Slope landslide geological disaster monitoring method and system based on image intelligent identification

Abstract

The invention relates to the technical field of slope disaster monitoring, particularly provides a slope landslide geological disaster monitoring method and system based on image intelligent identification, and solves the problems that image identification and hyperspectral analysis are depended, multi-modal data fusion is lacked, and the monitoring accuracy is high. The method comprises the following steps: acquiring multi-modal data through an unmanned aerial vehicle hyperspectral camera, a ground image sensor, a vibration sensor, a displacement sensor and a meteorological sensor; carrying out denoising, calibration and formatting processing on the acquired multi-modal data; extracting landslide related features from the image data, the hyperspectral data, the vibration data, the displacement data and the meteorological data; fusing the multi-modal data through an intelligent fusion algorithm to generate a landslide risk assessment result; and early warning information is generated according to a landslide risk assessment result, and an emergency response mechanism is triggered, so that the precision and efficiency of image recognition and hyperspectral data analysis are improved, and the real-time performance and response speed of the system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of slope disaster monitoring, and in particular to a slope landslide geological disaster monitoring method and system based on image intelligent recognition. Background Art

[0002] In the field of geological disaster monitoring, slope landslides are a common natural disaster that poses a serious threat to people's lives and property. Traditional monitoring methods rely on manual inspections and topographic surveys, which are not only inefficient but also susceptible to weather and environmental conditions. With the development of intelligent image recognition technology, slope landslide geological disaster monitoring methods based on image recognition have emerged, greatly improving the accuracy and real-time nature of monitoring.

[0003] Intelligent image recognition technology uses algorithms to automatically analyze and interpret information in images, enabling rapid detection of unusual slope changes, such as the appearance of cracks and signs of soil movement. This technology enables 24-hour monitoring, significantly reducing labor costs and the potential for human misjudgment. More importantly, the intelligent recognition system can provide early warning of potential landslide risks based on historical data and learned characteristic patterns, buying valuable time for evacuation and preventative measures.

[0004] Prior art one, Chinese patent, application number 201811264328.9 discloses a geological disaster intelligent group defense monitoring system based on rapid image recognition. The present invention belongs to the field of geotechnical engineering and provides a group defense intelligent slope monitoring and evaluation system, including a server, a mobile terminal, a GPS positioning module, a picture upload module and a data processing module; wherein, the mobile terminal connects the picture upload module and the prediction and warning module via a wireless network, and the system server connects the picture upload module, the data processing module and the prediction and warning module via a wireless network. The monitoring system of the present invention has low cost, a complete system structure, and simple and convenient operation. It can enhance the rapid identification of geological disasters such as landslides and ground deformations, and enable more users to participate in the geological disaster monitoring system. It is very suitable for popular use, is conducive to the collection of massive monitoring information, and realizes intelligent monitoring of the safety situation and dynamic changes of soil deformation, further realizes group defense and group control of geological disasters, and enhances urban public safety functions:

[0005] Prior art two, Chinese patent, application number 202510029079.9 discloses a method for monitoring dam slope landslides using a drone equipped with a hyperspectral camera. The core equipment is a multi-rotor drone equipped with a global positioning system and a hyperspectral camera. This method is mainly based on the monitoring of vegetation changes using hyperspectral imaging technology to identify potential landslide areas. It is mainly suitable for areas in tropical islands where vegetation is less affected by climate. By performing regular flights over the vegetation on the dam slope by the drone, hyperspectral image data is obtained, and the changes in the relevant spectral characteristics of the vegetation and soil are analyzed. Abnormal spectral changes in vegetation are often related to geological instability and can therefore be used as an indicator for landslide early warning. The method of the present invention can perform efficient, rapid, non-contact monitoring of large-area slopes, and has high precision and real-time performance, which helps to improve the accuracy and response speed of landslide disaster early warning.

