Multi-scale quantitative identification method for debris flow source

Abstract

The invention belongs to the technical field of information, and particularly discloses a debris flow source multi-scale quantitative identification method, which comprises the following steps: grouping regional spatial feature parameters to obtain a scale division threshold set; carrying out macroscopic grid division and determining a total resource amount estimation value in each unit; for the over-threshold unit, analyzing a channel and slope intersection in combination with a terrain model so as to locate potential key resource coordinates; scanning the slope accumulation bodies by using an image recognition algorithm, recognizing monomers meeting starting conditions, judging that the volume and the position of each accumulation body meet the starting conditions, and obtaining a microscopic resource detail data set; and finally, multi-level data are fused, and a consistency resource total amount evaluation result is obtained through interpolation and cyclic verification. And mapping information at macroscopic, mesoscopic and microscopic levels by using the conversion matrix, and judging that the mapped data meets the supply relationship requirement to obtain a multi-scale resource identification model. The method can effectively improve the precision and integrity of mineral resource identification in complex mountainous areas.

Description

Technical Field

[0001] This application belongs to the field of information technology, and more specifically, relates to a multi-scale quantitative identification method for debris flow sources. Background Technology

[0002] Debris flow source investigation and assessment is the most fundamental and decisive link in the disaster chain in mountainous areas. Its accuracy directly affects subsequent hazard zoning, selection of engineering prevention and control schemes, estimation of engineering quantities, and the scientific nature of monitoring and early warning systems. In reality, the spatial distribution of debris flow sources is extremely uneven and multi-scale nested, ranging in scale from large landslides at the watershed level to scattered loose deposits at the slope level, with a huge spatial scale span.

[0003] Currently, most debris flow source surveys still heavily rely on the personal experience of investigators and are constrained by project funding and timelines, generally employing fixed-scale methods. When using a smaller scale for a general survey of the entire watershed, it's easy to overlook small but critically located loose deposits that could significantly amplify the disaster; conversely, using a large-scale, detailed survey makes it difficult to cover the entire watershed under limited conditions, leading to the neglect of numerous potential sources and causing serious biases in the total source assessment. A more fundamental problem is that source data obtained at different scales exhibit systematic incompatibility in terms of spatial coverage, statistical units, accuracy, and weighting criteria, making effective overlay, summarization, and verification difficult. This results in vastly different total source assessments for the same debris flow gully in different surveys, severely impacting the reliability of prevention engineering design and the scientific rigor of decision-making.

[0004] The underlying reason for the above problems lies in the lack of a widely accepted quantitative scale classification standard based on measurable watershed characteristic parameters. Due to the absence of such an objective scale classification basis, it is impossible to scientifically set the minimum identification unit, survey precision, and data conversion and correction principles between different survey levels (macro, meso, and micro), resulting in survey results remaining in a state of "scale dependence, incomparability, and difficulty in verification."

[0005] Therefore, improving the accuracy and completeness of mineral resource identification in complex mountainous areas is an urgent problem to be solved. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the purpose of this application is to provide a multi-scale quantitative identification method for debris flow sources, which can effectively improve the accuracy and completeness of mineral resource identification in complex mountainous areas.

[0007] To achieve the above objectives, in a first aspect, this application provides a multi-scale quantitative identification method for debris flow sources, comprising the following steps:

[0008] S10, collect regional spatial feature parameters through remote sensing data, and use a clustering algorithm to group the regional spatial feature parameters to obtain a set of scale division thresholds; S20, Divide the threshold set according to the scale, apply the grid segmentation method to divide the entire region at the macro level to obtain multiple units, and determine the estimated total resource value in each unit; S30. If the estimated total amount of resources within a unit exceeds a preset threshold, a terrain model is used to analyze the intersection of the channel length and the slope range at the meso-level to obtain the distribution coordinates of potential key resources. S40, using the distribution coordinates, an image recognition algorithm is applied to scan the accumulation within the slope area at the micro level, and it is determined that the volume and position of each accumulation meet the activation conditions, thus obtaining a micro-resource detail dataset; S50, Match and fuse the micro-resource detail dataset with the distribution coordinates, use a data interpolation algorithm to fill the missing information between different levels, and obtain the total resource evaluation result after consistency verification by verification cycle comparison and adjustment. S60, for the total resource assessment results after the consistency verification, a transformation matrix is ​​applied to map the information at the macro, meso and micro levels, and it is determined that the mapped data meets the replenishment relationship requirements, thus obtaining a multi-scale resource identification model.

