Debris flow source identification method and system

Abstract

The invention discloses a debris flow source identification method and system, and the method comprises the steps: obtaining at least one second source attribute feature of to-be-identified debris flow source information, building the feature description of the to-be-identified debris flow source information on a plurality of source attribute data volumes, and carrying out the recognition of the to-be-identified debris flow source information according to the at least one second source attribute feature. And determining directory information of each byte in the debris flow source information to be identified. According to the embodiment of the invention, the cleaning unit adapted to different object source attribute data volumes is dynamically generated according to the to-be-identified debris flow object source information, and the weight of the cleaning unit is dynamically generated according to the to-be-identified debris flow object source information, so that the adaptability is better; therefore, the feature description of the established debris flow source information to be identified on a plurality of source attribute data volumes is more accurate, and the accuracy of the directory information of each byte in the determined debris flow source information to be identified is higher.

Description

Technical Field

[0001] This disclosure relates to the technical field of landslide layer structure identification, and in particular to a method and system for identifying the source of debris flows. Background Technology

[0002] Debris flows, a common and sudden geological disaster in mountainous areas, depend on a sufficient supply of debris. The type, quantity, distribution, and stability of the debris source directly determine the scale, intensity, and severity of the debris flow. Accurately identifying debris flow source information (such as material composition, spatial distribution, and potential recharge areas) is a core prerequisite for understanding the formation mechanism of debris flows, assessing disaster risks, and formulating prevention and control measures. It is crucial for improving the foresight and effectiveness of debris flow disaster prevention and control.

[0003] With the rapid development of technologies such as remote sensing, UAV aerial surveying, and ground-penetrating radar, debris flow source investigation has shifted from traditional ground exploration to multi-source data fusion analysis, resulting in massive amounts of source attribute data (such as topographic data, stratigraphic lithology data, and vegetation cover data). However, current debris flow source identification technology still faces many technical bottlenecks, making it difficult to meet the needs of accurate identification in complex scenarios.

[0004] Therefore, how to overcome technical challenges such as insufficient dynamic adaptation, inefficient data cleaning, and weak multi-scale feature fusion, and construct an identification method based on dynamic adjustment of material source attributes to achieve accurate feature extraction, efficient cleaning, and standardized identification of debris flow material source information has become a core issue that urgently needs to be addressed in the field of debris flow disaster prevention and control. It is also a key technical support for promoting the transformation of material source identification technology from "fixed mode" to "dynamic intelligence". Summary of the Invention

[0005] To address the technical problems existing in related technologies, this disclosure provides a method and system for identifying the source of debris flows.

[0006] A method for identifying the source of debris flows, comprising: The first source attribute features for obtaining the source information of the debris flow to be identified; Based on the first source attribute characteristics and at least one pre-set cleaning window, at least one cleaning unit is generated respectively. The cleaning unit is used to build a feature description of the different source attribute data volume corresponding to the debris flow source information to be identified. Based on the first source attribute feature and the at least one cleaning unit, at least one second source attribute feature of the debris flow source information to be identified is obtained; Based on the at least one second source attribute feature, determine the directory information of each byte in the debris flow source information to be identified; The step of determining the directory information of each byte in the debris flow source information to be identified based on the at least one second source attribute feature includes: Based on the first source attribute feature and the at least one second source attribute feature, determine the directory information of each byte in the debris flow source information to be identified; The step of generating at least one cleaning unit based on the first material source attribute features and at least one pre-set cleaning window specifically includes: extracting range material source blockage information from the first material source attribute features according to the pre-set cleaning window, and generating the cleaning unit based on the range material source blockage information.

[0007] In one independently implemented embodiment, the larger the pre-set cleaning window, the greater the amount of debris flow source information to be identified generated by the cleaning unit, and the greater the feature description of the source attribute data.

[0008] In one standalone embodiment, the weights of the cleaning units are generated based on the first source attribute characteristics.

[0009] In one independently implemented embodiment, obtaining at least one second source attribute feature of the debris flow source information to be identified based on the first source attribute feature and the at least one cleaning unit includes: The first source attribute features after convolution are simplified to obtain the simplified first source attribute features. The simplified first source attribute feature is subjected to depth convolution with the at least one cleaning unit to obtain at least one third source attribute feature. The at least one third source attribute feature is convolved to obtain at least one second source attribute feature of the debris flow source information to be identified.