[0006] The existing technologies currently rely on image recognition and hyperspectral analysis, lack the fusion of multimodal data, and thus lack comprehensiveness in monitoring results. Therefore, the present invention provides a method and system for monitoring landslide geological disasters based on intelligent image recognition. Summary of the Invention

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] In one aspect of the present invention, a method for monitoring landslide geological disasters based on intelligent image recognition is provided, comprising:

[0009] Collect multimodal data through drone hyperspectral cameras, ground image sensors, vibration sensors, displacement sensors, and meteorological sensors;

[0010] De-noising, calibration and formatting of collected multimodal data;

[0011] Extract landslide-related features from image data, hyperspectral data, vibration data, displacement data, and meteorological data;

[0012] The multimodal data is fused through intelligent fusion algorithms to generate landslide risk assessment results;

[0013] Generate early warning information based on the landslide risk assessment results and trigger the emergency response mechanism.

[0014] In an optional embodiment, the multimodal data processing includes using low-pass filtering and median filtering to remove high-frequency noise and impulse noise, identifying and processing outliers through statistical methods, performing radiation calibration, zero point calibration and sensitivity calibration on hyperspectral cameras, vibration sensors and displacement sensors, unifying time stamps for data collected by sensors and equipment, normalizing data from different sensors to the same dimension and range, and storing the preprocessed data in a relational or non-relational database.

[0015] In an optional embodiment, the collected image data is subjected to a wave differential noise reduction process, a database is established based on the collected image data, the image data is segmented and resized, the image data in the database is counted into a data set, and mapping is performed based on the timestamp;

[0016] The denoising formula for image data is as follows;

[0017] g(x)=ω -1 ·ω·f(x,y)·h(u,v),

[0018] Among them, ω·f(x,y) is the wave differential transformation operation, ω -1 is the inverse wave differential transformation operation, and h(u,v) is the filter kernel.

[0019] In an optional embodiment, the extraction of landslide-related features includes identifying cracks and edge features of collapsed areas in the image, analyzing the texture features of the image, identifying possible landslide areas, detecting changes in vegetation cover through color changes, identifying potential landslide areas, performing radiation correction on hyperspectral data, eliminating the influence of the atmosphere and sensors, selecting characteristic bands related to landslides, identifying changes in the composition of vegetation, soil, and rock materials through spectral analysis, assessing landslide risks, analyzing the time series of rainfall, identifying rainfall intensity and duration, assessing the impact of rainfall on landslides, fusing features extracted from images, hyperspectral, vibration, displacement and meteorological data, comprehensively assessing landslide risks, performing weighted fusion according to the importance of each feature, improving the accuracy and comprehensiveness of landslide identification, constructing a landslide risk assessment model based on the fused features, predicting the possibility of landslide occurrence, setting an early warning threshold based on the risk assessment results, and triggering an early warning mechanism.

[0020] In an optional implementation, hyperspectral data is combined with image data to analyze the correlation between vegetation cover and landslide areas, vibration data is combined with displacement data to evaluate the dynamic stability of the slope, meteorological data is combined with landslide risk assessment results, and the warning threshold is dynamically adjusted. The UAV flight path and hyperspectral camera parameters are adjusted according to terrain conditions, the sampling frequency of vibration sensors and displacement sensors is adjusted according to climatic conditions, and the monitoring frequency and warning threshold are dynamically adjusted according to the slope stability status.

[0021] In an optional embodiment, the early warning mechanism includes performing fusion analysis based on multimodal data to assess the landslide risk level of the slope, which is set as follows;

[0022] Low risk, stable slopes, no signs of landslide;

[0023] Medium risk: potential landslide risk exists on the slope and monitoring needs to be strengthened;

[0024] High risk: the slope has obvious signs of landslide and emergency measures must be taken immediately;

[0025] Alarm thresholds for different risk levels are set based on historical data. When the monitoring data exceeds the set threshold, the system automatically triggers an alarm and enters the early warning state. Based on real-time data from drones, ground sensors and meteorological equipment, the system dynamically updates the landslide risk assessment results, adjusts the alarm level and emergency measures, collects feedback information on emergency responses, evaluates the effectiveness of the early warning mechanism, and optimizes the early warning strategy.

