A feature factor system construction method for high-position remote landslide identification

Abstract

This invention discloses a method for constructing a feature factor system for identifying high-altitude long-distance landslides, comprising: collecting multiple known high-altitude long-distance landslide cases as positive samples and ordinary slope cases with similar terrain as negative samples; establishing a sample library containing a digital elevation model (DEM), radar deformation data, optical remote sensing images, engineering geological rock group maps, and fault distribution maps for each case; identifying the potential initiation zone, the lower slope, the downstream channel, and the deformation zone for each case in the positive and negative samples; extracting candidate feature factors for four analysis objects, calculating the information contribution of each candidate feature factor to its matched analysis object, retaining feature factors whose information contribution meets preset screening conditions, and obtaining a factor set; constructing the factor set and the calculation formula for each feature factor into a feature index system for identifying high-altitude long-distance landslides; this scheme is the first to construct a multi-factor feature system for identifying high-altitude long-distance landslides.

Description

Technical Field

[0001] This invention relates to the field of geological disaster identification technology, specifically to a method for constructing a feature factor system for identifying high-altitude, remote landslides. Background Technology

[0002] High-altitude long-distance landslides refer to landslides that develop on the upper part of steep slopes, with high shear exit positions (the height difference between the shear exit and the leading edge is generally greater than 100m), long sliding distances, large volumes, and fast movement speeds. High-altitude long-distance landslides can be divided into three geographical blocks from top to bottom: the initiation zone, the lower slope, and the downstream channel.

[0003] Its initiation zone is usually located on the upper part of a ridge or slope, with high gravitational potential energy at the beginning. After instability under the action of gravity, the landslide mass achieves high-speed movement with huge initial potential energy (peak speed can reach more than 70 m / s). Through mechanisms such as scraping and erosion, dynamic fracturing, air cushion effect and boundary layer effect, the energy loss during the movement is significantly reduced, thus generating a movement distance far exceeding that of ordinary landslides. During the movement, the landslide mass often transforms into debris flow or even debris flow, forming a chain disaster of landslide-debris flow-debris flow. It has the characteristics of strong kinetic energy and strong fragmentation-fluidization, and poses a huge threat to mountain towns, road and bridge infrastructure and water conservancy and hydropower projects.

[0004] Existing technologies mostly identify potential landslide hazards in high-altitude, remote areas by superimposing the values ​​of single or multiple influencing factors. If a single influencing factor meets an abnormal threshold, it is marked as high risk. There is no complete system of characteristic factors for identifying potential landslide hazards in high-altitude, remote areas. Summary of the Invention

[0005] The purpose of this invention is to provide a method for constructing a feature factor system for identifying high-altitude, remote landslides, in order to solve the technical problems in the existing technology where the determination of regional high-altitude, remote landslides still relies on the interpretation and analysis of technical personnel, and there are no relevant identification rules or quantitative models available for reference, which are characterized by strong subjectivity, low efficiency, and high uncertainty.

[0006] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution:

[0007] A method for constructing a feature factor system for identifying high-altitude, long-distance landslides includes the following steps:

[0008] Step 100: Collect multiple known high-altitude remote landslide cases as positive samples and ordinary slope cases with similar terrain as negative samples, and establish a sample library containing digital elevation model (DEM), radar monitoring deformation data, optical remote sensing images, engineering geological rock group maps and fault distribution maps for each case.

[0009] Step 200: Combining the digital elevation model (DEM), radar deformation data, and optical remote sensing images of the positive and negative sample cases, identify the potential initiation zone, lower slope, downstream channel, and deformation zone for each positive and negative sample case;

[0010] Step 300: Extract candidate feature factors of the four analysis objects. Based on the numerical distribution of each candidate feature factor in the positive and negative sample sets in the sample library, calculate the information contribution of each candidate feature factor to its matched analysis object. Retain feature factors whose information contribution meets the preset screening conditions to obtain a factor set.

[0011] Step 400: Construct the set of factors and the calculation formula of each feature factor into a feature index system for identifying high-altitude remote landslides.

[0012] Compared with the prior art, the present invention has the following advantages:

[0013] (1) This invention constructs a special characteristic index system for high-altitude long-distance landslides: For the whole process of high-altitude long-distance landslides, namely “high-altitude initiation - high-speed movement - long-distance accumulation”, the analysis object is divided into four major elements: potential initiation area, lower slope, downstream channel and deformation area for the first time, and 18 core characteristic factors are extracted, filling the gap in the existing technology in the characteristic system of high-altitude long-distance landslides.

[0014] (2) This invention incorporates both static geographic objects (topography, geology) and dynamic monitoring objects (InSAR deformation), overcoming the limitations of traditional methods that rely solely on static interpretation of historical landslides or single deformation information, and can more comprehensively depict the occurrence conditions, movement capabilities and activity levels of high-altitude long-distance landslides.