[0010] In one standalone implementation, the method is implemented through an identification model, and the method further includes: configuring the identification model according to a pre-set configuration set, the configuration set including: at least one example source attribute and the original directory information of each byte in the example source attribute; The recognition model includes a feature extraction module, at least one dynamic cleaning unit generation module, and a splicing module. Configuring the recognition model according to a pre-set configuration set includes: At least one example source attribute is loaded into the feature extraction module to obtain the first source attribute feature; The first source attribute feature is loaded into the at least one dynamic cleaning unit generation module to obtain at least one cleaning unit, and at least one second source attribute feature of the example source attribute is obtained according to the at least one cleaning unit and the first source attribute feature respectively. Load the at least one second source attribute feature into the splicing module to determine the first directory information of each byte in the example source attribute; Based on the original directory information of each byte and the first directory information of each byte in the example source attribute, the generation quantification index of the recognition model is determined; Configure the recognition model based on the generated quantitative indicators.

[0011] In one standalone implementation, the feature extraction module is pre-configured to obtain a network for extracting source attribute features, and / or the splicing module is a pre-configured recognition model for identifying byte directories.

[0012] The technical solutions provided by the embodiments of this disclosure may include the following beneficial effects.

[0013] This application discloses a method and system for identifying the source of debris flows. After obtaining a first source attribute feature of the debris flow source information to be identified, at least one cleaning unit with different convolutional window sizes can be generated based on the first source attribute feature and at least one pre-set cleaning window. Based on the at least one cleaning unit and the first source attribute feature, at least one second source attribute feature of the debris flow source information to be identified is obtained, thereby constructing a feature description of the debris flow source information on multiple source attribute data volumes. Then, based on the at least one second source attribute feature, the directory information of each byte in the debris flow source information to be identified is determined. The embodiments of this disclosure dynamically generate cleaning units adapted to different source attribute data volumes based on the debris flow source information to be identified. The weights of these cleaning units are dynamically generated based on the debris flow source information to be identified, resulting in better adaptability. Consequently, the feature description of the debris flow source information to be identified on multiple source attribute data volumes is more accurate, leading to higher accuracy in determining the directory information of each byte in the debris flow source information to be identified.

[0014] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this disclosure. Attached Figure Description

[0015] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the specification, serve to explain the principles of this application.

[0016] Figure 1A flowchart illustrating a debris flow source identification method provided in an embodiment of this application; Detailed Implementation

[0017] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0018] Based on the above, please refer to the following: Figure 1 This is a flowchart illustrating a debris flow source identification method provided in an embodiment of this application. Further, a debris flow source identification method may specifically include the content described in steps S11-S14.

[0019] In step S11, the first source attribute features of the debris flow source information to be identified are obtained.

[0020] For example, a pre-configured convolutional recognition model can be used to extract source attribute features from the debris flow source information to be identified, thereby obtaining the first source attribute features of the debris flow source information to be identified. The configuration process of the convolutional recognition model will be described in detail in the embodiments corresponding to the network configuration method, and will not be repeated here.

[0021] In step S12, at least one cleaning unit is generated based on the first material source attribute characteristics and at least one pre-set cleaning window.

[0022] For example, the pre-set cleaning window can be the size of the convolution window for the cleaning units to be generated. Different pre-set cleaning windows can generate cleaning units with different convolution window sizes. Range-based source blockage information from the first source attribute features can be dynamically extracted based on the pre-set cleaning window, and cleaning units can be generated based on this range-based source blockage information. In other words, the weights of the cleaning units are generated based on the first source attribute features. When multiple pre-set cleaning windows exist, range-based source blockage information from the first source attribute features can be extracted separately for each pre-set cleaning window, and multiple cleaning units can be generated based on the extracted range-based source blockage information.

[0023] In one possible implementation, different pre-set cleaning windows correspond to different amounts of source attribute data. That is, multiple cleaning units generated according to different pre-set cleaning windows can be used to build feature descriptions of the debris flow source information to be identified corresponding to different amounts of source attribute data. In this way, cleaning units with different convolution window sizes can be generated through different pre-set cleaning windows, and feature descriptions of the debris flow source information to be identified on different amounts of source attribute data can be built through cleaning units with different convolution window sizes.

[0024] In step S13, at least one second source attribute feature of the debris flow source information to be identified is obtained based on the first source attribute feature and the at least one cleaning unit.

[0025] For example, the first source attribute feature can be processed by at least one generated cleaning unit to construct a second source attribute feature corresponding to at least one source attribute data volume of debris flow source information to be identified.

[0026] In step S14, the directory information of each byte in the debris flow source information to be identified is determined based on the at least one second source attribute feature.

[0027] For example, the feature descriptions (second source attribute features) of the debris flow source information to be identified, built by various cleaning units, on different source attribute data volumes can be combined to determine the directory information of each byte in the debris flow source information to be identified.

[0028] In one possible implementation, the above-described determination of the directory information of each byte in the debris flow source information to be identified based on the at least one second source attribute feature may include: Based on the first source attribute feature and at least one second source attribute feature, the directory information of each byte in the debris flow source information to be identified is determined.

