputting the analysis results corresponding to the debris flow source data set to be identified. Specifically, the source data set data analysis network is trained using an example source data set and the category catalog it carries; the input data to the source data set data analysis network is the global data monitoring features corresponding to the debris flow source data set to be identified.

[0131] After obtaining the data monitoring features corresponding to each debris flow attribute data, feature selection can be performed on these monitoring features. For example, feature fusion can be performed on the monitoring features corresponding to related debris flow attribute data to obtain combined debris flow source attribute features. Furthermore, the combined debris flow source attribute features and the monitoring features corresponding to each debris flow attribute data can be processed using multi-debris source attribute methods to obtain global data monitoring features corresponding to the debris flow source data set to be identified. These global data monitoring features are then loaded into a source data set analysis network. This network performs forward computation on the global data monitoring features to obtain the analysis results corresponding to the debris flow source data set to be identified.

[0132] This application provides a flowchart illustrating a debris flow source monitoring method. It is understood that this debris flow source monitoring method can be executed by a computer device, which can be a server or a terminal device; this application does not limit this. The debris flow source monitoring method may include the following steps S201 to S209:

[0133] Step S201: Obtain X debris flow attribute data corresponding to the debris flow source data set to be identified.

[0134] Step S202: Convert the key material source type data in the X debris flow attribute data into material source weight features, perform convolution processing on the material source weight features to obtain the hidden danger weight features, and use the material source weight features and the hidden danger weight features as the data monitoring features corresponding to the key material source type data.

[0135] Step S203: Perform convolution processing on the historical data of the removed debris flow source in the X debris flow attribute data to obtain the source factor features corresponding to the historical data of the removed debris flow source.

[0136] In this embodiment, the historical source data of the undisturbed debris flow attribute data can be loaded into a source data convolutional unit. By performing convolution processing on the historical source data of the undisturbed debris flow attribute data through the source data convolutional unit, the source factor features corresponding to the historical source data of the undisturbed debris flow attribute data can be obtained. The source data convolutional unit can be any of the following: convolutional neural network, residual network, graph neural network, or other deep learning network.

[0137] Step S204: Analyze the historical data of the interference removal source to obtain the feature analysis results. Use the source factor characteristics and the feature analysis results as the data monitoring characteristics corresponding to the historical data of the interference removal source.

[0138] For example, a computer device can perform data detection on historical data of the interference source to obtain data detection results, clean the historical data of the interference source based on the data detection results to obtain cleaning results, classify the cleaning results to obtain classification results, analyze the classification results to obtain feature analysis results corresponding to the historical data of the interference source, and identify these feature analysis results to convert them into the content covered in the historical data of the interference source.

[0139] Step S205: Obtain the confidence integration information of the debris flow source hazard standard data in X debris flow attribute data, and perform convolution processing on the confidence integration information and the debris flow source hazard standard data to obtain the data monitoring features corresponding to the debris flow source hazard standard data.

[0140] Step S206: Obtain debris flow source attribute matching data between the data monitoring features corresponding to each debris flow attribute data. Based on the debris flow source attribute matching data, perform feature integration processing on the data monitoring features corresponding to each debris flow attribute data to obtain the global data monitoring features corresponding to the debris flow source data set to be identified.

[0141] Step S207: Load the global data monitoring features into the first data analysis network, and output the first regression analysis results corresponding to the debris flow source data set to be identified through the first data analysis network.

[0142] In this embodiment of the application, after feature selection in step S206 above, global data monitoring features corresponding to the debris flow source data set to be identified can be obtained. These global data monitoring features can be input into the source data set data analysis network vehicle. Through the source data set data analysis network, the analysis results corresponding to the debris flow source data set to be identified can be determined. The number of source data set data analysis networks can be one or more, and this application does not limit this.

[0143] For ease of understanding, the following description assumes that there are two data analysis networks for the source dataset. In this case, the data analysis networks can include a first data analysis network and a second data analysis network. The first and second data analysis networks are trained for the same source dataset analysis and processing task. For example, the output layer dimension of the first data analysis network is the same as the output layer dimension of the second data analysis network. The first and second data analysis networks can have different network structures; this application does not limit the specific network structures of the first and second data analysis networks.

[0144] The global data monitoring feature can be loaded into the first data analysis network. The first data analysis network performs forward computation on this global data monitoring feature, and the regression analysis result output by the output layer of the first data analysis network is called the first regression analysis result. This first regression analysis result can be a vector with a dimension equal to the number of categories in the debris flow source data set that the first data analysis network can identify. The numerical value in the first regression analysis result can be used to represent the probability that the debris flow source data set to be identified belongs to each analysis result.