[0015] (3) The selection of all feature factors and the determination of thresholds in this invention are based on statistical analysis of the sample library rather than on expert experience, which eliminates the problems of strong subjectivity and individual differences, and improves the reliability and reproducibility of the identification results.

[0016] (4) Each feature factor of the present invention corresponds to a specific link (starting, movement, accumulation, timing) in the landslide chain process, providing clear input parameters and reference thresholds for the subsequent establishment of a quantitative identification model, which can effectively guide the identification of high-altitude long-distance landslides. Attached Figure Description

[0017] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the process for constructing a feature factor system according to an embodiment of the present invention; Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] like Figure 1 As shown, this invention provides a method for constructing a feature factor system for identifying high-altitude long-distance landslides. This embodiment constructs a special feature index system for high-altitude long-distance landslides. Based on the characteristics of high-altitude long-distance landslides, a special feature system covering four major elements is designed, including the potential initiation zone, the slope below, the downstream channel, and the deformation zone. Key discriminative features such as "distance from the potential initiation zone to the channel", "channel gradient", and "deformation location" are introduced, filling the gap in the existing technology for feature systems of high-altitude long-distance landslides.

[0021] Specifically, the following steps are included:

[0022] Step 100: Collect multiple known high-altitude remote landslide cases as positive samples and ordinary slope cases with similar terrain as negative samples, and establish a sample library containing digital elevation model (DEM), radar monitoring deformation data, optical remote sensing images, engineering geological rock group maps and fault distribution maps for each case.

[0023] High-altitude long-distance landslides refer to landslides that develop on the upper part of steep slopes, with high shear exit locations (the height difference between the shear exit and the leading edge is generally greater than 100m), long sliding distances, large volumes, and high speeds. Their initiation zones are usually located on ridges or the upper part of slopes, initially possessing high gravitational potential energy. After instability under gravity, the landslide mass achieves high-speed movement (peak speeds can reach over 70 m / s) thanks to its enormous initial potential energy. Through mechanisms such as scraping and erosion, dynamic fracturing, air cushion effects, and boundary layer effects, it significantly reduces energy loss during movement, resulting in a migration distance far exceeding that of ordinary landslides. During movement, the landslide mass often transforms into debris flows or even mudflows, forming a landslide-debris flow-mudflow chain disaster. It is characterized by strong kinetic energy and intense fragmentation-fluidization, posing a significant threat to mountain towns, road and bridge infrastructure, and water conservancy and hydropower projects.

[0024] In this embodiment, ordinary slope cases with terrain similar to the positive sample cases are further used as negative samples, so as to facilitate the construction of a feature factor system for high-altitude long-distance landslide identification by analyzing the data of positive and negative samples.

[0025] Step 200: Combining the digital elevation model (DEM), radar deformation data, and optical remote sensing images of the positive and negative sample cases, identify the potential starting area, lower slope, downstream channel, and deformation zone for each positive and negative sample case.

[0026] In step 200, potential starting areas and underlying slopes in each case are identified using optical remote sensing images;

[0027] Identify downstream channels in each case using a digital elevation model (DEM);

[0028] By monitoring deformation data with radar, the deformation zone in each case is identified, and the overlap between the deformation zone and the potential initiation zone and the underlying slope is determined by combining digital elevation model (DEM) and optical remote sensing imagery.

[0029] This invention incorporates both static geographic objects (topography, geology) and dynamic monitoring objects (InSAR deformation), overcoming the limitations of traditional methods that rely solely on static interpretation of historical landslides or single deformation information. It can more comprehensively characterize the occurrence conditions, movement capabilities, and activity levels of high-altitude, long-distance landslides.

[0030] The specific process for identifying the potential initiation zone, lower slope, downstream channel, and deformation zone for each case of positive and negative samples is as follows:

[0031] Interpreting Potential Initiation Zones: In optical remote sensing imagery, potential initiation zones are typically located on ridges, peak shoulders, or above slope breaks, often exhibiting steep cliffs or high-slope landforms. Images show exposed fresh rock surfaces, tensional fracture zones, lateral shear fractures, and rock mass fracture surfaces, often accompanied by small collapses or debris accumulation, presenting an overall macroscopic precursor morphology of high-level rock mass shearing. Compared to the surrounding stable mountains, these zones are lighter in tone, brighter, and have coarser textures. Vegetation is either absent in strips or has low coverage.