[0029] For example, the aforementioned first source attribute features and at least one second source attribute feature can be integrated. Based on the integrated source attribute features, the directory information of each byte of the debris flow source information to be identified can be identified, thus obtaining the directory information of each byte in the debris flow source information to be identified. In this way, a cleaning unit adapted to different source attribute data volumes is dynamically generated based on the debris flow source information to be identified. The weight of this cleaning unit is dynamically generated based on the debris flow source information to be identified, resulting in better adaptability. Consequently, the feature description of the debris flow source information to be identified across multiple source attribute data volumes is more accurate, leading to higher accuracy of the directory information of each byte in the debris flow source information to be identified determined based on the second source attribute feature.

[0030] In one possible implementation, the above-described determination of the directory information of each byte in the debris flow source information to be identified based on the first source attribute feature and the at least one second source attribute feature may include: The at least one second source attribute feature is integrated with the first source attribute feature to obtain the integrated source attribute feature; The integrated material source attribute features are subjected to convolution processing to obtain the directory information of each byte in the debris flow material source information to be identified.

[0031] For example, after constructing a feature description (second source attribute feature) of the debris flow source information to be identified on multiple source attribute data volumes by generating multiple cleaning units and the first source attribute feature, the second source attribute feature can be integrated with the first source attribute feature. The integrated source attribute feature is then subjected to convolution processing (e.g., using a pre-configured identification model for identifying byte directory information to perform convolution processing on the integrated source attribute feature) to obtain the directory information of each byte in the debris flow source information to be identified.

[0032] In this way, after obtaining the first source attribute feature of the debris flow source information to be identified, at least one cleaning unit with a different convolution window size can be generated based on the first source attribute feature and at least one pre-set cleaning unit. Based on the at least one cleaning unit and the first source attribute feature, at least one second source attribute feature of the debris flow source information to be identified is obtained, thereby constructing a feature description of the debris flow source information to be identified across multiple source attribute data volumes. Then, based on the at least one second source attribute feature, the directory information of each byte in the debris flow source information to be identified is determined. The source attribute processing method disclosed in this embodiment can dynamically generate cleaning units adapted to different source attribute data volumes based on the debris flow source information to be identified. The weights of these cleaning units are dynamically generated based on the debris flow source information to be identified, resulting in better adaptability. Consequently, the feature description of the debris flow source information to be identified across multiple source attribute data volumes is more accurate, leading to higher accuracy in the determined directory information of each byte in the debris flow source information to be identified.

[0033] In one possible implementation, generating at least one cleaning unit based on the first source property characteristics and at least one pre-set cleaning window may include: According to the at least one pre-set cleaning window, the first material source attribute features are pooled to obtain at least one material source blockage information; The at least one source blockage information is convolved to obtain the at least one cleaning unit.

[0034] In this way, the first source attribute feature description of debris flow source information of any size to be identified can be pooled into a feature description (source blockage information) of a specific size (pre-set cleaning window) through pooling operation. Cleaning units can be dynamically generated based on the source blockage information of debris flow source information to be identified, and the content of multiple source attribute data and feature descriptions of multiple source attribute data can be captured through the generated cleaning units of different sizes.

[0035] In one possible implementation, obtaining at least one second source attribute feature of the debris flow source information to be identified based on the first source attribute feature and the at least one cleaning unit may include: The first source attribute features after convolution are simplified to obtain the simplified first source attribute features. The simplified first source attribute feature is subjected to depth convolution with the at least one cleaning unit to obtain at least one third source attribute feature. The at least one third source attribute feature is convolved to obtain at least one second source attribute feature of the debris flow source information to be identified.

[0036] Cleaning units can be dynamically generated based on the debris flow source information to be identified, adapting to different amounts of source attribute data. The weights of these cleaning units are dynamically generated based on the debris flow source information to be identified, resulting in better adaptability and more accurate feature descriptions of the debris flow source information to be identified across multiple source attribute data volumes.

[0037] In one possible implementation, the convolution processing of the source blockage information and / or the first source attribute feature and / or the third source attribute feature is performed by convolution processing the source blockage information and / or the first source attribute feature and / or the third source attribute feature through a 1×1 convolution window.

[0038] In one possible implementation, the above-mentioned source attribute processing method can be implemented by an identification model. The method may further include: configuring the identification model according to a pre-set configuration set, wherein the configuration set includes: at least one example source attribute and the original directory information of each byte in the example source attribute.

[0039] For example, in this embodiment of the disclosure, an identification model can be pre-configured through a configuration set. This identification model can extract source attribute features from the debris flow source information to be identified, generate cleaning units corresponding to different source attribute data volumes based on the extracted source attribute features, and then construct feature descriptions of the debris flow source information to be identified on different source attribute data volumes based on the cleaning units. By assembling the feature descriptions of the debris flow source information to be identified on different source attribute data volumes, the directory information of each byte in the debris flow source information to be identified can be obtained.