[0026] In an optional embodiment, the landslide risk assessment model is constructed by comparing regions in the image with known objects or categories and using a machine learning algorithm to identify actual objects in the image, as shown below:

[0027]

[0028] Where ω is the normal vector of the hyperplane, α i is the Lagrange multiplier, y i is the class label, x i is the feature vector of the sample;

[0029]

[0030] Where b is the intercept of the hyperplane, y i is the category label, is the characteristic vector of the sample time point, x k is the support vector on the boundary, y k is the true label, T corresponds to the time point.

[0031] In an optional embodiment, the generation of early warning information includes setting corresponding alarm thresholds according to the landslide risk level, setting up a multi-level alarm mechanism, generating specific early warning information, providing a real-time visual early warning interface on the monitoring platform, displaying landslide risk areas, alarm levels and emergency recommendations, initiating corresponding emergency plans according to the alarm levels, dynamically updating landslide risk assessment results based on real-time data from drones, ground sensors and meteorological equipment, adjusting alarm levels and emergency measures, collecting feedback information on emergency responses, evaluating the effectiveness of the early warning mechanism, and optimizing early warning strategies. According to changes in terrain, climate and slope conditions, the alarm thresholds and emergency response plans are automatically adjusted to improve the adaptability of the system, all monitoring data, alarm records and emergency response information are stored in a database, and risk analysis reports are generated regularly.

[0032] In an optional embodiment, the early warning strategy setting includes preprocessing hyperspectral data, image data, vibration data, displacement data and meteorological data and evaluating the landslide risk level, which is calculated as follows:

[0033] y0=β0+β1·x1+β2·x2+β3·x3+…+β n ·x n +∈,

[0034] Among them, y0 is the landslide risk score, β n is the regression coefficient, which indicates the influence of each characteristic on landslide risk, ∈ is the regression bias constant, and the regression coefficient β n When the risk score y0 exceeds the set threshold through least squares estimation, an alarm is triggered;

[0035] The landslide risk is divided into three levels: low, medium and high. The support vector is used for division and the calculation is as follows:

[0036] K(x,x′)=exp(-γ·‖xx′‖ 2 ),

[0037] Among them, γ is the kernel parameter, which controls the width of the kernel function.

[0038] Another aspect of the present invention provides a slope and landslide geological disaster monitoring system based on image intelligent recognition, comprising:

[0039] Data acquisition module, including a hyperspectral camera, ground image sensor, vibration sensor, displacement sensor, and meteorological sensor carried by the UAV;

[0040] The data transmission module is used to transmit the collected multimodal data to the edge computing node or cloud server via a wireless network;

[0041] The data processing module includes an image recognition submodule, a hyperspectral data analysis submodule, and a vibration / displacement data processing submodule, which is used to preprocess, extract features, and perform fusion analysis on multimodal data;

[0042] Intelligent fusion algorithm module, used to perform multimodal fusion of image data, hyperspectral data, vibration data, displacement data and meteorological data to generate landslide risk assessment results;

[0043] The early warning module generates early warning information based on the landslide risk assessment results and notifies users via mobile terminals, text messages or voice;

[0044] Adaptive monitoring strategy module that dynamically adjusts monitoring frequency, sensor configuration, and data processing strategy based on terrain, climate, and slope conditions.

[0045] The slope landslide geological disaster monitoring method and system based on image intelligent recognition of the present invention collects multimodal data through unmanned aerial vehicle hyperspectral cameras, ground image sensors, vibration sensors, displacement sensors and meteorological sensors; denoises, calibrates and formats the collected multimodal data; extracts landslide-related features from the image data, hyperspectral data, vibration data, displacement data and meteorological data; fuses the multimodal data through an intelligent fusion algorithm to generate a landslide risk assessment result; generates early warning information according to the landslide risk assessment result, and triggers an emergency response mechanism, thereby improving the accuracy and efficiency of image recognition and hyperspectral data analysis, and improving the real-time performance and response speed of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0047] Figure 1 A flowchart of a method for identifying landslide geological hazards based on image intelligence is provided in Example 1 of the present invention;

[0048] Figure 2 This is a framework diagram of slope landslide geological disaster monitoring based on image intelligent recognition provided in Example 3 of the present invention;

[0049] Figure 3 This is a block diagram of an electronic device provided in Embodiment 7 of the present invention;

[0050] Figure 4 This is a block diagram of the computer-readable storage medium provided in Example 8 of the present invention. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present invention will be described below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0052] In the following, the terms "first," "second," etc., are used for descriptive convenience only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature identified with "first," "second," etc., may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise specified, "plurality" means two or more.

