Automatic identification method for popular shallow soil landslide disaster risk area

Abstract

The invention discloses an automatic identification method for a popular shallow soil landslide disaster risk area. The method comprises a landslide risk area identification method and an intra-area landslide threat house identification method. The method comprises the following steps: firstly, acquiring terrain, soil layer structure, vegetation and rock-soil distribution information of a region through a remote sensing technology, and performing data acquisition and processing by adopting remote sensing images, infrared images, laser radar data and synthetic aperture radar data to generate a high-precision digital elevation model and a geologic structure map; through data fusion and analysis, in combination with information such as soil layers, terrains and vegetation, a potential landslide risk area is identified, and houses in the landslide risk area are identified and evaluated. According to the method, the identification precision of the landslide risk area and the threatened house is improved, and efficient and accurate technical support is provided for early warning and prevention of landslide disasters. The system has the advantages of automation, high efficiency and accuracy, and can be widely applied to real-time monitoring, risk assessment and emergency management of landslide disasters.

Description

Technical Field

[0001] The present invention belongs to the field of natural disaster monitoring and early warning, and in particular is an automatic identification method for risk areas of group shallow soil landslide disasters. Background Art

[0002] Landslides, a typical geological hazard, typically occur in mountainous, hilly, or sloping areas and are highly sudden and destructive. With global climate change and intensified human activities, landslides are becoming increasingly frequent and devastating, severely impacting people's lives and property, and causing substantial economic losses and social impacts. Traditional landslide monitoring and assessment methods rely heavily on geological exploration, field surveys, and manual analysis. While these methods can assess landslide risk to a certain extent, they present challenges such as high workload, high cost, low efficiency, and poor accuracy when monitoring over large areas, complex terrain, and at high frequencies.

[0003] With the recent development of remote sensing, radar, and artificial intelligence technologies, landslide hazard monitoring and assessment methods based on remote sensing data and geographic information systems (GIS) have become a hot topic of research. These technologies enable real-time monitoring over large areas through various means, including drones, satellites, and aerial platforms. They provide precise information on topography, soil layers, vegetation, and rock and soil distribution, providing powerful technical support for landslide hazard identification and early warning. In particular, technologies such as synthetic aperture radar (SAR) and laser ranging (LiDAR) can generate high-precision three-dimensional ground data, enabling the identification of potential landslide risk areas and hazard sources.

[0004] However, the existing technology has the following main problems: Complexity of data collection and processing: Existing landslide hazard identification methods usually rely on multiple remote sensing methods and complex geological analysis. Errors are prone to occur during data processing and fusion, and it is difficult to achieve automation and efficient processing in large areas.

[0005] Limitations of landslide risk assessment: Most existing methods identify landslide areas through traditional geological surveys and analysis of remote sensing images. However, when dealing with landslide risks in large-scale and complex terrain, there is a lack of efficient risk assessment models and systems, which cannot accurately reflect the dynamic changes of landslide hazards.

[0006] Insufficient identification of houses threatened by landslides: In the prediction and early warning of landslide disasters, although landslide risk areas can be identified, there is often a lack of assessment of the degree of threat to specific buildings (such as houses) within the landslide threat range, making it impossible to provide timely and accurate information support for post-disaster emergency response. Summary of the Invention

[0007] To address the issues raised in the aforementioned background technology, the present invention provides an automated method for identifying clustered shallow soil landslide risk zones. This method not only rapidly identifies landslide risk zones by establishing specific identification rules and combining high-precision remote sensing data with digital height difference models, but also further identifies houses within landslide-threatened areas, providing more accurate disaster warnings and risk assessments. This method can significantly improve the accuracy of landslide warnings and response efficiency, providing a more scientific and practical technical approach for disaster prevention and control.

[0008] To achieve the above object, the present invention provides the following technical solutions: A method for automatically identifying clustered shallow soil landslide risk areas includes a method for identifying landslide risk areas and a method for identifying houses threatened by landslides within the area. The method for identifying houses threatened by landslides within the area uses a high-precision digital elevation model (DEM) and house vector data to automatically identify potentially threatened houses and delineate potential risk areas using geographic information system (GIS) software. The method includes the following steps: S1. Obtain high-precision DEM data and prepare housing vector data; S2. Generate slope, elevation, and aspect layers; S3. Mark the turning points of the houses, calculate the height difference and slope, and determine whether the slope cut threatens the house; S4. Generate extension lines of the house boundaries, analyze the relationship between the extension lines and the slope and aspect, and identify clusters of shallow soil landslides threatening the house. S5. Extract eligible houses as threatened house files and output them for subsequent on-site investigation and risk zone delineation.

[0009] Among them, the method for identifying houses threatened by landslides in the area generates slope and elevation layers from a high-precision DEM, extracts inflection points from house vector files, uses elevation and slope values to determine the threat of slope cutting and building, and marks inflection points of houses with elevation differences greater than 10 meters and slopes greater than 8 degrees; on this basis, potentially threatened houses are identified.

[0010] Among them, the method for identifying houses threatened by landslides in the area generates outward extension lines of the house boundaries and extracts the slope, slope direction and height difference information of each extension line. Combined with the distance between the slope foot and the house, the direction angle of the extension line and the slope conditions, the dangerous lines with landslide threats are identified; when a house has one or more dangerous lines, the house is determined to be a threatened house, and the corresponding threatened house vector surface data is generated.

[0011] The landslide risk area identification method includes the following steps: S1. Obtain information on regional topography, geology, slope structure, vegetation distribution, etc. through remote sensing technology and regional geological data; S2. Analyze vegetation types, distribution, size, and age using optical image analysis techniques to assess the anchoring effect of roots of different vegetation types on shallow soils. S3. Obtain local data from the database, including historical landslide data, meteorological data, soil moisture, precipitation, etc.; S4. Integrate and analyze the data from S1 to S3, use the risk assessment model to comprehensively assess the risk level of landslide disasters, and generate landslide risk areas and risk levels.

[0012] The remote sensing technology in step S1 includes visible light remote sensing images, infrared remote sensing images, lidar technology and synthetic aperture radar technology, which are used to obtain high-resolution terrain, soil layer and vegetation distribution information of the region.

[0013] Step S2 uses multispectral or hyperspectral images to evaluate the anchoring effect of vegetation roots on shallow soil, and calculates vegetation coverage and root strength through image classification and analysis technology.