[0145] Step S208: Load the global data monitoring features into the second data analysis network, and output the second regression analysis results corresponding to the debris flow source data set to be identified through the second data analysis network.

[0146] For example, global data monitoring features can be input into a second data analysis network. This network performs forward computation on these features, and the regression analysis result output by the output layer of the second data analysis network is called the second regression analysis result. This second regression analysis result can be a vector with a dimension equal to the number of categories in the debris flow source data set identifiable by the second data analysis network. The numerical value in the second regression analysis result can represent the probability that the debris flow source data set to be identified belongs to each analysis result. It should be understood that the first and second regression analysis results can be vectors with the same dimension.

[0147] Step S209: Perform function processing on the first regression analysis results and the second regression analysis results to obtain the analysis results corresponding to the debris flow source data set to be identified.

[0148] This application provides an embodiment of a schematic diagram of debris flow source data set analysis and processing. After obtaining the global data monitoring features corresponding to the debris flow source data set to be identified, the global data monitoring features can be sequentially loaded into a first data analysis network and a second data analysis network. By performing forward calculation on the global data monitoring features through the first data analysis network, a first regression analysis result of the debris flow source data set to be identified can be obtained. By performing forward calculation on the global data monitoring features through the second data analysis network, a second regression analysis result of the debris flow source data set to be identified can be obtained.

[0149] In this embodiment, multiple feature extractors can be used to extract different types of data monitoring features corresponding to the same debris flow attribute data. This allows for the acquisition of data monitoring features for each debris flow attribute data, improving the diversity of these features. Based on the matching data between the data monitoring features corresponding to each debris flow attribute data, feature integration processing is performed on these features to obtain global data monitoring features. This enhances the characterization capability of the global data monitoring features for the debris flow source data set to be identified. By inputting this global data monitoring feature into a source data set analysis network, and then analyzing and processing this global data monitoring feature through multiple source data set analysis networks, and combining the regression analysis results of multiple source data set analysis networks, the analysis results of the source data set to be identified can be determined, which can improve the accuracy of source data set analysis and processing. Since the input data of the source data set analysis network can include multiple debris flow attribute data, that is, not limited to data of a specific debris flow source attribute, it can be applied to all source data sets, improving the coverage of source data set analysis and processing. After the source data set analysis network is online, the analysis and processing of all source data sets does not require manual intervention, reducing the operating costs of the source data set management platform.

[0150] This application provides a flowchart illustrating a debris flow source monitoring method. It is understood that this debris flow source monitoring method can be executed by a computer device, which can be a server or a terminal device; this application does not limit this. The debris flow source monitoring method may include the following steps S301 to S305:

[0151] Step S301: Obtain the attribute data of X sample debris flows corresponding to the sample material source data set.

[0152] The processing procedure during network training is similar for all exemplary source data sets in the exemplary dataset. For ease of understanding, this embodiment uses any one exemplary source data set in the exemplary dataset as an example. X exemplary debris flow attribute data corresponding to the exemplary source data set can be obtained. The process of obtaining the X exemplary debris flow attribute data corresponding to this exemplary source data set, and the process of obtaining the X debris flow attribute data corresponding to the debris flow source data set to be identified in step S101, will not be elaborated here.

[0153] Step S302: Extract features from the debris flow attribute data of each example to obtain the example data monitoring features corresponding to each example debris flow attribute data, and obtain feature matching data between the example data monitoring features corresponding to each example debris flow attribute data.

[0154] Step S303: Based on the feature matching data, perform feature integration processing on the example data monitoring features corresponding to each example debris flow attribute data to obtain the global debris flow source attribute example features corresponding to the example source data set.

[0155] In particular, the feature integration processing of the example data monitoring features corresponding to each example debris flow attribute data in step S303, and the feature integration processing of the data monitoring features corresponding to each debris flow attribute data in step S103; that is to say, the process of obtaining the global debris flow source attribute example features corresponding to the example source data set is similar to the process of obtaining the global data monitoring features corresponding to the debris flow source data set to be identified, and will not be described again here.

[0156] Step S304: Load the global debris flow source attribute example features into the pre-trained network, and analyze and process the global debris flow source attribute example features through the pre-trained network to obtain the example category corresponding to the example source data set.

[0157] Step S305: Based on the anomalies between the example category and the example directory corresponding to the example source data set, optimize the network coefficients of the pre-trained network, and use the pre-trained network covering the optimized network coefficients as the data analysis network for the source data set.