[0053] In the present invention, unless otherwise clearly specified and limited, the term "connection" should be understood in a broad sense. For example, "connection" can be a fixed mechanical connection, a detachable mechanical connection, or an integrated connection; or, "connection" can be a direct connection or an indirect connection through an intermediate medium. In addition, unless otherwise clearly specified and limited, the term "coupling" should be understood in a broad sense. For example, "coupling" can be a direct electrical connection, such as physical contact and electrical conduction between two components, or it can be understood as the electrical connection between different components in a circuit structure through a physical line that can transmit electrical signals, such as printed circuit board (PCB) copper foil or wire, to transmit electrical signals; or, "coupling" can be an indirect electrical connection between two components through an intermediate medium; or, "coupling" can be an electrical connection between two components in an airless / non-contact manner, such as electrical connection between two components using capacitive coupling to transmit electrical signals.

[0054] In an embodiment of the present invention, directional terms such as "up", "down", "left" and "right" may be defined including but not limited to the orientation relative to the schematic placement of the components in the drawings. It should be understood that these directional terms may be relative concepts, which are used for relative description and clarification, and may change accordingly according to changes in the orientation of the components in the drawings.

[0055] Example 1:

[0056] like Figure 1 As shown, an embodiment of the present invention provides a method for monitoring landslide geological disasters based on image intelligent recognition, comprising the following steps:

[0057] Step S1: Collect multimodal data through UAV hyperspectral camera, ground image sensor, vibration sensor, displacement sensor and meteorological sensor;

[0058] Step S2: De-noising, calibration and formatting the collected multimodal data;

[0059] Step S3: extracting landslide-related features from image data, hyperspectral data, vibration data, displacement data, and meteorological data;

[0060] Step S4: fusing the multimodal data through an intelligent fusion algorithm to generate a landslide risk assessment result;

[0061] Step S5: Generate early warning information based on the landslide risk assessment results and trigger the emergency response mechanism.

[0062] In the above-mentioned embodiment, a drone equipped with a hyperspectral camera is used to acquire hyperspectral image data of the slope area, analyze changes in vegetation cover and soil spectral characteristics, and identify potential landslide areas. Based on deep learning algorithms (such as convolutional neural networks (CNN) or the object detection algorithm YOLO), landslide characteristics are identified from ground image data, including the detection of abnormal phenomena such as cracks, displacements, and collapses. This improves the accuracy and efficiency of image recognition and hyperspectral data analysis. Combining multiple data sources such as images, hyperspectral data, and vibration data, a comprehensive landslide monitoring system is constructed, improving the system's real-time performance and response speed, enhancing the drone's endurance and monitoring efficiency, and building an efficient, integrated monitoring and early warning system that supports the collaborative operation of multiple devices.

[0063] Example 2:

[0064] On the basis of Example 1, in step S1 provided in this embodiment of the present invention,

[0065] Step S101: multimodal data processing includes removing high-frequency noise and impulse noise by using low-pass filtering and median filtering;

[0066] Step S102: Identify and process outliers through statistical methods, and perform radiation calibration, zero point calibration, and sensitivity calibration on the hyperspectral camera, vibration sensor, and displacement sensor;

[0067] Step S103: unify the timestamps of the data collected by the sensors and devices, and normalize the data of different sensors to the same dimension and range;

[0068] Step S104: Storing the pre-processed data in a relational or non-relational database

[0069] In the above embodiments, algorithms such as median filtering, bilateral filtering, or wavelet transform are used to remove noise from the image. Radiation correction and spectral smoothing algorithms, such as Savitzky-Golay filtering, are applied to reduce spectral noise. Low-pass filtering and median filtering are used to remove high-frequency noise and impulse noise. Outliers are identified and processed through statistical methods (such as Z-score detection). Radiation calibration, zero point calibration, and sensitivity calibration are performed on hyperspectral cameras, vibration sensors, and displacement sensors. Ensure that the data collected by all sensors and devices have a unified timestamp. Convert data in different formats into a unified format (such as CSV, JSON).