[0014] The soil stability and landslide susceptibility are analyzed through terrain data, and the soil slope, slope aspect and soil moisture factors are calculated. The landslide risk heat map is generated by combining historical landslide event data and meteorological data to evaluate the landslide susceptibility of each area.

[0015] Step S4 includes the fusion of remote sensing images, radar data and optical images, and combines geographic information system technology and machine learning methods to use local relevant data in the database for model training to automatically evaluate the landslide disaster risk index of the region.

[0016] The risk assessment model is based on machine learning methods such as random forest, support vector machine or convolutional neural network. It is trained with historical landslide data and real-time data, and combined with meteorological data, geological data and historical landslide records in the database to predict the probability of landslide occurrence.

[0017] The landslide risk assessment results are a probability distribution map or heat map of landslide occurrence, and output landslide risk levels, which include low risk, medium risk, high risk, and extremely high risk. The risk assessment results are output after comprehensive analysis of historical landslide data, precipitation data, meteorological forecast data, etc.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. By combining a high-precision digital elevation model (DEM), building vector data, and advanced GIS technology, this method can accurately analyze factors such as regional topographic changes, slope, aspect, and elevation differences, effectively identifying potential landslide risk areas and threatened buildings. Compared to traditional methods that rely on manual on-site surveys, this method offers significant advantages in spatial coverage and accuracy, avoiding the high error rates associated with traditional methods. The automated process for generating identification results significantly improves data processing efficiency, particularly in large-scale applications. This reduces manual intervention and time consumption, speeds up identification, and allows for the identification of more potential risk areas and affected buildings in a shorter time.

[0019] By identifying threatened homes and landslide risk areas in advance, accurate data can be provided for subsequent disaster warnings and risk assessments. This data allows management departments to develop more precise emergency response and evacuation plans, minimizing casualties and property losses caused by landslides. This method, which does not rely on historical disaster sites or hidden danger reports, automatically identifies potential risk targets based on data such as home location and surrounding terrain, avoiding areas of risk that may be missed or underestimated using traditional methods. Precisely calibrating the threat posed to each home improves the accuracy and relevance of disaster management.

[0020] Through GIS-based automated analysis, this method can identify and delineate potential disaster areas in advance, reducing the workload of on-site surveys. Traditional manual surveys can be cumbersome and inefficient, especially in large mountainous areas or disaster-prone areas. This method significantly improves survey efficiency and reduces personnel costs. By identifying threatened homes in advance, this method can optimize disaster warning processes, avoid ineffective evacuations or emergency response operations, and thus reduce costs and resource waste for management departments.

[0021] This method integrates high-precision DEM data, building vector data, and analysis of multiple factors such as slope and aspect to comprehensively assess landslide potential without relying on historical data or a single assessment dimension. This not only enhances understanding of landslide mechanisms but also improves the accuracy of risk assessment, avoiding important geological factors that may be overlooked by traditional methods. By identifying different types of hazards, such as clustered shallow landslides and slope-cut construction, this method can analyze different types of landslide threats separately, avoiding the bias caused by categorizing all risks as the same hazard, and ensuring comprehensiveness and accuracy of the assessment results.

[0022] 2. Traditional methods typically rely on static geological survey data and historical information for landslide risk assessment. This approach fails to fully consider the temporal and spatial variations of environmental factors. This is especially true when dynamic factors such as precipitation are involved, making traditional methods unable to respond to environmental changes in real time.

[0023] The present invention uses high-resolution remote sensing images and radar data, as well as dynamic analysis of real-time meteorological data, to perform refined spatiotemporal forecasts and capture changes in different time periods and regions. This makes the assessment results not only more accurate, but also reflects the dynamic characteristics of landslide risks as they change over time and in the environment. The occurrence of landslides can be predicted under conditions of changes in factors such as precipitation and temperature fluctuations. This dynamic forecasting method has strong real-time and responsiveness, and can help to respond in advance. Especially under extreme climatic conditions, the present invention can accurately predict high-risk periods for landslides, provide specific time periods and spatial regions, and provide precise guidance for disaster emergency response.

[0024] 3. Traditional landslide risk assessment methods often ignore the impact of vegetation roots on soil stability and landslide occurrence, or consider vegetation only as surface cover. They lack in-depth analysis of the anchoring role of vegetation in soil stability.

[0025] This technology uses optical imaging to analyze vegetation types, distribution, size, and age, assesses the strength and stability of plant roots, and analyzes landslide risk based on the interaction between plant roots and the soil layer. This not only considers the surface coverage of vegetation but also deeply analyzes the reinforcement effect of plant roots on the soil layer. This technology can effectively distinguish the different effects of different types of vegetation (such as herbaceous plants and trees) on soil stability. This is particularly important for risk assessment in mountainous and hilly areas, helping to accurately identify potential landslide areas.

[0026] 4. Traditional methods often use crude regional divisions (e.g., broad high-risk, medium-risk, and low-risk areas) to assess landslide risk. This lacks detailed and personalized analysis, and fails to meet the risk management needs of diverse regions and environments. This invention, through comprehensive analysis of multiple dimensions (topography, slope structure, vegetation, meteorology, etc.), establishes a personalized risk assessment model for each specific area, enabling refined landslide risk classification and management. This refined risk assessment provides more specific and actionable decision support.

[0027] 5. Traditional landslide monitoring and risk assessment methods often require the deployment of a large number of sensors, underground detection equipment, or monitoring systems. This not only requires a large amount of hardware equipment, but also involves complex installation and maintenance work. This is especially true in inaccessible or dangerous mountainous environments, where the deployment of sensors is difficult and risky. The present invention achieves high efficiency and accuracy in landslide risk assessment through non-invasive data acquisition methods such as remote sensing technology, radar data, optical imaging, and meteorological data. These technologies can obtain information on the environmental and geological characteristics of a region without the need for a large number of sensors. Through remote sensing methods such as drones and satellite images, data can be acquired simultaneously over a large area, avoiding the need for a large number of on-site equipment and sensor deployments, reducing the cost of project implementation, and avoiding complex construction and equipment maintenance issues. Large areas can be quickly covered, especially in areas that are difficult to reach by manpower, such as mountainous areas and canyons. There is no need to worry about sensor failures or errors.