[0009] As a further preferred option, step S10 includes: The original image data of the target area is acquired by remote sensing satellites, and OpenCV is used to denoise and geometrically correct the original image data to obtain corrected image data. Based on the corrected image data, spatial feature parameters within the region are extracted to determine the set of spatial feature parameters. The calculation formula is as follows: Where A represents the spatial area characteristic parameter of the region, and R represents the region. This represents the corrected image data, where (i, j) represents the pixel position; For the set of spatial feature parameters, the K-means algorithm is applied to group the parameters. If the distance between a parameter and the cluster center is less than a preset threshold, it is assigned to the corresponding category, and a preliminary grouping result is obtained. From the preliminary grouping results, obtain the parameter distribution range of each category, use NumPy to calculate the mean and standard deviation of each category as boundary values, and determine the initial threshold for scale division; Based on the initial threshold, the preliminary grouping results are verified a second time. If the parameter distribution of a certain category exceeds the boundary value range, the excess parameters are moved to the adjacent category to obtain the optimized grouping category. By optimizing the grouping categories, the threshold under the distribution pattern is calculated using the percentile function of SciPy, and the final scale division threshold set is obtained.

[0010] As a further preferred option, the calculation formulas for denoising and geometric correction of the original image data using OpenCV are as follows:

[0011]

[0012] In the formula, This represents the pixel value of the image after denoising. This represents the pixel values ​​of the original remote sensing satellite image; This represents the Gaussian filter kernel function. Indicates standard deviation, k Indicates the radius of the filtering window. m and n Indicates the offset within the window; Represents the coordinates after geometric correction; Indicates the coordinates before correction; This represents the parameters of the affine transformation matrix.

[0013] As a further preferred embodiment, the formula for calculating the spatial feature parameter and the distance to the cluster center is:

[0014] In the formula, d represents the distance between the spatial feature parameter and the cluster center; Represents the i-th spatial feature parameter; The cluster centers represent the k-th cluster; m represents the parameter dimension. This represents the value of the j-th dimension of the i-th parameter; Let represent the value of the j-th dimension at the k-th center.

[0015] As a further preferred option, step S20 includes: The grid size is divided according to the threshold set defined by the scale. Multiple grid cells are obtained from the divided grid size, and spatial cutting is performed on the target area using ArcGIS tools; Extract the set of pixels covered by each grid cell from the corrected image data; Count the number of resource pixels for the pixel set of each grid cell; The total resource quantity of each grid cell is estimated based on the number of resource pixels. The total resource estimates of all grid cells are aggregated to form a set of regional resource distributions.

[0016] As a further preferred option, step S30 includes: If the estimated total resource volume within a unit exceeds a preset threshold, then meso-level regional data is obtained from QGIS, and geographic filtering is performed to determine the analysis scope. Based on the analysis scope, the channel length data in the digital elevation model of ArcGIS is extracted to obtain the geographical interval; By combining geographical regions with slope range data obtained from the database, the Intersect tool in ArcGIS is used to overlay the data to obtain the boundaries of the intersecting areas. Using satellite imagery data from the boundaries of the intersecting areas, resource distribution feature values ​​are extracted in ENVI to obtain the first distribution feature. Based on the distribution characteristics, spatial clustering points are obtained by using the K-means algorithm to obtain location clusters. Feature vectors are obtained from the location information of spatial aggregation points, and then classified in a support vector machine to obtain high-priority locations. By combining high-priority locations with terrain auxiliary data obtained from ArcGIS, buffer verification is performed to obtain the final resource distribution coordinates.