[0070] Normalize data from different sensors to the same dimension and range for subsequent analysis. Use interpolation methods (such as linear interpolation) or machine learning methods to fill in missing data. Segment the data by time or space and mark key areas. Store the preprocessed data in a relational or non-relational database. Save it in a file format to facilitate data sharing and transmission. Establish indexes to improve query efficiency and perform regular backups to ensure data security. Check whether the noise in the data is as expected. Ensure that data from different sensors is consistent. Confirm that the data is not missing or damaged. Assess data quality through statistical methods and data visualization. Through the above steps, the quality and consistency of multimodal data can be effectively improved, providing reliable data support for subsequent landslide monitoring and risk assessment.

[0071] Example 3:

[0072] Based on Example 2, the steps provided in this embodiment of the present invention include performing a wave differential noise reduction process on the collected image data, collecting and establishing a database based on the collected image data, segmenting the image data and unifying the image size, and statistically adding the image data in the database to the data set, and performing a mapping process based on the timestamp.

[0073] The denoising formula for image data is as follows;

[0074] g(x)=ω -1 ·ω·f(x,y)·h(u,v),

[0075] Among them, ω·f(x,y) is the wave differential transformation operation, ω -1 is the inverse wave differential transformation operation, and h(u,v) is the filter kernel.

[0076] In the above embodiment, an original image signal containing noise is acquired. A first- or second-order differentiation operation is performed on the signal, and its rate of change is calculated. The differentiated signal is analyzed to identify the noise component. Noise typically manifests as high-frequency fluctuations, while the signal changes more gradually. The identified noise is suppressed or eliminated through filtering or other methods. The denoised signal is integrated or reconstructed to restore the original signal. Salt and pepper noise or Gaussian noise in the image is eliminated, while image edges and details are preserved. High-frequency noise is effectively identified and suppressed. Important signal features, such as edges and details, are preserved. This method is applicable to various types of signal processing. Differentiation operations may amplify noise, requiring compensation in combination with other techniques. Certain assumptions are made about the type and distribution of noise, and may not be effective in certain situations. The computational complexity is high and may require significant computing resources. The signal is smoothed before differentiation to reduce the amplification effect of noise. Filter parameters are dynamically adjusted based on signal changes to improve the noise reduction effect. Methods such as neural networks are used to learn the characteristics of the signal and noise, achieving more intelligent noise reduction.

[0077] Example 4:

[0078] Based on the embodiments, the steps provided in the embodiments of the present invention for extracting landslide-related features include identifying cracks in images, edge features of collapsed areas, analyzing texture features of images, identifying possible landslide areas, detecting changes in vegetation cover through color changes, identifying potential landslide areas, performing radiation correction on hyperspectral data, eliminating the influence of the atmosphere and sensors, selecting characteristic bands related to landslides, identifying changes in the composition of vegetation, soil, and rock materials through spectral analysis, assessing landslide risks, analyzing the time series of rainfall, identifying rainfall intensity and duration, assessing the impact of rainfall on landslides, fusing features extracted from images, hyperspectral, vibration, displacement and meteorological data, comprehensively assessing landslide risks, performing weighted fusion according to the importance of each feature, improving the accuracy and comprehensiveness of landslide identification, constructing a landslide risk assessment model based on the fused features, predicting the possibility of landslide occurrence, setting an early warning threshold based on the risk assessment results, and triggering an early warning mechanism.