[0032] The potential initiation zone is the location where the landslide body initially fractures, becomes unstable, and shears out. The potential initiation zone refers to the source area of ​​a high-altitude, long-distance landslide, located near a ridge or on the upper part of a steep slope. It is the location where the landslide body initially becomes unstable, disintegrates, and initiates. Its geological conditions and geometric shape directly determine whether a landslide can occur, the magnitude of the initial potential energy, and the degree of rock mass disintegration.

[0033] Interpreting the lower slope: In optical remote sensing imagery, the lower slope is a transitional slope connecting the initiation area and the deposition area. The terrain is characterized by a gently sloping or moderately sloping slope with a continuous and intact surface. The leading edge of the slope naturally connects to the downstream water system or gully, without obvious protrusions or depressions. The slope surface retains its original weathered deposition or colluvial layer structure, with natural vegetation distribution, growing either downhill or perpendicular to the slope. The image has a uniform tone, without any discolored or dark patches, and the tone transitions naturally with adjacent slopes, without obvious boundaries or bands.

[0034] The lower slope refers to the sloping area located below the potential initiation zone, connecting the potential initiation zone and the downstream channel. The lower slope serves as the channel and path for the movement of the landslide body after initiation. Its topographical features control the landslide body's speed, distance, and mode of movement. It is the key path for the landslide body to transition from initial sliding to high-speed movement. During this stage, the landslide body gains kinetic energy through gravity acceleration and interacts with the underlying strata, generating scraping and entrainment effects, further expanding the scale of the disaster. If the lower slope is too gentle or contains obstacles, the landslide body will slow down and accumulate there, unable to enter the channel for long-distance movement.

[0035] Interpreting downstream channels: Based on DEM data extraction and analysis, these channels are generally located at the nearest downstream water system to the leading edge of the slope or the nearest confluence water system to the slope body. In the imagery, these water systems primarily appear as clearly defined channels or distinctly dissected gullies. The channels generally extend downstream along the slope, their planar morphology controlled by the topography at the leading edge of the slope. Slight bending or deflection due to depositional pressure is visible in some areas. The channel bed is relatively dark in tone, with continuous linear or banded textures.

[0036] Downstream channels refer to the valley systems located at the leading edge of the slope below and connected to the direction of landslide movement. They serve as the conduit for the continued movement of high-altitude, long-distance landslides and the final deposition site. After entering a channel, the landslide mass is constrained by the terrain, causing its movement direction to become more uniform and its energy to become more concentrated. At the same time, a large amount of loose deposits within the channel are carried away, leading to a rapid expansion of the disaster's scale.

[0037] Deformation Zone Interpretation: Based on radar-monitored deformation data, the deformation zone is mainly distributed in the potential initiation zone, the underlying slope, or localized areas thereof. Compared to the surrounding area, the deformation rate or deformation value within the zone exhibits a distinct and continuous high-value anomaly, sometimes forming a significant gradient change characteristic. The time-series deformation curve shows a typical "three-stage" evolution characteristic: the early stage is a low-rate linear or near-linear creep stage; the middle stage enters a nonlinear acceleration stage, with the slope of the deformation rate curve gradually increasing; before the slippage, a significant slope abrupt change and acceleration peak appear, and the deformation rate increases sharply, reflecting the complete connection of the shear zone.

[0038] Unlike static topography and geological factors, deformation zones refer to areas of significant surface displacement identified based on deformation monitoring technology. These zones have a spatial correspondence with potential landslide initiation areas, downstream slopes, or downstream channels. Deformation information reflects the current deformation state and activity intensity of the landslide body and is an important basis for determining whether a landslide is in an active phase and identifying potential sliding surfaces. The spatial distribution, intensity, and development trend of deformation zones can reveal the deformation location, movement direction, and stability state of the landslide.

[0039] The four analytical objects do not exist in isolation, but form an organic whole. The potential initiation zone provides the energy base and determines the generation of high-altitude long-distance landslides; the lower slope determines the energy conversion efficiency and the sliding distance of high-altitude long-distance landslides; the downstream channel controls the disaster-causing range and determines the impact range of high-altitude long-distance landslides; and the deformation zone indicates the urgency of the risk and determines the timing of high-altitude long-distance landslides.

[0040] Existing technologies either focus only on a single aspect (such as analyzing only the initiation zone) or analyze each aspect in isolation, failing to form a complete identification system. This implementation method integrates these four core aspects into a unified analytical framework, achieving for the first time a full-dimensional quantitative identification of high-altitude remote landslides, from "probability of occurrence" to "remote disaster area" and "risk urgency".

[0041] This invention incorporates both static geographic objects (topography, geology) and dynamic monitoring objects (InSAR deformation), overcoming the limitations of traditional methods that rely solely on static interpretation of historical landslides or single deformation information. It can more comprehensively characterize the occurrence conditions, movement capabilities, and activity levels of high-altitude, long-distance landslides.