[0028] 6. In traditional landslide monitoring, especially when relying on sensor deployment and manual surveys, human interference cannot be ignored. For example, factors such as sensor installation deviations, operator errors, and environmental misjudgments can all affect data accuracy and reliability. This present invention reduces human interference and provides more objective and stable data. This data is largely processed automatically, reducing the potential for human error. This provides a higher level of automation, especially when collecting data over large areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a flow chart of the method for identifying houses threatened by landslides in an area according to the present invention; Figure 2 Schematic diagram of the flow of the landslide risk area identification method of the present invention. DETAILED DESCRIPTION

[0030] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0031] See also Figure 1 and 2 , an automatic identification method for risk areas of group shallow soil landslide disasters, including a landslide risk area identification method and a landslide-threatened house identification method in the area.

[0032] See also Figure 2 ,Specifically, the landslide risk area identification method includes the ,following steps: S1. Obtain regional topography, soil structure, vegetation, and rock and soil distribution information through remote sensing technology 1. Remote sensing data collection Remote sensing technology collects various surface and underground information through remote detection. Data can be collected using sensors carried by satellites, drones, and aerial platforms. For information on topography, soil structure, vegetation, and rock and soil distribution, the following methods are used: 1.1 Obtaining visible light remote sensing images Use high-resolution satellite remote sensing images or RGB optical cameras mounted on drones. If higher precision is required, use aerial image acquisition platforms (such as high-resolution aerial cameras).

[0033] Data processing and analysis: Preprocessing remote sensing images through radiometric and atmospheric corrections eliminates the effects of atmospheric and lighting conditions. Image classification algorithms (maximum likelihood classification, support vector machines, etc.) are used to extract soil structure, vegetation cover, and terrain features from the images.

[0034] 1.2 Acquiring infrared remote sensing images Use satellites with infrared sensors, drones (with infrared sensors on board), or aerial platforms.

[0035] Data processing and analysis: Extracting areas of high reflectivity from infrared images generally indicates important information such as vegetation, soil moisture, and soil temperature. Combined with vegetation indices (NDVI, EVI, etc.), soil moisture and vegetation health are analyzed, indirectly assessing soil stability.

[0036] 1.3 LiDAR technology to obtain high-resolution terrain data Use LiDAR (LiDAR) equipment (such as airborne LiDAR and drone LiDAR systems) to obtain three-dimensional information about the ground, vegetation, and buildings. LiDAR emits a laser beam and measures the reflection time to accurately determine the height and shape of the target object. This allows for the extraction of details such as topography, slope, and surface features.

[0037] Data processing and analysis: Point cloud data processing software is used to generate high-precision digital elevation models (DEMs) and digital surface models (DSMs). Parameters such as slope, aspect, and surface roughness are extracted from LiDAR data. These parameters are essential for assessing landslide hazard.

[0038] 1.4 Synthetic Aperture Radar (SAR) Technology Continuous ground monitoring using SAR satellites. SAR can penetrate clouds, rain, fog, and other adverse weather conditions, enabling it to obtain ground information. SAR technology is particularly suitable for monitoring soil subsidence, surface deformation, landslide activity, and changes in water systems.

[0039] Data Processing and Analysis: Interferometric Synthetic Aperture Radar (InSAR) technology uses multi-temporal SAR imagery to perform differential analysis and extract ground deformation information. This technology is particularly effective for monitoring landslides and soil subsidence. Radial Basis Function (Kriging) and multi-temporal analysis methods can help identify subtle changes in soil layers and further assess landslide risk areas.

[0040] 2. Data processing and fusion Through the collaborative work of multiple remote sensing technologies, comprehensive information on soil layers, topography, rock and soil distribution, and vegetation can be obtained. Further data fusion and processing are required: 2.1 Data Fusion Data from various sources (satellite remote sensing imagery, LiDAR point clouds, SAR imagery, etc.) is spatially registered and fused within a GIS platform. After fusion, the information is stored in raster and vector formats for easy analysis and presentation. For example, combining terrain data from LiDAR with soil and vegetation information from remote sensing imagery can better identify potential landslide-prone areas.

[0041] 2.2 Soil structure analysis Combining SAR data, infrared remote sensing data, and high-precision topographic maps using LiDAR technology, we analyze the regional soil structure, specifically its thickness, density, and type. By comparing different geological data and using remote sensing data to identify geological faults and geotechnical structures, we generate geotechnical type maps, providing a crucial basis for subsequent landslide risk assessments.

[0042] 2.3 Vegetation distribution and soil stability assessment Remote sensing imagery and spectral data are used to analyze vegetation coverage and root distribution within a region. Combined with surface relief information obtained through LiDAR or SAR technology, this analysis examines whether vegetation effectively anchors the soil and prevents soil sliding. Vegetation indices (NDVI, EVI, etc.) are combined with indicators such as relative soil stability and soil moisture to assess the stabilizing effect of vegetation within the region.

[0043] 3. Data Output Topographic data: slope, aspect, surface deformation, etc.

[0044] Soil layer data: soil structure, soil layer thickness, moisture, etc.

[0045] Vegetation data: vegetation coverage, NDVI value, root density, etc.

[0046] These data will serve as basic input for subsequent steps to support the analysis and prediction of landslide risks.

[0047] S2. Use radar technology to obtain information on soil faults, soil density distribution, underground cavities, and water systems Radar technology (including ground-based, airborne, or satellite-based radar) is used to survey the target area, acquiring information on soil faults, density distribution, underground cavities, and water systems. This data is crucial for assessing soil stability, landslide risk, and groundwater flow. This process involves adjusting radar parameters such as frequency and power to ensure accurate data at varying target depths.

[0048] 1. Select the appropriate radar technology Select the appropriate radar technology based on the size of the detection area and the depth of the soil layer: Ground radar: Suitable for smaller areas, with a detection depth of usually 0-30 meters. It has high spatial resolution and is suitable for detecting soil faults and underground cavities.

[0049] Aerial radar: Suitable for larger-scale detection, especially for determining soil density distribution and obtaining water system information over large areas. Aerial radar operates at a lower frequency and can detect depths exceeding 50 meters.

[0050] Satellite radar: Suitable for deep detection in large areas, capable of detecting depths from 10 meters to hundreds of meters, suitable for large-scale soil density distribution and macro-monitoring of water systems.