[0079] Combine hyperspectral data with image data to analyze the correlation between vegetation cover and landslide areas, combine vibration data with displacement data to evaluate the dynamic stability of slopes, combine meteorological data with landslide risk assessment results, dynamically adjust the warning threshold, adjust the drone flight path and hyperspectral camera parameters according to terrain conditions, adjust the sampling frequency of vibration sensors and displacement sensors according to climatic conditions, and dynamically adjust the monitoring frequency and warning threshold according to the slope stability status.

[0080] The early warning mechanism includes fusion analysis based on multimodal data to assess the landslide risk level of the slope, which is set as follows;

[0081] Low risk, stable slopes, no signs of landslide;

[0082] Medium risk: potential landslide risk exists on the slope and monitoring needs to be strengthened;

[0083] High risk: the slope has obvious signs of landslide and emergency measures must be taken immediately;

[0084] Alarm thresholds for different risk levels are set based on historical data. When the monitoring data exceeds the set threshold, the system automatically triggers an alarm and enters the early warning state. Based on real-time data from drones, ground sensors and meteorological equipment, the system dynamically updates the landslide risk assessment results, adjusts the alarm level and emergency measures, collects feedback information on emergency responses, evaluates the effectiveness of the early warning mechanism, and optimizes the early warning strategy.

[0085] In an optional embodiment, the landslide risk assessment model is constructed by comparing regions in the image with known objects or categories and using a machine learning algorithm to identify actual objects in the image, as shown below:

[0086]

[0087] Where ω is the normal vector of the hyperplane, α i is the Lagrange multiplier, y i is the class label, x i is the feature vector of the sample;

[0088]

[0089] Where b is the intercept of the hyperplane, y i is the category label, is the characteristic vector of the sample time point, x k is the support vector on the boundary, y k is the true label, T corresponds to the time point

[0090] In the above-mentioned embodiments, drones, hyperspectral cameras, vibration sensors, displacement sensors, and weather stations are used to collect image data, hyperspectral data, vibration data, displacement data, and meteorological data. The collected data is denoised, calibrated, and formatted to ensure accuracy and consistency, laying the foundation for subsequent analysis. Landslide-related features are extracted from the preprocessed data. For example, crack and texture features are extracted from image data; soil moisture and vegetation index are extracted from hyperspectral data; vibration frequency and energy are analyzed from vibration data; displacement rate is detected from displacement data; and rainfall and temperature are analyzed from meteorological data.

[0091] Based on the extracted features, machine learning algorithms (such as random forests, support vector machines) or deep learning methods (such as convolutional neural networks, recurrent neural networks) are used to build a landslide risk assessment model to predict landslide risks. Based on the assessment results, landslide risks are divided into low risk, medium risk, and high risk levels. According to the landslide risk level, the corresponding alarm threshold is set. For example, when the displacement rate exceeds a certain threshold, an alarm is triggered. A multi-level alarm mechanism is set, such as yellow alarm (medium risk), orange alarm (high risk), and red alarm (extremely high risk). Different levels of alarms correspond to different emergency measures. Generate specific warning information, including risk level, affected area, recommended measures, etc.

[0092] Disseminate warning information to relevant users and departments through a variety of means. Send real-time warning information and recommended measures to registered users. Send text messages or voice alerts to key personnel (such as emergency management personnel and local governments). Disseminate warning information through social media platforms. Disseminate warning information through emergency broadcast systems. Provide a real-time visual warning interface on the monitoring platform, displaying landslide risk areas, alert levels, and emergency response recommendations.

[0093] Activate the corresponding emergency plan according to the alert level. For example:

[0094] When the risk is medium, strengthen monitoring, arrange personnel inspections, and prepare emergency supplies.

[0095] When the risk is high, organize personnel evacuation, seal off dangerous areas, and coordinate emergency rescue forces.

[0096] When the risk is extremely high, implement emergency evacuation, initiate rescue operations, and coordinate with fire, medical and other departments.

[0097] Linkage mechanism: Establish a linkage mechanism with the government, emergency management departments, fire departments and medical institutions to ensure that early warning information can be quickly transmitted and responded to.