[0017] As a further preferred option, step S40 includes: Using the aforementioned distribution coordinates, the target area within the slope range is initially divided at the micro level. OpenCV is then used to scan the accumulated bodies within the divided area one by one to obtain preliminary accumulated volume data. Based on the preliminary accumulation volume data, the position coordinates of each accumulation body were recorded and compared with the distribution coordinates to calculate the Euclidean distance. ,in x 1, y 1 represents the initial position's x and y coordinates. x 2, y 2 represents the horizontal and vertical coordinates of the distribution coordinates, determining the accuracy of the location coordinates; If the Euclidean distance between the location coordinates and the distribution coordinates exceeds a preset threshold, the scan data of the accumulation body will be re-acquired and corrected by combining it with the terrain data within the slope area to obtain the corrected location coordinates. By using the corrected position coordinates, the volume data of each accumulation is extracted a second time to determine whether the volume data meets the preset start conditions, and a list of accumulations that meet the conditions is obtained. Based on the list of eligible accumulations, extract the resource detail data for each accumulation, including comprehensive information on volume and location coordinates, to determine the final set of micro-resource details; From the final set of micro-resource details, the resource detail data for each stack is classified and organized, and stored using PostgreSQL to obtain a structured dataset. By using a structured dataset, the resource distribution within the slope area is labeled in layers to generate a micro-level resource distribution layer, thus determining the final resource distribution view.

[0018] As a further preferred embodiment, step S50 includes: Based on the micro-resource detail dataset, combined with detailed data and distribution coordinate information, the resource distribution hierarchy information is initially organized, and the missing parts are supplemented by linear interpolation using the griddata function of SciPy to obtain preliminary supplemented resource distribution information. Based on the preliminary information on resource distribution, the differences between data layers are calculated. The NumPy subtract function is used to obtain the difference value. If the difference value exceeds the preset threshold of 0.05, Kriging interpolation is performed on the missing information to determine the continuity information of data between layers. By combining the data continuity information between layers with the distribution coordinates and detailed data, the total amount of resources is calculated by layer-by-layer summation to obtain the resource distribution subsets at each layer and thus obtain the total resource data at each layer. Based on the total amount of resources in each layer, the overall trend is calculated using the mean function of NumPy. The deviation of each layer of data is compared. If the deviation of a certain layer exceeds the preset threshold of 0.1, the data of that layer is adjusted by weighted average to obtain the adjusted consistent resource data. By adjusting the consistent resource data, the Pandas groupby function is used to perform hierarchical mapping of the relationship between micro-resources and total resources, and to determine whether the resource distribution at each level conforms to the preset normal distribution law, thus obtaining the mapped resource distribution information. Based on the mapped resource distribution information, combined with the consistency value and evaluation data, the total amount of resources is summed and calculated to determine the final total resource evaluation result.

[0019] As a further preferred option, step S60 includes: Obtain the numerical array from the total resource quantity assessment results after consistency verification; The transformation matrix was constructed using the NumPy library to perform multiplication on the numerical array, resulting in macroscopic, mesoscopic, and microscopic mapping data. Based on the numerical distribution in the macro-level mapping data and the numerical distribution in the meso-level mapping data, determine whether the macro-level resource quantity is greater than or equal to the total meso-level resource quantity. If the condition is met, retain the macro-level mapping data, meso-level mapping data, and micro-level mapping data. If the condition is not met, terminate the current process and mark it as an exception. By retaining the macro-level mapping data, meso-level mapping data, and micro-level mapping data, the replenishment relationship between the meso-level mapping data and the micro-level mapping data is verified to determine whether the meso-level resource quantity is equal to the sum of the corresponding micro-level resource quantity, thus obtaining the preliminary macro-level data, preliminary meso-level data, and preliminary micro-level data that satisfy the replenishment relationship. For the preliminary macro-level data, preliminary meso-level data, and preliminary micro-level data that satisfy the supply relationship, the linear regression function of the SciPy library is used to adjust the ratio coefficient between the macro-level and meso-level data, so as to obtain the proportionally adjusted macro-level distribution data, proportionally adjusted meso-level distribution data, and proportionally adjusted micro-level distribution data. Based on the adjusted macro-level distribution data, adjusted meso-level distribution data, and adjusted micro-level distribution data, the proportion of resource quantity at each level to the total resource assessment result is calculated, and the proportion distribution set is obtained. By using the proportion distribution set, we can determine whether the sum of the proportion values ​​of the three layers is equal to the preset unit value, and obtain resource distribution data that meets multi-scale constraints.

[0020] In a second aspect, this application provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method as described in any one of the above descriptions.