[0158] For example, after obtaining the example categories corresponding to the example source dataset through a pre-trained network, anomalies between the example categories and the example directories corresponding to the example source dataset can be calculated. Based on these anomalies, the loss of the pre-trained network can be calculated. The network coefficients of the pre-trained network can be iteratively trained based on this loss. When the loss of the pre-trained network meets the training termination condition, training can be stopped, and the pre-trained network at the point of training termination can be used as the trained data analysis network for the source dataset. For example, the data analysis network for the source dataset can be the first data analysis network mentioned above, or it can be the second data analysis network, or it can be both the first and second data analysis networks.

[0159] In this embodiment, some open-source databases can be used for pre-training to obtain a pre-trained network. Then, in conjunction with the material source data set analysis and processing task, an example dataset carrying an example catalog is collected. The network coefficients of the pre-trained network are fine-tuned using the example dataset to obtain the material source data set analysis network for obtaining material source hazard analysis results. This can reduce the catalog data in the network training stage, thereby reducing the training cost of the material source data set analysis network.

[0160] Based on the above, a debris flow source monitoring 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.

[0161] Based on the above, a computer-readable storage medium is also provided, on which a computer program stored implements the above method during runtime.

[0162] In summary, based on the above scheme, for each debris flow source data set to be identified, X debris flow attribute data corresponding to that data set can be obtained. Feature extraction is performed on each debris flow attribute data to obtain the corresponding data monitoring features. Then, based on the matching data between the data monitoring features corresponding to each debris flow attribute data, feature integration processing is performed on the data monitoring features corresponding to each debris flow attribute data to obtain the global data monitoring features corresponding to the debris flow source data set to be identified. Analyzing and processing these global data monitoring features yields the analysis results for the debris flow source data set to be identified. In the task of analyzing and processing source data sets, combining the data monitoring features corresponding to multiple debris flow attribute data in the debris flow source data set to be identified can yield more representative global data monitoring features. These global data monitoring features can accurately obtain the analysis results of source hazard issues, thus improving the accuracy of source data set analysis and processing.

[0163] 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).

[0164] 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 monitoring the source of debris flows, characterized in that, include: Obtain X debris flow attribute data corresponding to the debris flow source data set to be identified; among which, the X debris flow attribute data include six categories: source type, spatial distribution, geometric morphology, physical mechanics, reserves and recharge, and dynamic evolution; Feature extraction is performed on each debris flow attribute data to obtain the data monitoring features corresponding to each debris flow attribute data, and debris flow source attribute matching data between the data monitoring features corresponding to each debris flow attribute data is obtained. Based on the debris flow source attribute matching data, feature integration processing is performed on the data monitoring features corresponding to each debris flow attribute data to obtain the global data monitoring features corresponding to the debris flow source data set to be identified. The global data monitoring features are analyzed and processed to obtain the analysis results corresponding to the debris flow source data set to be identified; The acquisition of X debris flow attribute data corresponding to the data set of debris flow sources to be identified includes: Obtain the range of debris flow source attribute types in the debris flow source monitoring command, and based on the debris flow source attribute types within the range of debris flow source attribute types, parse the debris flow source data set to be identified to obtain the original data that matches the debris flow source attribute types within the range of debris flow source attribute types. The original data is filtered to obtain X debris flow attribute data corresponding to the debris flow source data set to be identified.

2. The method as described in claim 1, characterized in that, The process of filtering the raw data yields X debris flow attribute data corresponding to the debris flow source data set to be identified, including: The quality of the mountain-related information in the original data is verified to obtain the quality verification results of the mountain-related information. If the quality verification result of the mountain matter indicates that the verification is qualified, the mountain matter information will be standardized to obtain standard data of debris flow source hazards; Key source type data are obtained from the source type data covered by the original data, and the source historical data in the original data is subjected to interference removal processing to obtain interference-removed source historical data. Based on the standard data of debris flow source hazards, the key source type data, and the historical data of the de-interference source, X debris flow attribute data corresponding to the data set of debris flow sources to be identified are obtained.

3. The method as described in claim 1, characterized in that, The X debris flow attribute data include key source type data, historical data of de-interferenced source materials, and standard data on debris flow source material hazards. The step of extracting features from each debris flow attribute data to obtain the data monitoring features corresponding to each debris flow attribute data includes: The key material source type data is transformed into material source weight features. The material source weight features are then convolved to obtain hidden danger weight features. The material source weight features and the hidden danger weight features are used as data monitoring features corresponding to the key material source type data. The source historical data of the de-interference material is subjected to convolution processing to obtain the source factor features corresponding to the source historical data of the de-interference material; The historical data of the interference removal source is analyzed to obtain the feature analysis results. The source factor features and the feature analysis results are used as the data monitoring features corresponding to the historical data of the interference removal source. The confidence level integration information associated with the standard data of debris flow source hazards is obtained, and the confidence level integration information and the standard data of debris flow source hazards are convolved to obtain the data monitoring features corresponding to the standard data of debris flow source hazards.