[0042] Step 300: Extract candidate feature factors of the four analysis objects. Based on the numerical distribution of each candidate feature factor in the positive and negative sample sets in the sample library, calculate the information contribution of each candidate feature factor to its matched analysis object. Retain feature factors whose information contribution meets the preset screening conditions to obtain a factor set.

[0043] When extracting candidate feature factors for the four analysis objects, for the four major analysis elements of potential starting area, lower slope, downstream channel and deformation area divided in step 200, candidate feature factors with clear physical meaning and which can be calculated from multi-source remote sensing data (DEM, InSAR, geological map) are initially selected from aspects such as topography, geological structure, hydrological conditions and deformation characteristics, with a total of no less than 40 items.

[0044] When calculating the information contribution of each candidate feature factor to the complete chain process of high-altitude remote landslides, the sample library (including high-altitude remote landslides and ordinary slopes) constructed in step 100 is used. The candidate feature factor values ​​corresponding to each case are taken as input, and algorithms such as random forest feature importance assessment, recursive feature elimination (RFE) or principal component analysis (PCA) are used to calculate the sensitivity or information contribution of each candidate factor to characterizing the typical pattern of "high-altitude initiation-high-speed movement-remote accumulation" when high-altitude remote landslides occur. Therefore, this embodiment selects the feature factor set by quantifying the independent discrimination ability of each feature factor in distinguishing between high-altitude remote landslides and non-landslides / ordinary slopes, and constructs a feature factor system. This feature factor system can identify the probability of occurrence of high-altitude remote landslides.

[0045] Then, for each selected feature factor, a specific calculation formula based on DEM, InSAR and other data is defined, and the mean, standard deviation and percentile distribution of each factor are statistically analyzed using high-altitude remote landslide samples from historical case data to obtain its anomaly judgment threshold (such as "starting area slope anomaly threshold 30°~50°"), which serves as a reference benchmark for subsequent identification.

[0046] The above 18 characteristic factors, their calculation methods, and typical thresholds are integrated into a dedicated characteristic index system for identifying high-altitude remote landslides. This system integrates multi-dimensional information from static geographic objects (potential initiation areas, lower slopes, and downstream channels) and dynamic monitoring objects (deformation zones), and can be used to guide the automated or semi-automated identification of high-altitude remote landslides.

[0047] In step 300, the candidate feature factors of the potential initiation zone, the lower slope, the downstream channel, and the deformation zone are divided into continuous factors and discrete factors. The information contribution of the continuous factor to the matched analysis object is obtained by calculating the corresponding numerical distribution results of the continuous factor in the positive and negative samples. The information contribution of the discrete factor is calculated by calculating the occurrence ratio of the discrete factor in the positive samples.

[0048] For the potential start-up area, the lower slope, and the downstream channel continuity factors, calculate their information contribution.

[0049] For the positive sample set Xpos and the negative sample set Xneg, calculate the effect size:

[0050] ;

[0051] in, This represents the mean of the candidate feature factors on the positive sample set Xpos. This represents the mean of the candidate feature factors in the negative sample set Xneg;

[0052] This represents the standard deviation of the candidate feature factor in the positive sample set Xpos. This represents the standard deviation of the candidate feature factor in the negative sample set Xneg;

[0053] N is the number of positive samples, and M is the number of negative samples.

[0054] Then normalize d to the interval [0,1]: = .

[0055] For numerical factors in the deformation region, the method for calculating their information contribution is as follows:

[0056] Identify the peak acceleration time Tpeak of a certain numerical factor in the deformation zone and the actual landslide occurrence time Toccur, and calculate the time difference ΔT=Toccur. Tpeak;

[0057] The ΔT of all positive samples in the sample library is normalized, and the normalized value is used as the timing contribution Stime, which is calculated using the following formula:

[0058] Stime={max(ΔT) ΔT} / {max(ΔT) min(ΔT);

[0059] Where max(ΔT) and min(ΔT) are the maximum and minimum time differences of all positive samples in the historical database, respectively;

[0060] The smaller ΔT is, the closer Stime is to 1, and the higher the information contribution of the numerical factors in the deformation region; conversely, the larger ΔT is, the closer Stime is to 0, and the lower the information contribution of the numerical factors in the deformation region.

[0061] For discrete factors in the potential initiation zone, the lower slope, the downstream channel, and the deformation zone, the method for calculating their information contribution is as follows:

[0062] = , where N1 is the frequency of occurrence of the discrete factor in positive sample cases, and N is the number of positive sample cases.

[0063] For each analysis object, candidate feature factors are sorted from largest to smallest according to their corresponding information contribution. Feature factors with a contribution greater than a preset threshold (e.g., 0.5) are retained, and at least 3 feature factors are retained for each analysis object.