[0021] This application offers the following advantages: It discloses a multi-scale fusion method for accurate mineral resource identification. Addressing the operational challenge of traditional methods struggling to simultaneously consider macroscopic distribution patterns, mesoscopic channel control, and microscopic accumulation initiation characteristics in complex terrain, leading to significant deviations in total resource assessment and severe underestimation of key enrichment areas, this method extracts regional spatial feature parameters from remote sensing data and clusters them to obtain a set of scale division thresholds. Based on this, the study area is divided into gridded macroscopic units, and the total resource volume is initially estimated. For units exceeding the threshold, at the mesoscopic level, terrain models are used to analyze the intersection of channels and slopes to determine the coordinates of potential key resources. Then, at the microscopic level, image recognition technology is used to accurately scan the volume and location of accumulations, filtering resource details that meet the initiation conditions. Finally, through multi-level data matching and fusion, interpolation filling, and consistency verification, combined with a transformation matrix, an effective mapping and replenishment relationship verification of macroscopic-mesoscopic-microscopic information is achieved, thereby constructing a logically consistent and scale-coordinated multi-scale resource identification model. This application significantly improves the accuracy and completeness of mineral resource identification in complex mountainous areas, providing reliable technical support for efficient exploration deployment. Attached Figure Description

[0022] Figure 1 This is a flowchart of the multi-scale quantitative identification method for debris flow sources provided in the embodiments of this application; Figure 2 This is a schematic diagram of the multi-scale quantitative identification method for debris flow sources provided in the embodiments of this application; Figure 3 This is another technical principle diagram of the multi-scale quantitative identification method for debris flow sources provided in the embodiments of this application. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0024] This application's research reveals a strong hierarchical dependence and scale effect between the development and distribution characteristics of debris flow sources and basic geometric and geomorphological parameters such as watershed area, main channel length, average slope, tributary density, and elevation difference. Currently, the fields of engineering geology and disaster prevention still lack a widely accepted quantitative scale classification standard based on these directly measurable watershed characteristic parameters. Due to the lack of such objective scale classification criteria, it is impossible to set scientifically reasonable spatial minimum identification unit area thresholds, survey accuracy requirements, and source quantity conversion and correction principles for the macroscopic (watershed-wide), mesoscopic (main channel and major tributary levels), and microscopic (individual slope or local accumulation body in the gully bed) levels. Without these standards, source survey results remain in an awkward state of "scale dependence, incomparability, and difficulty in verification," severely hindering the transformation of debris flow disaster prevention from experience-based judgment to quantitative and standardized approaches.

[0025] In this regard, such as Figures 1-3 This application provides a multi-scale quantitative identification method for debris flow sources, comprising the following steps: Step S101: Spatial feature parameters of the region are collected by remote sensing data, and the parameters are grouped by a clustering algorithm to obtain a set of scale division thresholds.

[0026] The original image data of the target area is acquired by remote sensing satellites, and OpenCV is used to denoise and geometrically correct the original image data to obtain corrected image data.

[0027]

[0028] In the formula, This represents the pixel value of the image after denoising. This represents the pixel values ​​of the original remote sensing satellite image; This represents the Gaussian filter kernel function. Indicates standard deviation, k Indicates the radius of the filtering window.m and n This represents the offset within the window; this formula is used in OpenCV denoising to remove image noise.

[0029] In the formula, Represents the coordinates after geometric correction; Indicates the coordinates before correction; This represents the parameters of the affine transformation matrix; this formula is used in OpenCV geometric correction to achieve image spatial registration. Based on the corrected image data, spatial feature parameters within the region are extracted to determine the set of spatial feature parameters.

[0030] In the formula, A represents the spatial area characteristic parameter of the region, and R represents the region. Let (i, j) represent the corrected image data, (i, j) represent the pixel position, and sum represent the summation of the corrected image intensities within the region, used to extract spatial scale parameters. For the set of spatial feature parameters, the K-means algorithm is applied to group the parameters. If the distance between a parameter and the cluster center is less than a preset threshold, it is assigned to the corresponding category, resulting in preliminary grouping results.