[0051] 2. Radar technical parameter adjustment According to the target depth and specific needs of detection, adjust the radar parameters such as frequency and power to ensure that soil layer information at different depths can be effectively detected.

[0052] Frequency selection: High frequency (500 MHz - 1 GHz): Suitable for shallow detection (0 - 10 meters), with higher resolution, capable of providing clearer detailed information such as faults and underground cavities.

[0053] Low frequency (50 MHz - 500 MHz): Suitable for deeper detection (10 - 50 meters). Although the resolution is low, it has strong penetration ability and is suitable for obtaining soil density distribution and deep water system information.

[0054] Power adjustment: High power (30W - 50W): Increases signal strength, improves penetration depth, and is suitable for deep detection.

[0055] Low power (5W - 20W): Suitable for shallow areas, reduces interference signals and noise, and improves signal quality.

[0056] Scan Mode: Single scan: used to quickly scan a small area, such as detecting a specific soil fault or underground cavity.

[0057] Continuous scanning: Suitable for monitoring large areas, generating comprehensive soil density distribution maps and water system distribution maps.

[0058] 3. Data Collection Use radar for ground detection and data collection at appropriate frequency and power settings. The following data are mainly collected: Soil fault information: The distribution and depth of soil faults are determined by the time and intensity of radar wave reflections. Radar waves reflect differently when encountering different media, and by analyzing the delay time and intensity of the echoes, the boundaries and location of the faults can be determined.

[0059] Soil density distribution: The reflection intensity of radar waves is related to the dielectric constant of the soil. The density and moisture content of the soil significantly affect the propagation of radar waves. Based on the intensity of the reflected signal, the density distribution of the soil layer can be estimated, and thus the stability of the soil layer can be analyzed.

[0060] Underground cavities: The presence of underground cavities can cause abnormal radar reflections, manifesting as weak signal echoes or abnormal echo times. Multiple scans and comparisons at different frequencies can accurately locate the distribution and depth of underground cavities.

[0061] Water system information: Radar waves will produce obvious reflections in water bodies. By analyzing the reflected signals of the water system, we can obtain information such as the distribution of groundwater flow, water level, flow direction, etc.

[0062] 4. Data Processing and Analysis The data collected by the radar needs to be processed and analyzed, including the following steps: Data preprocessing: De-noising, smoothing, filtering and other processing are performed on the collected raw radar data to eliminate unnecessary interference signals.

[0063] Time and frequency domain analysis: Time domain analysis: Analyze the arrival time and duration of the echo signal to infer the depth, thickness and fault conditions of the soil layer.

[0064] Frequency domain analysis: By analyzing the spectrum of the reflected wave, the electromagnetic characteristics of the soil layer are extracted, and then the density distribution of the soil layer and the water system conditions are analyzed.

[0065] Inversion and Modeling: Using inversion algorithms, radar wave reflection data is used to generate a two-dimensional or three-dimensional model of the soil layer, accurately depicting the distribution of underground structures. Common inversion methods are used, including least squares and genetic algorithms.

[0066] Underground cavity detection: Detecting underground cavities uses the reflection difference method to identify them. The reflection intensity and signal propagation time in the cavity area are significantly different. Through comparative analysis, the cavity boundary and location can be accurately determined.

[0067] Soil density analysis: Soil density can be estimated by analyzing the relative propagation speed and reflection intensity of radar waves. Areas with higher reflection intensity generally indicate denser soil layers, while areas with lower reflection intensity may indicate looser soil layers.

[0068] 5. Result Output Soil fault location and direction: provide fault information for subsequent landslide risk assessment.

[0069] Soil layer density distribution: reflects soil layer stability and guides subsequent soil stability analysis.

[0070] Distribution of underground cavities: Identify potential landslide triggers, especially underground cavities that may cause geological instability.

[0071] Water system information: provides data support for further analysis of hydrological impacts, precipitation impacts, etc.

[0072] These data will serve as input for subsequent landslide risk assessments and help build accurate landslide risk models.

[0073] S3. Analyze vegetation type, distribution, size, and age characteristics using optical image analysis techniques to assess the anchoring effect of roots of different vegetation types on shallow soils. Optical image analysis technology (multispectral or hyperspectral image analysis) is used to classify and analyze vegetation in the area, extract information such as vegetation type, distribution, size and age, and evaluate the anchoring effect of vegetation roots on shallow soil based on these characteristics, providing auxiliary support for landslide risk assessment.

[0074] 1. Optical Image Acquisition and Processing Optical images (including multispectral and hyperspectral images) can be obtained through remote sensing platforms (such as satellites and drones). Different bands correspond to different vegetation characteristics, such as the red band and the near-infrared band.

[0075] Multispectral imagery: Typically contains red, green, blue, and near-infrared bands. Commonly used for vegetation analysis and classification.

[0076] Hyperspectral imagery: Acquires data in multiple continuous bands, providing more detailed information than multispectral imagery and suitable for accurate vegetation analysis.

[0077] The acquired optical images are pre-processed before further analysis, including radiation correction, atmospheric correction, and geometric correction.

[0078] NDVI (Normalized Difference Vegetation Index): ; Among them, NIR: reflectance of near infrared band; Red: reflectance of red band; NDVI value reflects the growth of vegetation, and the value ranges from -1 to 1, where a value close to 1 indicates good vegetation growth. Vegetation coverage calculation: Use the NDVI index to estimate the vegetation coverage of different areas. Vegetation coverage It is calculated by classifying the pixels in the image.

[0079] ;in, : vegetation coverage (%); Number of pixels in vegetation area: the number of pixels classified as vegetation; Total number of pixels: the total number of pixels in the image; 2 Tree age and species analysis Estimating tree age generally relies on the growth characteristics of trees and the chlorophyll content of vegetation. Hyperspectral imagery analyzes the reflectance of different bands to extract information such as chlorophyll content, leaf area index (LAI), and leaf thickness. These parameters are closely related to tree age. Using the spectral reflectance characteristics of vegetation, the following methods are applied: Chlorophyll content estimation formula (based on hyperspectral reflectance data): ;in, Red band reflectivity; Near-infrared band reflectivity; The regression coefficient, obtained by fitting the experimental data, shows that chlorophyll content is closely related to factors such as tree age, tree health, and soil moisture. Therefore, the age of the tree can be inverted by the chlorophyll content in the image.