[0064] This implementation method breaks through the limitations of traditional landslide identification, which only focuses on a single topographic / geological indicator. It constructs an eighteen-item core feature factor system that combines "static topographic and geological indicators + dynamic deformation indicators" and introduces exclusive identification indicators such as "deformation location" and "angle between downstream channel and sliding direction".

[0065] The final identified characteristic factors of the potential initiation zone include: slope of the potential initiation zone, area of ​​the potential initiation zone, elevation difference of the potential initiation zone, elevation difference between the leading edge of the potential initiation zone and the lower slope toe, engineering geological rock group of the potential initiation zone, and distance to the structure.

[0066] The final determined characteristic factors of the lower slope include: lower slope gradient, lower slope gradient difference, lower slope area, length in the direction of movement, engineering geological rock group of the lower slope, and distance between the leading edge of the slope and the downstream channel along the direction of movement.

[0067] The final determined characteristic factors of the downstream channel include: the angle between the downstream channel and the sliding direction, and the channel gradient;

[0068] The final determined characteristic factors of the deformation zone include: maximum deformation rate, average deformation rate, deformation location, and the angle between the deformation zone and the sliding direction.

[0069] Among them, the engineering geological rock group of the potential start-up area, the distance to the structure, the engineering geological rock group of the slope below, and the deformation location are discrete factors, while the other characteristic factors are continuous factors.

[0070] The 18 factors selected above can be stably and accurately extracted from multi-source remote sensing data (optical images, InSAR, DEM, geological maps), and have clear physical dimensions and calculation methods. The selection of all feature factors and the determination of thresholds in this invention are based on statistical analysis of the sample library (such as Cohen's d), rather than relying on expert experience, which eliminates the problems of strong subjectivity and individual differences, and improves the reliability and reproducibility of the identification results.

[0071] Step 400: Construct the set of factors and the calculation formula of each feature factor into a feature index system for identifying high-altitude remote landslides.

[0072] The selection of all feature factors and the determination of thresholds in this invention are based on statistical analysis of the sample database, rather than relying on expert experience. This eliminates the problems of strong subjectivity and individual differences, and improves the reliability and reproducibility of the identification results.

[0073] Specifically, in step 400, the slope value θ of all raster cells within the polygonal range of the potential start-up area is extracted based on the digital elevation model. i And calculate the slope value θ of all raster cells. iThe arithmetic mean of the potential start-up area is given by the formula: where n is the total number of raster cells in the potential start-up area. The slope of the potential start-up area is calculated using the formula:

[0074] Slope init = θ i ;

[0075] Based on the vector polygons of the identified potential start-up regions, the number N of raster cells they cover is calculated. pixel Multiply by the area A of a single cell pixel This refers to the projected area on the horizontal plane. The formula for calculating the area of ​​the potential initiation zone is:

[0076] Area init =N pixel ×A pixel ;

[0077] The elevation value H of the highest point within the polygonal area of ​​the potential start-up zone is extracted based on the digital elevation model. max Elevation value H of the lowest point min The difference between the two is calculated as the elevation difference of the potential starting area. The specific calculation formula is as follows:

[0078] H init =H max H min ;

[0079] Elevation value H of the lowest point at the leading edge of the potential starting area was extracted based on a digital elevation model. front Elevation value H of the lowest point of the slope toe below. toe The difference between the two is calculated as the elevation difference between the leading edge and the lower toe of the slope. The specific calculation formula is as follows:

[0080] ΔH=H front H toe ;

[0081] Spatial overlay analysis is performed between the vector polygon of the potential start-up area and the vector map of the engineering geological rock group. The overlap area with each rock group type is calculated, and the rock group type code or name with the largest overlap area ratio is extracted as the discrete factor of the engineering geological rock group in the potential start-up area, i.e.:

[0082] R init =argmax r∈R (Area init ∩Area r );

[0083] Calculate the polygonal centroid Cinit of the potential initiation zone to the nearest fault line P. faultThe shortest spatial straight-line distance to any point p is used as the distance to the construction site. The specific formula for calculating the distance to the construction site is as follows:

[0084] D fault = .

[0085] Based on the digital elevation model, the slope value θ of all raster cells within the polygonal area of ​​the lower slope region is extracted. j And calculate the slope value θ of all raster cells. j The arithmetic mean of the values, where n is the total number of raster cells in the lower slope, is used to calculate the slope of the lower slope.