[0031]

[0032] In the formula, d represents the distance between the spatial feature parameter and the cluster center. Represents the i-th spatial feature parameter. Let m represent the cluster center of the k-th cluster, and m represent the parameter dimension. This represents the value of the j-th dimension of the i-th parameter. Let represent the value of the j-th dimension at the k-th center. From the initial grouping results, obtain the parameter distribution range for each category. Use NumPy to calculate the mean and standard deviation of each category as boundary values ​​to determine the initial threshold for scale division. Based on the initial threshold, perform a secondary verification on the initial grouping results. If the parameter distribution of a certain category exceeds the boundary value range, move the excess parameters to adjacent categories to obtain optimized grouping categories. By optimizing the grouping categories, use the percentile function of SciPy to calculate the threshold under the distribution pattern to obtain the final set of scale division thresholds.

[0033] Step S102: Divide the threshold set according to the scale, apply the grid segmentation method to divide the entire region at the macro level, obtain multiple units, and determine the estimated total resource value in each unit.

[0034] The grid size is divided using a pre-determined scale threshold set. Multiple grid cells are obtained from the divided grid size, and spatial cutting is performed on the target area using ArcGIS tools. The pixel set covered by each grid cell is extracted from the corrected image data. The number of resource pixels is counted for the pixel set of each grid cell. An estimated total resource value for each grid cell is determined based on the number of resource pixels. The estimated total resource values ​​for all grid cells are then aggregated to form a regional resource distribution set.

[0035] Step S103: If the estimated total resource amount within the unit exceeds a preset threshold, then for the meso-level, a terrain model is used to analyze the intersection of the channel length and the slope range to obtain the distribution coordinates of potential key resources.

[0036] If the estimated total resource volume within a unit exceeds a preset threshold, meso-level regional data is obtained from QGIS, and geographic filtering is performed to determine the analysis scope. Based on the analysis scope, channel length data from the digital elevation model (DEM) in ArcGIS is extracted to obtain geographic intervals. These geographic intervals are then combined with slope range data obtained from the database, and overlaid using the Intersect tool in ArcGIS to obtain the boundaries of intersecting areas. From these intersecting area boundaries, resource distribution feature values ​​are extracted using satellite imagery data in ENVI to obtain distribution features. Based on these distribution features, location clustering is performed using the K-means algorithm to obtain spatial cluster points. Feature vectors are obtained from the location information of these spatial cluster points and classified using a support vector machine to obtain high-priority locations. Finally, these high-priority locations are combined with terrain auxiliary data obtained from ArcGIS for buffer validation to obtain the final resource distribution coordinates.

[0037] Step S104: Using distributed coordinates, an image recognition algorithm is applied to scan the accumulation bodies within the slope area at the micro level to determine whether the volume and position of each accumulation body meet the activation conditions, thereby obtaining a micro-resource detail dataset.

[0038] By using distributed coordinates, the target area within the slope is initially divided at a microscopic level. OpenCV is then used to scan each accumulation within the divided area to obtain preliminary accumulation volume data. Based on this preliminary accumulation volume data, the position coordinates of each accumulation are recorded and compared with the distributed coordinates to calculate the Euclidean distance. Where x1 and y1 are the initial location coordinates, and x2 and y2 are the distribution coordinates, the accuracy of the location coordinates is determined. If the Euclidean distance between the location coordinates and the distribution coordinates exceeds a preset threshold, the scan data of that accumulation is re-acquired and corrected by combining it with the terrain data within the slope area to obtain the corrected location coordinates. Using the corrected location coordinates, the volume data of each accumulation is extracted a second time to determine whether the volume data meets the preset start conditions, resulting in a list of accumulations that meet the conditions. Based on the list of accumulations that meet the conditions, the resource detail data of each accumulation is extracted, including comprehensive information on volume and location coordinates, to determine the final set of micro-resource details. From the final set of micro-resource details, the resource detail data of each accumulation is classified and organized, and stored using PostgreSQL to obtain a structured dataset. Using the structured dataset, the resource distribution within the slope area is hierarchically labeled to generate a micro-level resource distribution layer, determining the final resource distribution view.

[0039] Step S105: Match and fuse the micro-resource detail dataset with the distribution coordinates, use a data interpolation algorithm to fill the missing information between different levels, and obtain the total resource assessment result after consistency verification by cyclic comparison and adjustment.