[0098] On-site disposal: Emergency personnel take corresponding on-site disposal measures based on early warning information and monitoring data, such as reinforcing slopes and clearing obstacles.

[0099] Dynamic Adjustment and Optimization: Based on real-time data from drones, ground sensors, and meteorological equipment, landslide risk assessments are dynamically updated to adjust alert levels and emergency response measures. Feedback from emergency responses is collected to evaluate the effectiveness of the early warning mechanism and optimize the early warning strategy. Alarm thresholds and emergency response plans are automatically adjusted based on changes in terrain, climate, and slope conditions, enhancing system adaptability. All monitoring data, alert records, and emergency response information are stored in a database for subsequent analysis and tracing. Risk analysis reports are regularly generated to summarize landslide monitoring and early warning experience and provide reference for future monitoring efforts. Typical landslide cases are entered into a case library to provide data support for future optimization of the early warning mechanism. Ensure the proper operation of all sensors and equipment, and promptly replace or repair faulty equipment. Regularly update the monitoring system and early warning algorithms to improve system performance and accuracy. Train system operators and emergency management personnel to ensure they are proficient in using the early warning system and taking appropriate emergency measures.

[0100] Generating early warning information and triggering emergency response mechanisms based on landslide risk assessment results is a systematic process involving multiple steps, including data collection, risk assessment, warning generation, information dissemination, and emergency response. Through multimodal data fusion, intelligent analysis, and linkage mechanisms, the reliability and effectiveness of the early warning mechanism can be significantly improved, providing strong support for the prevention and control of landslide geological hazards. Furthermore, dynamic adjustment and optimization of the system, as well as the recording and analysis of historical data, will help continuously improve the accuracy and response capabilities of the early warning mechanism, ensuring the safety of people's lives and property.

[0101] Example 5:

[0102] like Figure 1 As shown, based on Example 4, the generation of warning information in the steps provided in the embodiment of the present invention includes setting corresponding alarm thresholds according to the landslide risk level, setting a multi-level alarm mechanism, generating specific warning information, providing a real-time visual warning interface on the monitoring platform, displaying landslide risk areas, alarm levels and emergency recommendations, initiating corresponding emergency plans according to the alarm levels, dynamically updating landslide risk assessment results based on real-time data from drones, ground sensors and meteorological equipment, adjusting alarm levels and emergency measures, collecting feedback information on emergency responses, evaluating the effectiveness of the warning mechanism, and optimizing the warning strategy. According to changes in terrain, climate and slope conditions, the alarm thresholds and emergency response plans are automatically adjusted to improve the adaptability of the system, all monitoring data, alarm records and emergency response information are stored in a database, and risk analysis reports are generated regularly.

[0103] The early warning strategy setting includes preprocessing hyperspectral data, image data, vibration data, displacement data and meteorological data and evaluating the landslide risk level, which is calculated as follows:

[0104] y0=β0+β1·x1+β2·x2+β3·x3+…+β n ·x n +∈,

[0105] Among them, y0 is the landslide risk score, β n is the regression coefficient, which indicates the influence of each characteristic on landslide risk, ∈ is the regression bias constant, and the regression coefficient β n When the risk score y0 exceeds the set threshold through least squares estimation, an alarm is triggered;

[0106] The landslide risk is divided into three levels: low, medium and high. The support vector is used for division and the calculation is as follows:

[0107] K(x,x′)=exp(-γ·‖xx′‖ 2 ),

[0108] Among them, γ is the kernel parameter, which controls the width of the kernel function.

[0109] In the above embodiment, landslide risk levels are assessed based on preprocessed data. Machine learning models, such as linear regression and logistic regression, are used, as are machine learning models like random forests and support vector machines (SVMs). Deep learning models, such as convolutional neural networks (CNNs) and long short-term memory networks (LSTMs), are used to analyze image data (e.g., cracks and landslide signs).