[0086] Slope lower = θ j ;

[0087] The difference between the average slope of the potential initiation zone and the average slope of the lower slope is calculated as the slope difference of the lower slope. A positive value indicates that the potential initiation zone is steeper than the lower slope, which is conducive to landslide initiation. The slope difference is calculated as follows:

[0088] ΔSlope=Slope init Slope lower ;

[0089] Based on the identified vector polygon of the lower slope, the number of raster cells M covered by this vector polygon is determined. pixel Multiply by the area A of a single cell pixel This is the projected area of ​​the slope below, and the specific calculation formula is as follows:

[0090] Area lower =M pixel A pixel ;

[0091] Along the main sliding direction vector of the landslide, the lower slope boundary (Boundary) lower Projecting all points onto this direction, the difference between the point with the maximum and minimum projection values ​​is calculated as the maximum straight-line distance along the direction of motion. Thus, the formula for calculating the length of the lower slope is:

[0092] L lower = ∥proj dir (a) proj dir (b)∥;

[0093] Spatial overlay analysis is performed on the polygon of the lower slope area and the raster or vector map of the engineering geological rock group, and the rock group type code or name with the largest overlapping area is extracted as the engineering geological rock group of the lower slope.

[0094] R lower =argmax r∈R (Area lower ∩Area);

[0095] Extract the lowest point P at the leading edge of the lower slope. front Calculate P front To the nearest trench centerline channel The shortest spatial straight-line distance from any point q is used as the leading edge distance / channel distance. The specific calculation formula is as follows:

[0096] D channel .

[0097] Calculate the acute angle between the overall direction vector of the downstream channel centerline and the overall sliding direction vector of the landslide, with a value of [0°, 90°]. Specifically, the overall sliding direction vector of the landslide is from the centroid of the potential initiation zone to the centroid of the deposition zone or the leading edge of the lower slope. This angle is taken as the angle between the downstream channel and the sliding direction.

[0098] Select the downstream channel segment associated with the landslide, and determine the starting elevation H of this channel segment. start Elevation H at the endpoint end The difference between the two points and the horizontal projection distance L horizontal The ratio is used as the channel gradient, and the specific calculation formula is as follows:

[0099] J=(H start H end) / L horizontal.

[0100] Based on deformation monitoring data from interferometric radar, the deformation rate values ​​r of all pixels within the deformation zone are extracted. k The maximum value among them is taken as the maximum deformation rate of the deformation region. The specific calculation formula is as follows:

[0101] R max= r k );

[0102] The arithmetic mean of the deformation rates of all pixels within the deformation zone is calculated as the average deformation rate. The specific calculation formula is as follows:

[0103] = ;

[0104] Based on the spatial distribution characteristics of deformation rate, spatial clustering analysis was used to identify areas with deformation rates significantly higher than the background value. Combined with the landslide element division results, the range of the initiation zone Z was determined. init The slope range Z below lower Accumulation area Z accum To determine whether the main deformation area belongs to the trailing edge, middle edge, or leading edge, the specific calculation formula is as follows:

[0105] DeformPart=f cluster (D, Z) init Z lower Z accum );

[0106] Based on the vector polygon of the deformation region, the major axis direction vector of the deformation region can be extracted using principal component analysis or the minimum bounding rectangle method. deform This direction is compared with the overall sliding direction vector of the landslide. slide The acute angle between them is used as the angle between the deformation zone and the sliding direction.

[0107] Existing technologies largely rely on the static interpretation of historical landslides, failing to identify developing potential landslide hazards. This invention introduces radar-monitored deformation data, combining static disaster-causing condition analysis with dynamic deformation monitoring, achieving a technological leap from identifying existing landslides to providing early warnings of potential landslides.

[0108] Therefore, this application constructs a special characteristic index system for high-altitude long-distance landslides. Based on the characteristics of high-altitude long-distance landslides, a special characteristic system covering four major elements is designed, including the potential initiation zone, the slope below, the downstream channel, and the deformation zone. Key discrimination features such as the distance from the potential initiation zone to the channel, the channel gradient, and the deformation location are introduced, filling the gap in the existing technology for the characteristic system of high-altitude long-distance landslides.

[0109] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.

Claims

1. A method for constructing a feature factor system for identifying high-altitude, long-distance landslides, characterized in that, Includes the following steps: Step 100: Collect multiple known high-altitude remote landslide cases as positive samples and ordinary slope cases with similar terrain as negative samples, and establish a sample library containing digital elevation model (DEM), radar monitoring deformation data, optical remote sensing images, engineering geological rock group maps and fault distribution maps for each case. Step 200: Combining the digital elevation model (DEM), radar deformation data, and optical remote sensing images of the positive and negative sample cases, identify the potential initiation zone, lower slope, downstream channel, and deformation zone for each positive and negative sample case; Step 300: Extract candidate feature factors of the four analysis objects. Based on the numerical distribution of each candidate feature factor in the positive and negative sample sets in the sample library, calculate the information contribution of each candidate feature factor to its matched analysis object. Retain feature factors whose information contribution meets the preset screening conditions to obtain a factor set. Step 400: Construct the set of factors and the calculation formula of each feature factor into a feature index system for identifying high-altitude remote landslides.