[0040] Based on micro-level resource data, combined with detailed data and distribution coordinate information, the hierarchical information of resource distribution is initially organized. SciPy's `griddata` function is used to perform linear interpolation to fill in missing parts, obtaining preliminary incomplete resource distribution information. Based on this preliminary incomplete information, the differences between hierarchical data are calculated. NumPy's `subtract` function is used to obtain the difference values. If the difference value exceeds a preset threshold of 0.05, Kriging interpolation is performed on the missing information to determine the continuity of data between hierarchical levels. Using the continuity information between hierarchical levels, combined with distribution coordinates and detailed data, the total resource amount is calculated hierarchically to obtain subsets of resource distribution at each level, resulting in hierarchical total resource data. Based on the hierarchical total resource data, NumPy's `mean` function is used to calculate the overall trend. Deviations are compared between each level of data. If the deviation at a certain level exceeds a preset threshold of 0.1, a weighted average adjustment is performed on the data at that level to obtain adjusted consistent resource data. By adjusting the consistent resource data, Pandas' groupby function is used to perform a hierarchical mapping between micro-level resources and total resource volume. This determines whether the resource distribution at each level conforms to a pre-defined normal distribution, yielding the mapped resource distribution information. Based on this mapped resource distribution information, combined with the consistency value and evaluation data, the total resource volume is summed and calculated to determine the final resource volume evaluation result.

[0041] Step S106: Based on the total resource assessment results after consistency verification, a transformation matrix is ​​applied to map information at the macro, meso, and micro levels, and the mapped data is judged to meet the replenishment relationship requirements to obtain a multi-scale resource identification model.

[0042] Obtain a numerical array from the total resource assessment results after consistency verification. Use the NumPy library to construct a transformation matrix and multiply the numerical array to obtain macro-level, meso-level, and micro-level mapping data. Based on the numerical distributions in the macro-level and meso-level mapping data, determine if the macro-level resource quantity is greater than or equal to the sum of the meso-level resources. If so, retain the macro-level, meso-level, and micro-level mapping data; otherwise, terminate the current process and mark an exception. Using the retained macro-level, meso-level, and micro-level mapping data, verify the supply relationship between the meso-level and micro-level mapping data to determine if the meso-level resource quantity equals the sum of the corresponding micro-level resources, thus obtaining preliminary macro-level, preliminary meso-level, and preliminary micro-level data that satisfy the supply relationship. For the preliminary macro-level, preliminary meso-level, and preliminary micro-level data that satisfy the replenishment relationship, the linear regression function of the SciPy library is used to adjust the ratio coefficients between the macro-level and meso-level data, resulting in proportionally adjusted macro-level, meso-level, and micro-level distribution data. Based on these proportionally adjusted data, the proportion of resource quantity at each level to the total resource assessment result is calculated, resulting in a set of proportion distributions. Using this set, it is determined whether the sum of the proportions at the three levels equals a preset unit value, thus obtaining resource distribution data that conforms to multi-scale constraints.

[0043] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A multi-scale quantitative identification method for debris flow sources, characterized in that, Includes the following steps: S10, collect regional spatial feature parameters through remote sensing data, and use a clustering algorithm to group the regional spatial feature parameters to obtain a set of scale division thresholds; S20, Divide the threshold set according to the scale, apply the grid segmentation method to divide the entire region at the macro level to obtain multiple units, and determine the estimated total resource value in each unit; S30. If the estimated total amount of resources within a unit exceeds a preset threshold, a terrain model is used to analyze the intersection of the channel length and the slope range at the meso-level to obtain the distribution coordinates of potential key resources. S40, using the distribution coordinates, an image recognition algorithm is applied to scan the accumulation within the slope area at the micro level, and it is determined that the volume and position of each accumulation meet the activation conditions, thus obtaining a micro-resource detail dataset; S50, Match and fuse the micro-resource detail dataset with the distribution coordinates, use a data interpolation algorithm to fill the missing information between different levels, and obtain the total resource evaluation result after consistency verification by verification cycle comparison and adjustment. S60, for the total resource assessment results after the consistency verification, a transformation matrix is ​​applied to map the information at the macro, meso and micro levels, and it is determined that the mapped data meets the replenishment relationship requirements, thus obtaining a multi-scale resource identification model.