4. The method as described in claim 3, characterized in that, The process of converting the key source type data into source weight features includes: The key material source type data is transformed into W hazard factors to obtain the hazard vectors corresponding to the W hazard factors; wherein, the W hazard factors include real-time rainfall hazard factors, accumulated material hazard factors, and blockage hazard factors; Based on the important information of the W hidden danger factors in the key material source type data, the feature vectors corresponding to the W hidden danger factors are obtained. Based on the location of the hidden danger areas in the key material source type data using the W hidden danger factors, the location vectors corresponding to the W hidden danger factors are obtained. The hazard vector, the feature vector, and the location vector are combined to obtain the material source weight feature corresponding to the key material source type data.

5. The method as described in claim 3, characterized in that, The step of performing convolution processing on the historical data of the de-interference source to obtain the source factor features corresponding to the historical data of the de-interference source includes: The historical data of the de-interference material source is loaded into the material source data convolutional unit to obtain the input features of the m-th feature extraction module in the material source data convolutional unit; when m is 1, the input features of the m-th feature extraction module are the historical data of the de-interference material source; the m features include: real-time rainfall features, sediment features and blockage features; Based on one or more feature extraction layers in the m-th feature extraction module, the input features of the m-th feature extraction module are processed to obtain the processing result; Based on the weight coefficients corresponding to the simplification layer in the m-th feature extraction module, the processing result is simplified to obtain simplified features. The simplified features and the input features of the m-th feature extraction module are combined to obtain the output features of the m-th feature extraction module. The output features of the Y-th feature extraction module in the source data convolution unit are used as the source factor features corresponding to the de-interference source historical data.

6. The method as described in claim 3, characterized in that, The analysis of the historical data of the interference source to obtain feature analysis results includes: Data detection is performed on the historical data of the interference removal source to obtain data detection results. Based on the data detection results, the historical data of the interference removal source is cleaned to obtain cleaning results. The cleaning results are classified to obtain classification results. The classification results are then analyzed to obtain feature analysis results corresponding to the historical data of the interference source.

7. The method as described in claim 1, characterized in that, The step involves integrating the data monitoring features corresponding to each debris flow attribute data point based on the debris flow source attribute matching data to obtain the global data monitoring features corresponding to the debris flow source data set to be identified, including: If the debris flow source attribute matching data indicates that there is a correlation between the data monitoring feature corresponding to the a-th debris flow attribute data and the data monitoring feature corresponding to the b-th debris flow attribute data among the X debris flow attribute data, then the debris flow attribute data corresponding to the a-th debris flow attribute data and the data monitoring feature corresponding to the b-th debris flow attribute data are fused to obtain a debris flow source attribute combination feature; the remaining a debris flow attribute data are sample data in the database, and the b debris flow attribute data are real-time attribute data, representing the debris flow attribute data corresponding to the current object; The combined features of debris flow source attributes and the data monitoring features corresponding to each debris flow attribute data are processed to obtain the global data monitoring features corresponding to the set of debris flow source data to be identified. The step of analyzing and processing the global data monitoring features to obtain the analysis results corresponding to the debris flow source data set to be identified includes: The global data monitoring features are loaded into the first data analysis network, and the first regression analysis results corresponding to the debris flow source data set to be identified are output through the first data analysis network. The global data monitoring features are loaded into the second data analysis network, and the second regression analysis results corresponding to the data set of debris flow sources to be identified are output through the second data analysis network. The results of the first regression analysis and the second regression analysis are processed by a function to obtain the analysis results corresponding to the data set of debris flow sources to be identified.

8. The method according to any one of claims 1 to 7, characterized in that, The method further includes: Obtain the attribute data of X typical debris flows corresponding to the typical material source data set; Feature extraction is performed on the debris flow attribute data of each example to obtain the example data monitoring features corresponding to each example debris flow attribute data, and feature matching data between the example data monitoring features corresponding to each example debris flow attribute data is obtained. Based on the feature matching data, feature integration processing is performed on the example data monitoring features corresponding to each example debris flow attribute data to obtain the global debris flow source attribute example features corresponding to the example source data set. The global debris flow source attribute example features are loaded into a pre-trained network, and the global debris flow source attribute example features are analyzed and processed through the pre-trained network to obtain the example category corresponding to the example source data set. Based on the anomalies between the example category and the example directory corresponding to the example source data set, the network coefficients of the pre-trained network are optimized, and the pre-trained network covering the optimized network coefficients is used as the source data set data analysis network; the source data set data analysis network includes at least one of a first data analysis network and a second data analysis network.

9. A debris flow source monitoring 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