[0080] 3. Assessment of the anchoring effect of vegetation roots on shallow soil The anchoring effect of vegetation roots on shallow soil is a crucial factor in preventing landslides and soil erosion. Root anchoring is influenced by a variety of factors, including tree species, tree age, root depth, and root density. While optical imaging cannot directly observe root distribution, it can indirectly infer root growth characteristics and soil stability. Root anchoring is assessed using the following methods: Root strength assessment: root strength of vegetation It is closely related to the distribution density of the root system and the health of the tree. The root strength of the vegetation can be estimated by its growth, using the following relationship: ;in, : root strength (dimensionless); LAI: leaf area index (m 2 / m 2 ); Tree age: tree age (years); : Empirical constant, adjusted according to different tree species and soil types; Vegetation root density and soil stability: Root density It is usually positively correlated with soil stability and can be evaluated using the following formula: ;in, Root density (kg / m 2 ); Root mass: total mass of vegetation roots (kg); Area: area of the analysis region (m 2 ); The greater the root density, the stronger the anchoring effect of vegetation on the soil, the higher the stability of the soil, and the lower the probability of landslides.

[0081] Soil stability calculation: root density and soil shear strength ( ). It can be evaluated by the soil shear strength formula: ;in, Soil shear strength ; Soil cohesion (Pa); Normal pressure of soil (Pa); internal friction angle of soil (°); : root adhesion to soil (dimensionless); 4 Output Vegetation coverage table: records the vegetation coverage ratio in each study area. GIS technology can be used to generate heat maps or coverage maps.

[0082] Root strength data table: records the root strength of each type of vegetation and performs weighted calculations based on factors such as plant species and tree age.

[0083] Plant species and age data: Based on remote sensing image analysis and plant classification data, the species and ages of the main plants in the area are output.

[0084] Vegetation health status map: Draw a vegetation health status distribution map based on NDVI values.

[0085] This data will provide important input for subsequent landslide risk assessments, helping the system identify areas that may be at higher risk of landslides due to vegetation factors.

[0086] S4. Utilize locally relevant data in the database, including historical landslide data, meteorological data, soil moisture, precipitation, and seismic activity.

[0087] The goal of this process is to calculate the landslide risk level in different areas by combining historical records with real-time data, and to provide a basis for early warning and prevention of landslide disasters.

[0088] 1. Data Source and Data Preprocessing 1.1 Data Source: Historical landslide data: This includes detailed information on the time, location, cause, impact area, type, and volume of historical landslide events. This data typically comes from government departments, research institutions, and regional disaster record systems.

[0089] Meteorological data: including meteorological conditions such as precipitation, temperature, wind speed, humidity, etc., usually provided by weather stations, satellite remote sensing data and weather forecast systems.

[0090] Soil moisture data: Soil moisture data can be obtained through ground-based meteorological stations, remote sensing satellites, or soil moisture sensors.

[0091] Precipitation data: are obtained through weather stations and remote sensing techniques and typically include the spatial and temporal variations of monthly, seasonal, and annual precipitation.

[0092] 1.2 Data preprocessing: In order to ensure the validity and consistency of the data, the above data needs to be preprocessed as follows: Data cleaning: remove abnormal data, duplicate data, and fill or delete missing values.

[0093] Data standardization: When the dimensions of different data are inconsistent, standardization is required to facilitate comparison and model training.

[0094] Time synchronization: aligning data at different time scales (e.g., meteorological data, historical landslide events, real-time earthquake data, etc.) for joint analysis.

[0095] 2. Feature extraction and landslide risk factor construction After acquiring and cleaning the data, it is necessary to extract the key factors that affect landslide occurrence and construct appropriate features for landslide risk assessment. These factors include but are not limited to: Slope: Terrain slope is a key factor in landslide occurrence. The steeper the slope, the greater the likelihood of a landslide. Slope can be calculated from remote sensing imagery or terrain data such as digital elevation models.

[0096] Soil moisture: The higher the soil moisture, the weaker the soil's shear strength and the greater the likelihood of landslides. Soil moisture can be obtained through meteorological data, remote sensing data, and soil sensors.

[0097] Precipitation: The impact of precipitation on landslides is obvious. Excessive precipitation saturates the soil with water, which can lead to soil mobility and trigger landslides. Precipitation data can be analyzed based on historical and real-time meteorological data.

[0098] Historical landslide event data: Historical landslide events can help assess the landslide susceptibility of a region and understand the patterns of landslide occurrence, their temporal and spatial distribution patterns, and their relationship with meteorological, earthquake, soil and other factors.

[0099] 3. Output and decision support Meteorological data table: includes fields such as time, location, precipitation, temperature, and humidity.

[0100] Historical landslide event records: a database that includes information on the specific time, location, scale and type of landslides.

[0101] These data provide detailed environmental context and potential triggering factors for subsequent steps, can help determine which areas are most vulnerable to landslides, and provide support for the prediction and prevention of landslide risks.

[0102] S5. Data fusion and final risk assessment Data Fusion: The key to S5 lies in the effective integration of data from various sources (topographic data, radar data, vegetation data, meteorological data, etc.). Each data type provides a different perspective on landslide risk factors, such as terrain slope, soil structural stability, the impact of vegetation roots on the soil, and the triggering effects of meteorological factors. By integrating this data, we can obtain a more comprehensive and accurate landslide risk assessment model.

[0103] Risk Assessment Model Construction: Using machine learning or statistical methods, we combine the weights and influences of different data sources to build a multi-dimensional landslide risk prediction model. This model calculates the probability of landslide occurrence based on a combination of various factors and categorizes them according to risk level.

[0104] The specific steps are: 1. Data preprocessing and standardization: Ensure that data from different sources are comparable to prevent data discrepancies from affecting the final results.

[0105] Missing value processing: For missing geographical, meteorological or other data, interpolation, mean filling and other methods are used.

[0106] Data standardization: Standardize data of different scales and units to make them on the same scale to prevent the deviation of a certain data source from affecting the final model.

[0107] For example, remote sensing image data (image data) and meteorological data (numerical data) are standardized to ensure that they have equal influence in the model.