[0040] Configuring the recognition model according to the pre-set configuration set may include: In step S15, at least one example source attribute is loaded into the feature extraction module to obtain the first source attribute feature.

[0041] In step S16, the first material source attribute feature is loaded into the at least one dynamic cleaning unit generation module to obtain at least one cleaning unit, and at least one second material source attribute feature of the example material source attribute is obtained according to the at least one cleaning unit and the first material source attribute feature respectively. In step S17, the at least one second source attribute feature is loaded into the splicing module to determine the first directory information of each byte in the example source attribute.

[0042] At least one example source attribute can be loaded into the feature extraction module to extract source attribute features. The output of the feature extraction module is the first source attribute feature of the example source attribute. The first source attribute feature is loaded into at least one dynamic cleaning unit generation module to generate at least one cleaning unit adapted to different source attribute data volumes.

[0043] For example, the dynamic cleaning unit generation module can dynamically extract range-based source blockage information from the first source attribute features according to a pre-set cleaning window, and generate cleaning units based on this range-based source blockage information. Multiple dynamic cleaning unit generation modules can each extract range-based source blockage information from the first source attribute features according to different pre-set cleaning windows, and then generate multiple cleaning units based on the extracted range-based source blockage information. The pre-set cleaning window is the pre-set convolution window size of the cleaning unit to be generated.

[0044] It should be understood that this application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for identifying the source of debris flows, characterized in that, include: The first source attribute features for obtaining the source information of the debris flow to be identified; Based on the first source attribute characteristics and at least one pre-set cleaning window, at least one cleaning unit is generated respectively. The cleaning unit is used to build a feature description of the different source attribute data volume corresponding to the debris flow source information to be identified. Based on the first source attribute feature and the at least one cleaning unit, at least one second source attribute feature of the debris flow source information to be identified is obtained; Based on the at least one second source attribute feature, determine the directory information of each byte in the debris flow source information to be identified; The step of determining the directory information of each byte in the debris flow source information to be identified based on the at least one second source attribute feature includes: Based on the first source attribute feature and the at least one second source attribute feature, determine the directory information of each byte in the debris flow source information to be identified; The step of generating at least one cleaning unit based on the first material source attribute features and at least one pre-set cleaning window specifically includes: extracting range material source blockage information from the first material source attribute features according to the pre-set cleaning window, and generating the cleaning unit based on the range material source blockage information.

2. The method according to claim 1, characterized in that, The larger the pre-set cleaning window, the greater the amount of debris flow source information to be identified in the generated cleaning unit, and the greater the feature description of the source attribute data.

3. The method according to claim 1, characterized in that, The weights of the cleaning units are generated based on the first material source attribute characteristics.

4. The method according to claim 1, characterized in that, The step of obtaining at least one second source attribute feature of the debris flow source information to be identified based on the first source attribute feature and the at least one cleaning unit includes: The first source attribute features after convolution are simplified to obtain the simplified first source attribute features. The simplified first source attribute feature is subjected to depth convolution with the at least one cleaning unit to obtain at least one third source attribute feature. The at least one third source attribute feature is convolved to obtain at least one second source attribute feature of the debris flow source information to be identified.

5. The method according to any one of claims 1-4, characterized in that, The method is implemented by an identification model, and the method further includes: configuring the identification model according to a pre-set configuration set, the configuration set including: at least one example source attribute and the original directory information of each byte in the example source attribute; The recognition model includes a feature extraction module, at least one dynamic cleaning unit generation module, and a splicing module. Configuring the recognition model according to a pre-set configuration set includes: At least one example source attribute is loaded into the feature extraction module to obtain the first source attribute feature; The first source attribute feature is loaded into the at least one dynamic cleaning unit generation module to obtain at least one cleaning unit, and at least one second source attribute feature of the example source attribute is obtained according to the at least one cleaning unit and the first source attribute feature respectively. Load the at least one second source attribute feature into the splicing module to determine the first directory information of each byte in the example source attribute; Based on the original directory information of each byte and the first directory information of each byte in the example source attribute, the generation quantification index of the recognition model is determined; Configure the recognition model based on the generated quantitative indicators.

6. The method according to claim 5, characterized in that, The feature extraction module is pre-configured to obtain a network for extracting source attribute features, and / or the splicing module is a pre-configured recognition model for recognizing byte directories.

7. A debris flow source identification system, characterized in that, It includes a processor and a memory that communicate with each other, the processor being used to read a computer program from the memory and execute it to implement the method of any one of claims 1-6.

Images