2. The multi-scale quantitative identification method for debris flow sources as described in claim 1, characterized in that, Step S10 includes: The original image data of the target area is acquired by remote sensing satellites, and OpenCV is used to denoise and geometrically correct the original image data to obtain corrected image data. Based on the corrected image data, spatial feature parameters within the region are extracted to determine the set of spatial feature parameters. The calculation formula is as follows: Where A represents the spatial area characteristic parameter of the region, and R represents the region. This represents the corrected image data, where (i, j) represents the pixel position; For the set of spatial feature parameters, the K-means algorithm is applied to group the parameters. If the distance between a parameter and the cluster center is less than a preset threshold, it is assigned to the corresponding category, and a preliminary grouping result is obtained. From the preliminary grouping results, obtain the parameter distribution range of each category, use NumPy to calculate the mean and standard deviation of each category as boundary values, and determine the initial threshold for scale division; Based on the initial threshold, the preliminary grouping results are verified a second time. If the parameter distribution of a certain category exceeds the boundary value range, the excess parameters are moved to the adjacent category to obtain the optimized grouping category. By optimizing the grouping categories, the threshold under the distribution pattern is calculated using the percentile function of SciPy, and the final scale division threshold set is obtained.

3. The multi-scale quantitative identification method for debris flow sources as described in claim 2, characterized in that, The calculation formulas for denoising and geometric correction of raw image data using OpenCV are as follows: In the formula, This represents the pixel value of the image after denoising. This represents the pixel values ​​of the original remote sensing satellite image; This represents the Gaussian filter kernel function. Indicates standard deviation, k Indicates the radius of the filtering window. m and n Indicates the offset within the window; Represents the coordinates after geometric correction; Indicates the coordinates before correction; This represents the parameters of the affine transformation matrix.

4. The multi-scale quantitative identification method for debris flow sources as described in claim 2, characterized in that, The formula for calculating the spatial feature parameters and the distance to the cluster centers is as follows: In the formula, d represents the distance between the spatial feature parameter and the cluster center; Represents the i-th spatial feature parameter; The cluster centers represent the k-th cluster; m represents the parameter dimension. This represents the value of the j-th dimension of the i-th parameter; Let represent the value of the j-th dimension at the k-th center.

5. The multi-scale quantitative identification method for debris flow sources as described in claim 1, characterized in that, Step S20 includes: The grid size is divided according to the threshold set defined by the scale. Multiple grid cells are obtained from the divided grid size, and spatial cutting is performed on the target area using ArcGIS tools; Extract the set of pixels covered by each grid cell from the corrected image data; Count the number of resource pixels for the pixel set of each grid cell; The total resource quantity of each grid cell is estimated based on the number of resource pixels. The total resource estimates of all grid cells are aggregated to form a set of regional resource distributions.

6. The multi-scale quantitative identification method for debris flow sources as described in claim 1, characterized in that, Step S30 includes: If the estimated total resource volume within a unit exceeds a preset threshold, then meso-level regional data is obtained from QGIS, and geographic filtering is performed to determine the analysis scope. Based on the analysis scope, the channel length data in the digital elevation model of ArcGIS is extracted to obtain the geographical interval; By combining geographical regions with slope range data obtained from the database, the Intersect tool in ArcGIS is used to overlay the data to obtain the boundaries of the intersecting areas. Using satellite imagery data from the boundaries of the intersecting areas, resource distribution feature values ​​are extracted in ENVI to obtain the first distribution feature. Based on the distribution characteristics, spatial clustering points are obtained by using the K-means algorithm to obtain location clusters. Feature vectors are obtained from the location information of spatial aggregation points, and then classified in a support vector machine to obtain high-priority locations. By combining high-priority locations with terrain auxiliary data obtained from ArcGIS, buffer verification is performed to obtain the final resource distribution coordinates.