[0108] Data smoothing: Use filtering methods to smooth data, remove outliers and noise, and improve data quality and accuracy.

[0109] 2. Data fusion: combining multiple data sources (remote sensing images, radar data, optical images, meteorological data, etc.) into a unified, multi-dimensional dataset.

[0110] Including spatial data fusion: integrating spatial data such as terrain, radar data, vegetation distribution, etc., and integrating these data into a multi-level spatial model through the geographic information system (GIS).

[0111] For example, data such as vegetation cover, slope, soil type, humidity, and precipitation can be combined according to geographical regions to generate comprehensive risk characteristics for each region.

[0112] Time series data fusion: For meteorological data and historical landslide data, time series analysis can be combined with the time nodes of landslide occurrence to perform data fusion and identify the temporal and spatial distribution patterns of landslides.

[0113] 3. Feature Engineering: Extracting features related to landslide risk from different datasets.

[0114] Topographic features: data such as slope, aspect, and terrain curvature.

[0115] Soil characteristics: soil moisture, soil layer structure, soil type, etc.

[0116] Vegetation characteristics: vegetation coverage, root strength, vegetation health, root distribution depth, etc.

[0117] Meteorological characteristics: precipitation, temperature, humidity, air pressure, etc.

[0118] Characteristics of historical landslides: frequency, location, impact range, etc. of historical landslides.

[0119] 4. Model selection and training: Use appropriate machine learning models to predict landslide risks and determine risk levels.

[0120] Optional, Supervised learning algorithm: Use historical data of known landslide occurrences and non-occurrences to train the model. Commonly used models include: Random Forest: Random forest is constructed through decision trees and can effectively process multi-dimensional and nonlinear data. It is suitable for risk prediction of geological and meteorological data.

[0121] Support vector machine: Support vector machine can effectively classify complex boundary conditions and is suitable for delineating landslide risk areas.

[0122] Neural networks: Convolutional neural networks, in particular, can be used to process image data (such as remote sensing images) and extract complex spatial features.

[0123] Gradient Boosting Machine (GBM): GBM can optimize the selection of features and is suitable for accurate landslide risk prediction.

[0124] Unsupervised learning algorithm: For unlabeled data, clustering methods can be used to analyze the risk distribution within the region and divide the risk levels.

[0125] 5. Risk level assessment: Based on the output of the model, the risk level of the landslide is determined.

[0126] Probability threshold method: According to the probability of landslide occurrence output by the model, different thresholds are set to define the risk level. For example: Low risk: The probability of landslide is less than 30%.

[0127] Medium risk: The probability of landslide is between 30% and 60%.

[0128] High risk: The probability of landslide is between 60% and 90%.

[0129] Very high risk: The probability of a landslide occurring is greater than 90%.

[0130] Multi-factor weighted approach: Different factors (such as topography, vegetation, and weather) are assigned different weights and a comprehensive landslide risk value is derived through weighted summation. This comprehensive value is then used to determine the risk level.

[0131] 6. Risk area map generation: The risk assessment results are presented in the form of a map showing the landslide risk level of each area.

[0132] Using GIS technology, the risk levels of different areas are visualized as layers, and different risk levels are distinguished by color.

[0133] The spatial distribution and intensity of risks are displayed through heat maps, contour maps, etc.

[0134] Output regional maps of different risk levels for reference by decision makers in planning, building design, disaster prevention, etc.

[0135] S5 output: Landslide risk level (low risk, medium risk, high risk, extremely high risk): Different risk levels are divided according to the probability and risk indicators predicted by the model.

[0136] Landslide risk area map: Based on the model's risk assessment, a landslide risk distribution map is generated for each region. This map can be used as a reference for landslide monitoring, land use planning, disaster prevention, etc.

[0137] Risk assessment report: records the risk assessment results of each area in detail, including risk level, main influencing factors, predicted probability of landslide occurrence, etc., for reference by decision makers.

[0138] S5 relies on data collected from phases S1 through S4. By integrating data fusion and machine learning algorithms, and taking into account multiple factors such as topography, soil, vegetation, meteorology, and historical landslides, S5 accurately assesses landslide risk and generates a risk level map. The resulting risk area map will provide a valuable reference for engineering projects, disaster prevention, and emergency response.

[0139] See also Figure 1 ,Furthermore, after identifying the landslide risk area, if there are houses in the risk area, ,the landslide-threatened houses in the risk area are further ,identified as follows: 1. Data Collection and Preprocessing First, it is necessary to collect and process relevant basic data, including high-precision digital elevation models (DEMs) and house vector data. The quality and accuracy of this data directly affect the recognition results.

[0140] 1.1 High-precision DEM data Data source: Obtain elevation data of the target area through remote sensing technology or existing digital elevation data (such as LiDAR data, SRTM data, etc.).

[0141] Accuracy requirement: DEM accuracy is required to be no less than 2 meters or 5 meters, depending on the terrain characteristics of the target area.

[0142] Data preprocessing: DEM data is preprocessed using GIS software, including noise removal, filling in missing values, and correcting terrain anomalies.

[0143] 1.2 House vector data Data source: Obtain house vector data from satellite remote sensing images, drone aerial photography, or existing geographic information databases.

[0144] Data preprocessing: Convert house vector data into a format suitable for analysis in conjunction with DEM data to ensure that house boundaries are accurate.

[0145] 2. DEM data analysis and layer generation Use GIS software (such as ArcGIS, QGIS, etc.) to analyze high-precision DEM and generate slope, elevation difference and aspect layers related to house vector data.

[0146] 2.1 Generate slope and height difference layers Slope layer: Generates a slope layer by calculating the slope (i.e., the slope of elevation change) of each grid cell in the DEM data.

[0147] Height difference layer: Calculate the height difference between each grid cell in the DEM and the adjacent grid cells to obtain the height difference distribution within the area.

[0148] 2.2 Generate aspect layer Aspect layer: Based on DEM data, the aspect of each grid cell, that is, the direction of the slope surface, is calculated to help identify areas where landslides may occur.

[0149] 3. Identification of houses threatened by slope cutting and building 3.1 House inflection point extraction The inflection points of the house (i.e., the corners of the house boundary) are extracted from the house vector data, and the elevation difference and slope information within a 20-meter radius around each inflection point are extracted.