2. The method according to claim 1, characterized in that, In step 200, potential starting areas and underlying slopes in each case are identified using optical remote sensing images; Identify downstream channels in each case using a digital elevation model (DEM); By monitoring deformation data with radar, the deformation zone in each case is identified, and the overlap between the deformation zone and the potential initiation zone and the underlying slope is determined by combining digital elevation model (DEM) and optical remote sensing imagery.

3. The method for constructing a feature factor system for identifying high-altitude, long-distance landslides according to claim 1, characterized in that, In step 300, the candidate feature factors of the potential initiation zone, the lower slope, the downstream channel, and the deformation zone are divided into continuous factors and discrete factors. The information contribution of the continuous factor to the matched analysis object is obtained by calculating the corresponding numerical distribution results of the continuous factor in the positive and negative samples. The information contribution is calculated by calculating the occurrence ratio of the discrete factor in the positive samples.

4. The method according to claim 3, characterized in that, For continuous factors such as the potential starting area, the lower slope, and the downstream channel, the method for calculating their information contribution is as follows: For the positive sample set Xpos and the negative sample set Xneg, calculate the effect size: ； in, This represents the mean of the candidate feature factors in the positive sample cases Xpos. This represents the mean of the candidate feature factors in the negative sample case Xneg; This represents the standard deviation of the candidate feature factor in the positive sample cases Xpos. This represents the standard deviation of the candidate feature factor in the negative sample case Xneg; N is the number of positive sample cases, and M is the number of negative sample cases; Normalize d to the interval [0,1]: = ; For numerical factors in the deformation region, the method for calculating their information contribution is as follows: Identify the peak acceleration time Tpeak of a certain numerical factor in the deformation zone and the actual landslide occurrence time Toccur, and calculate the time difference ΔT=Toccur. Tpeak; The ΔT of all positive samples in the sample library is normalized, and the normalized value is used as the timing contribution Stime, which is calculated using the following formula: Stime={max(ΔT) ΔT} / {max(ΔT) min(ΔT); Where max(ΔT) and min(ΔT) are the maximum and minimum time differences of all positive samples in the historical database, respectively; The smaller ΔT is, the closer Stime is to 1, and the higher the information contribution of the numerical factors in the deformation region; conversely, the larger ΔT is, the closer Stime is to 0, and the lower the information contribution of the numerical factors in the deformation region. For discrete factors in the potential initiation zone, the lower slope, the downstream channel, and the deformation zone, the method for calculating their information contribution is as follows: = , where N1 is the frequency of occurrence of the discrete factor in positive sample cases, and N is the number of positive sample cases.

5. The method according to claim 4, characterized in that, For each analysis object, the candidate feature factors are sorted from largest to smallest according to their corresponding contribution. Feature factors with a contribution greater than the preset contribution threshold are retained, and at least 3 feature factors are retained for each analysis object.

6. The method according to claim 5, characterized in that, In step 300, the characteristic factors of the potential initiation zone that are finally determined include: the slope of the potential initiation zone, the area of ​​the potential initiation zone, the elevation difference of the potential initiation zone, the elevation difference between the leading edge of the potential initiation zone and the lower slope toe, the engineering geological rock group of the potential initiation zone, and the distance to the structure. The final determined characteristic factors of the lower slope include: lower slope gradient, lower slope gradient difference, lower slope area, length in the direction of movement, engineering geological rock group of the lower slope, and distance between the leading edge of the slope and the downstream channel along the direction of movement. The final determined characteristic factors of the downstream channel include: the angle between the downstream channel and the sliding direction, and the channel gradient; The final determined characteristic factors of the deformation zone include: maximum deformation rate, average deformation rate, deformation location, and the angle between the deformation zone and the sliding direction. Among them, the engineering geological rock group of the potential start-up area, the engineering geological rock group of the slope below, and the deformation location are discrete factors, while the other characteristic factors are continuous factors.