7. The multi-scale quantitative identification method for debris flow sources as described in claim 1, characterized in that, Step S40 includes: Using the aforementioned distribution coordinates, the target area within the slope range is initially divided at the micro level. OpenCV is then used to scan the accumulated bodies within the divided area one by one to obtain preliminary accumulated volume data. Based on the preliminary accumulation volume data, the position coordinates of each accumulation body were recorded and compared with the distribution coordinates to calculate the Euclidean distance. ,in x 1, y 1 represents the initial position's x and y coordinates. x 2, y 2 represents the horizontal and vertical coordinates of the distribution coordinates, determining the accuracy of the location coordinates; If the Euclidean distance between the location coordinates and the distribution coordinates exceeds a preset threshold, the scan data of the accumulation body will be re-acquired and corrected by combining it with the terrain data within the slope area to obtain the corrected location coordinates. By using the corrected position coordinates, the volume data of each accumulation is extracted a second time to determine whether the volume data meets the preset start conditions, and a list of accumulations that meet the conditions is obtained. Based on the list of eligible accumulations, extract the resource detail data for each accumulation, including comprehensive information on volume and location coordinates, to determine the final set of micro-resource details; From the final set of micro-resource details, the resource detail data for each stack is classified and organized, and stored using PostgreSQL to obtain a structured dataset. By using a structured dataset, the resource distribution within the slope area is labeled in layers to generate a micro-level resource distribution layer, thus determining the final resource distribution view.

8. The multi-scale quantitative identification method for debris flow sources as described in claim 1, characterized in that, Step S50 includes: Based on the micro-resource detail dataset, combined with detailed data and distribution coordinate information, the resource distribution hierarchy information is initially organized, and the missing parts are supplemented by linear interpolation using the griddata function of SciPy to obtain preliminary supplemented resource distribution information. Based on the preliminary information on resource distribution, the differences between data layers are calculated. The NumPy subtract function is used to obtain the difference value. If the difference value exceeds the preset threshold of 0.05, Kriging interpolation is performed on the missing information to determine the continuity information of data between layers. By combining the data continuity information between layers with the distribution coordinates and detailed data, the total amount of resources is calculated by layer-by-layer summation to obtain the resource distribution subsets at each layer and thus obtain the total resource data at each layer. Based on the total amount of resources in each layer, the overall trend is calculated using the mean function of NumPy. The deviation of each layer of data is compared. If the deviation of a certain layer exceeds the preset threshold of 0.1, the data of that layer is adjusted by weighted average to obtain the adjusted consistent resource data. By adjusting the consistent resource data, the Pandas groupby function is used to perform hierarchical mapping of the relationship between micro-resources and total resources, and to determine whether the resource distribution at each level conforms to the preset normal distribution law, thus obtaining the mapped resource distribution information. Based on the mapped resource distribution information, combined with the consistency value and evaluation data, the total amount of resources is summed and calculated to determine the final total resource evaluation result.

9. The multi-scale quantitative identification method for debris flow sources as described in claim 1, characterized in that, Step S60 includes: Obtain the numerical array from the total resource quantity assessment results after consistency verification; The transformation matrix was constructed using the NumPy library to perform multiplication on the numerical array, resulting in macroscopic, mesoscopic, and microscopic mapping data. Based on the numerical distribution in the macro-level mapping data and the numerical distribution in the meso-level mapping data, determine whether the macro-level resource quantity is greater than or equal to the total meso-level resource quantity. If the condition is met, retain the macro-level mapping data, meso-level mapping data, and micro-level mapping data. If the condition is not met, terminate the current process and mark it as an exception. By retaining the macro-level mapping data, meso-level mapping data, and micro-level mapping data, the replenishment relationship between the meso-level mapping data and the micro-level mapping data is verified to determine whether the meso-level resource quantity is equal to the sum of the corresponding micro-level resource quantity, thus obtaining the preliminary macro-level data, preliminary meso-level data, and preliminary micro-level data that satisfy the replenishment relationship. For the preliminary macro-level data, preliminary meso-level data, and preliminary micro-level data that satisfy the supply relationship, the linear regression function of the SciPy library is used to adjust the ratio coefficient between the macro-level and meso-level data, so as to obtain the proportionally adjusted macro-level distribution data, proportionally adjusted meso-level distribution data, and proportionally adjusted micro-level distribution data. Based on the adjusted macro-level distribution data, adjusted meso-level distribution data, and adjusted micro-level distribution data, the proportion of resource quantity at each level to the total resource assessment result is calculated, and the proportion distribution set is obtained. By using the proportion distribution set, we can determine whether the sum of the proportion values ​​of the three layers is equal to the preset unit value, and obtain resource distribution data that meets multi-scale constraints.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1 to 9.

Images