[0150] 3.2 Height difference and slope analysis For each house turning point, calculate the elevation difference of the point (i.e. the maximum elevation within a 20-meter radius of the point minus the elevation of the turning point) and the slope of the point.

[0151] Threshold setting: If the slope is greater than 8 degrees and the height difference is greater than 10 meters, the point is marked as a threatened point.

[0152] House screening: If there is one or more marking points at the turning point of a house, it means that the house is threatened by slope cutting and building, and the house is extracted as a threatened house.

[0153] 4. Identification of houses threatened by clustered shallow soil landslides 4.1 Extension Line Generation House boundary extension line: Generate straight lines extending outward based on the boundary of the house vector data, with one extension line generated every 10 meters. Four extension lines are generated at the turning points of the house, two lines perpendicular to the boundaries on both sides and another two lines dividing the house into three equal parts.

[0154] Extension line length: Set the extension line length to 50 meters.

[0155] 4.2 Analysis of extension line slope and aspect For each extension line, the slope, aspect, and elevation difference information of each grid it passes through are calculated. In particular, the slope changes in the area where the extension line passes through are used to identify potential landslide hazard areas.

[0156] 4.3 Danger Line Determination Slope foot calculation: Based on the extension line, determine whether there is a steep slope with a slope greater than 25 degrees (i.e., slope foot) in the area near the house, and calculate the distance from the slope foot to the house.

[0157] Calculation of direction angle and slope angle: Calculate the direction angle of the extension line and calculate the angle with the slope direction of the grid through which the extension line passes to determine whether the conditions for landslide occurrence are met.

[0158] Determination of landslide conditions: The distance between the slope foot and the house is less than 10 meters; The angle between the direction of the extension line and the slope aspect is less than 20 degrees; The extension line passes through an area with a slope greater than 25 degrees for a length exceeding 40 meters, and an area with a slope greater than 35 degrees for a length exceeding 20 meters; The total height difference of the extension line is greater than 30 meters.

[0159] Danger line marking: When an extension line meets all the above conditions, it is determined to be a danger line.

[0160] 4.4 Threatened House Extraction Danger line screening: If a house is passed by at least one danger line, the house is determined to be a threatened house and the threatened house vector data is generated.

[0161] 5. Output and risk assessment 5.1 Recognition result output The identification results are output as two files: a vector surface file of threatened houses and a corresponding danger line file for subsequent on-site investigation and risk zone delineation.

[0162] 5.2 Risk assessment and further analysis Based on the data of identified threatened houses, further disaster risk assessment is carried out. Combined with historical landslide data, meteorological data, etc., the risk level of each area is calculated, and a landslide risk heat map or probability distribution map is generated to guide disaster prevention and mitigation work.

[0163] 6. System Integration and Automation Integrate the above analysis process into an automated system and realize automated operation through GIS platform (such as ArcGIS, QGIS), reduce manual intervention, and improve processing efficiency and accuracy.

[0164] The system implements automated data loading, layer generation, analysis and calculation, and result output, supports batch processing, and is suitable for identifying houses threatened by landslides in large areas.

[0165] The following are representative application cases and processes based on the technology developed by the present invention, which can effectively verify the actual effect of the present invention in landslide risk assessment.

[0166] Case 1: Automated identification of landslide risk areas in mountainous areas A certain mountainous area has been under the threat of landslides for a long time and urgently needs to conduct risk assessment using remote sensing technology.

[0167] step: Remote sensing data collection: High-resolution satellite remote sensing imagery is used to obtain mountain terrain information. Unmanned aerial vehicles (UAVs) equipped with RGB optical cameras and infrared sensors are used to obtain vegetation coverage and soil moisture data. LiDAR technology is used to collect three-dimensional terrain data and generate digital elevation models (DEMs) and digital surface models (DSMs) to identify slopes and terrain features in mountainous areas.

[0168] Data processing and fusion: Utilizing a GIS platform, we spatially register and fuse satellite remote sensing images, LiDAR point clouds, and SAR images. We analyze soil structure, vegetation coverage, and soil moisture to identify potential landslide-prone areas.

[0169] Identification of houses threatened by landslides: Based on the identified landslide risk areas and combined with the existing house vector data in the Geographic Information System (GIS), it is determined which houses are located in high-risk landslide areas.

[0170] Result: High-risk landslide areas in the mountainous area were successfully identified and threatened houses were marked, helping the government to deploy disaster prevention measures in a timely manner and avoid potential losses.

[0171] Case 2: Landslide risk identification around cities There are hillside areas around a certain city. Soil cracks have recently appeared in some areas, and a landslide risk assessment is needed.

[0172] step: Remote sensing data collection: High-resolution satellite imagery is used to obtain topographic information around the city, combined with infrared remote sensing imagery to analyze soil moisture and identify potential landslide areas. Synthetic aperture radar (SAR) technology is used to monitor subtle changes in the soil layer and identify whether there is subsidence or deformation.

[0173] Data fusion and processing: LiDAR data, SAR imagery, and infrared remote sensing imagery are combined to analyze the soil structure and assess soil stability. The NDVI (Natural Dispersion Index) is used to determine whether vegetation cover is sufficient to prevent landslides.

[0174] Identification of houses threatened by landslides: Based on remote sensing data and geographic information systems, determine whether surrounding houses are in potential landslide areas and assess the possibility of house damage.

[0175] Results: This method accurately identified high-risk landslide areas around the city, assessed threatened houses, and issued timely warnings, ensuring the safety of people and property.

[0176] Case 3: Landslide risk assessment for engineering construction in mountainous areas Infrastructure construction is underway in a mountainous area, which poses a landslide risk and requires a landslide threat assessment.

[0177] step: Remote sensing data acquisition: UAVs equipped with LiDAR systems conduct high-precision terrain scanning to generate digital elevation models (DEMs) for analyzing the slope, aspect, and surface characteristics of the area. Infrared remote sensing imagery is used to assess soil moisture, and SAR technology is used to monitor ground deformation and soil subsidence.

[0178] Data Fusion and Analysis: Remote sensing images, LiDAR data, and SAR data are integrated to identify landslide-prone areas within the region. Landslide risk in the construction area is assessed based on soil structure and vegetation distribution, combined with the location of the project.