7. The method according to claim 6, characterized in that, In step 400, the slope value θ of all raster cells within the polygonal range of the potential starting area is extracted based on the digital elevation model. i And calculate the slope value θ of all raster cells. i The arithmetic mean of the values, where n is the total number of raster cells in the potential initiation zone, is used to calculate the slope of the potential initiation zone. Slope init = i i ; Based on the vector polygon of the identified potential start-up region, the number N of raster cells it covers is calculated. pixel Multiply by the area A of a single cell pixel That is, the projected area on the horizontal plane. The formula for calculating the area of ​​the potential initiation zone is: Area init =N pixel ×A pixel ; The elevation value H of the highest point within the polygonal range of the potential start-up area is extracted based on the digital elevation model. max Elevation value H of the lowest point min The difference between the two is calculated as the elevation difference of the potential start-up area. The specific calculation formula is as follows: H init =H max H min ; The elevation value H of the lowest point at the leading edge of the potential starting area is extracted based on the digital elevation model. front Elevation value H of the lowest point of the slope toe below. toe The difference between the two is calculated as the elevation difference between the leading edge and the lower toe of the slope. The specific calculation formula is as follows: ΔH=H front H toe ; Spatial overlay analysis is performed between the vector polygon of the potential start-up area and the engineering geological rock group vector map. The overlap area with each rock group type is calculated, and the rock group type code or name with the largest overlap area ratio is extracted as the discrete factor of the engineering geological rock group in the potential start-up area, i.e.: R init =argmax r∈R (Area init ∩Area r ); Calculate the polygonal centroid point Cinit of the potential initiation region to the nearest fault line P. fault The shortest spatial straight-line distance to any point p is used as the distance to the construction site. The specific formula for calculating the distance to the construction site is as follows: D fault = 。 8. The method according to claim 6, characterized in that, Based on the digital elevation model, the slope value θ of all raster pixels within the polygonal area of ​​the lower slope region is extracted. j And calculate the slope value θ of all raster cells. j The arithmetic mean of the values, where n is the total number of raster pixels in the lower slope, is used to calculate the slope of the lower slope. Slope lower = i j ; The difference between the average slope of the potential initiation zone and the average slope of the lower slope is calculated as the slope difference of the lower slope. A positive value indicates that the potential initiation zone is steeper than the lower slope, which is conducive to landslide initiation. The slope difference is calculated as follows: ΔSlope=Slope init Slope lower ; Based on the identified vector polygon of the lower slope, the number of raster cells M covered by the vector polygon is determined. pixel Multiply by the area A of a single cell pixel This is the projected area of ​​the slope below, and the specific calculation formula is as follows: Area lower =M pixel A pixel ; Along the main sliding direction vector of the landslide, the lower slope boundary (Boundary) lower Projecting all points onto this direction, the difference between the point with the maximum and minimum projection values ​​is calculated as the maximum straight-line distance along the direction of motion. Thus, the formula for calculating the length of the lower slope is: L lower = ∥proj dir (a) project dir (b)∥; Spatial overlay analysis is performed on the polygon of the lower slope area and the raster or vector map of the engineering geological rock group, and the rock group type code or name with the largest overlapping area is extracted as the engineering geological rock group of the lower slope. R lower =argmax r∈R (Area lower ∩Area); Extract the lowest point P at the leading edge of the lower slope. front Calculate P front To the nearest trench centerline channel The shortest spatial straight-line distance from any point q is used as the leading edge distance / channel distance. The specific calculation formula is as follows: D channel 。 9. The method according to claim 6, characterized in that, Calculate the acute angle between the overall direction vector of the downstream channel centerline and the overall sliding direction vector of the landslide, with a value of [0°, 90°]. Specifically, the overall sliding direction vector of the landslide is from the centroid of the potential initiation zone to the centroid of the deposition zone or the leading edge of the lower slope. This angle is taken as the angle between the downstream channel and the sliding direction. Select the downstream channel segment associated with the landslide, and determine the starting elevation H of this channel segment. start Elevation H at the endpoint end The difference between the two points and the horizontal projection distance L horizontal The ratio is used as the channel gradient, and the specific calculation formula is as follows: J=(H start H end) / L horizontal。 10. The method according to claim 6, characterized in that, Based on deformation monitoring data from interferometric radar, the deformation rate values ​​r of all pixels within the deformation zone are extracted. k The maximum value among them is taken as the maximum deformation rate of the deformation region. The specific calculation formula is as follows: R max= r k )； The arithmetic mean of the deformation rates of all pixels within the deformation zone is calculated as the average deformation rate. The specific calculation formula is as follows: = ； Based on the spatial distribution characteristics of deformation rate, spatial clustering analysis was used to identify areas with deformation rates significantly higher than the background value. Combined with the landslide element division results, the range of the initiation zone Z was determined. init The slope range Z below lower Accumulation area Z accum To determine whether the main deformation area belongs to the trailing edge, middle edge, or leading edge, the specific calculation formula is as follows: DeformPart=f cluster (D,Z init ,Z lower ,Z accum ); Based on the vector polygon of the deformation region, the major axis direction vector of the deformation region can be extracted using principal component analysis or the minimum bounding rectangle method. deform This direction is compared with the overall sliding direction vector of the landslide. slide The acute angle between them is used as the angle between the deformation zone and the sliding direction.

Images