[0179] Identification of buildings threatened by landslides: Using a geographic information system (GIS) to overlay known building information with risk areas, we can identify potentially affected buildings.

[0180] Results: Through this method, the landslide risk in the construction area was successfully assessed, and design adjustment suggestions were provided to the construction unit to avoid potential landslide disasters.

[0181] Case 4: Landslide Risk Assessment in Rural Mountainous Areas Small-scale landslides have occurred continuously in a rural mountainous area. It is necessary to assess the landslide risk in the area and propose preventive measures.

[0182] step: Remote sensing data acquisition: High-resolution satellite remote sensing images and drone remote sensing are used to obtain data on topography, soil structure, and vegetation distribution. SAR technology is used to detect subtle surface deformations and analyze soil layer changes.

[0183] Data Fusion and Analysis: Remote sensing imagery, LiDAR data, and SAR data are integrated to analyze geological faults and assess soil stability. Factors such as soil moisture, vegetation health, and slope are analyzed, combined with historical landslide data in the region, to determine the landslide risk level.

[0184] Identification of houses threatened by landslides: Using house vector data and landslide risk areas, we can identify threatened farmhouses and residential areas.

[0185] Results: High-risk landslide areas in the region were successfully identified, and the threat of landslide hazards to houses and people was reduced through land use planning and engineering construction measures.

[0186] Case 5: Landslide disaster warning near the coastline The low-lying areas near a certain coastline are facing landslide risks, and landslide disaster early warning work needs to be carried out.

[0187] step: Remote sensing data collection: SAR and LiDAR technologies are used to obtain coastal terrain information and analyze soil structure and moisture. Infrared remote sensing imagery is also used to assess soil moisture and vegetation distribution.

[0188] Data processing and fusion: SAR imagery, LiDAR point cloud data, and infrared remote sensing imagery are integrated to analyze the soil stability of the coastline. Landslide risk in the area is comprehensively assessed based on topographic characteristics, soil moisture, and vegetation cover.

[0189] Identification of houses threatened by landslides: Overlay house vector data with high-risk landslide areas to identify residential houses in risk areas and provide emergency evacuation recommendations based on the data.

[0190] Results: The proposed method effectively predicted landslide risk areas near the coastline and successfully identified threatened houses, providing strong decision-making support for the local government.

[0191] Through these actual cases, we can see the effectiveness of this automated identification method in different geographical environments and application scenarios, and it can provide accurate data support for landslide disaster prevention and emergency management.

[0192] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for automatically identifying risk areas of group shallow soil landslide disasters, characterized in that: The process includes identifying landslide risk areas and identifying houses threatened by landslides within the area. The identification of houses threatened by landslides in the area uses high-precision digital elevation models (DEMs) and house vector data, and implements automated identification through geographic information system (GIS) software to identify potentially threatened houses and delineate potential risk areas. The process includes the following steps: S1. Obtain high-precision DEM data and prepare housing vector data; S2. Generate slope, elevation, and aspect layers; S3. Mark the turning points of the houses, calculate the height difference and slope, and determine whether the slope cut threatens the house; S4. Generate extension lines of the house boundaries, analyze the relationship between the extension lines and the slope and aspect, and identify clusters of shallow soil landslides threatening the house. S5. Extract eligible houses as threatened house files and output them for subsequent on-site investigation and risk zone delineation.

2. The method according to claim 1, characterized in that The identification of houses threatened by landslides in the area is carried out by generating slope and height difference layers from high-precision DEM, and extracting the inflection points of the house vector files. The height difference and slope values are used to judge the threat of slope cutting and building houses, and the inflection points of houses with a height difference greater than 10 meters and a slope greater than 8 degrees are marked; on this basis, potentially threatened houses are identified.

3. The method according to claim 1, characterized in that The identification of houses threatened by landslides in the area is carried out by generating outward extension lines of the house boundaries and extracting the slope, slope direction and height difference information of each extension line. The dangerous lines with landslide threats are identified by combining the distance between the slope foot and the house, the direction angle of the extension line and the slope conditions. When a house has one or more dangerous lines, the house is determined to be a threatened house and the corresponding threatened house vector surface data is generated.

4. The method according to claim 1, wherein The landslide risk area identification includes the following steps: S1. Obtain information on regional topography, geology, slope structure, vegetation distribution, etc. through remote sensing technology and regional geological data; S2. Analyze vegetation types, distribution, size, and age using optical image analysis techniques to assess the anchoring effect of roots of different vegetation types on shallow soils. S3. Obtain local data from the database, including historical landslide data, meteorological data, soil moisture, precipitation, etc.; S4. Integrate and analyze the data from S1 to S3, use the risk assessment model to comprehensively assess the risk level of landslide disasters, and generate landslide risk areas and risk levels.

5. The method according to claim 4, characterized in that The remote sensing technology in step S1 includes visible light remote sensing images, infrared remote sensing images, lidar technology, and synthetic aperture radar technology, which are used to obtain high-resolution terrain, soil layer, and vegetation distribution information of the region; radar technology includes ground radar, airborne radar, or satellite radar, which is used to detect soil and rock characteristics; Step S2 uses multispectral or hyperspectral images to evaluate the anchoring effect of vegetation roots on shallow soil, and calculates vegetation coverage and root strength through image classification and analysis technology; Step S4 includes the fusion of remote sensing images, radar data and optical images, and combines geographic information system technology and machine learning methods to use local relevant data in the database for model training to automatically evaluate the landslide disaster risk index of the region.

6. The method according to claim 4, characterized in that The soil stability and landslide susceptibility are analyzed through terrain data, and the soil slope, slope aspect and soil moisture factors are calculated. The landslide risk heat map is generated by combining historical landslide event data and meteorological data to evaluate the landslide susceptibility of each area.

7. The method according to claim 4, characterized in that The risk assessment model is based on machine learning methods such as random forest, support vector machine or convolutional neural network. It is trained with historical landslide data and real-time data, and combined with meteorological data, geological data and historical landslide records in the database to predict the probability of landslide occurrence.

8. The method according to claim 4, characterized in that The landslide risk assessment results are a probability distribution map or heat map of landslide occurrence, and output landslide risk levels, which include low risk, medium risk, high risk, and extremely high risk. The risk assessment results are output after comprehensive analysis of historical landslide data, precipitation data, meteorological forecast data, etc.

Images