[0110] Network structure: Consists of convolutional, pooling, and fully connected layers. Different alert thresholds are set based on risk assessment results. Historical data analysis: Thresholds for different risk levels are determined based on historical landslide data. Expert experience: Thresholds are adjusted based on the opinions of geological experts.

[0111] Example:

[0112] Low risk: risk score y<30.

[0113] Medium risk: risk score 30≤y<70.

[0114] High risk: risk score y ≥ 70.

[0115] Alert generation and dissemination, alert levels: Yellow: Medium risk. Orange: High risk. Red: Extremely high risk. Mobile app push notifications. SMS / voice notifications. Emergency broadcast system. Social media platforms.

[0116] Establish an emergency response plan. Low-risk: Strengthen monitoring and arrange inspections. Medium-risk: Prepare emergency supplies and organize evacuation drills. High-risk: Immediately organize personnel evacuation and initiate rescue operations. Dynamically update risk assessment results and alert levels based on new monitoring data. Collect feedback from emergency responses to optimize early warning strategies and model parameters.

[0117] Establishing an early warning strategy is a systematic process involving multiple steps, including data collection, preprocessing, risk assessment, setting alarm thresholds, and initiating emergency responses. Effective landslide risk assessment and early warning can be achieved through the use of appropriate mathematical models (such as linear regression, support vector machines, and convolutional neural networks) and parameter settings. Furthermore, dynamic adjustment and optimization mechanisms can enhance the adaptability and accuracy of the early warning system, ensuring reliable operation in the face of varying geological conditions and climate change, and providing strong support for disaster prevention and control.

[0118] Example 6:

[0119] like Figure 2 As shown, based on Example 5, an embodiment of the present invention provides a slope landslide geological disaster monitoring system based on image intelligent recognition, comprising:

[0120] Data acquisition module, including a hyperspectral camera, ground image sensor, vibration sensor, displacement sensor, and meteorological sensor carried by the UAV;

[0121] The data transmission module is used to transmit the collected multimodal data to the edge computing node or cloud server via a wireless network;

[0122] The data processing module includes an image recognition submodule, a hyperspectral data analysis submodule, and a vibration / displacement data processing submodule, which is used to preprocess, extract features, and perform fusion analysis on multimodal data;

[0123] Intelligent fusion algorithm module, used to perform multimodal fusion of image data, hyperspectral data, vibration data, displacement data and meteorological data to generate landslide risk assessment results;

[0124] The early warning module generates early warning information based on the landslide risk assessment results and notifies users via mobile terminals, text messages or voice;

[0125] Adaptive monitoring strategy module that dynamically adjusts monitoring frequency, sensor configuration, and data processing strategy based on terrain, climate, and slope conditions.

[0126] In the above embodiment, the data acquisition module collects image data from the slope area in real time. A camera, installed in the slope area, regularly captures high-resolution images. Sensors, such as vibration sensors and displacement sensors, assist in collecting environmental data. Network equipment transmits the image data to a data processing center via wired or wireless networks. Data transmission utilizes a stable network connection to ensure real-time data transmission and avoid data loss or delays.

[0127] The data preprocessing module preprocesses the collected image data to improve the accuracy of subsequent recognition. Denoising uses algorithms such as median filtering and Gaussian filtering to remove noise from the image. Image enhancement adjusts the image's brightness, contrast, and saturation to highlight potential landslide signs. Feature extraction extracts landslide-related features such as cracks, displacement, and soil moisture. The output is a high-quality, preprocessed image that facilitates subsequent intelligent recognition.

[0128] The intelligent recognition module automatically identifies landslide signs using image recognition algorithms, such as convolutional neural networks (CNNs), to identify landslide features in images. A machine learning model is trained to accurately identify landslide signs. Feature matching matches pre-processed images with historical landslide images to improve recognition accuracy. The output is the recognition results, including the location, type, and severity of the landslide.

[0129] Based on the identification results, the early warning system module issues an alarm and initiates an emergency response. Alarm generation: When signs of a landslide are identified, detailed alert information is generated, including the location, type, and severity of the landslide. Alarm dissemination: Alert information is sent to relevant personnel via various means (such as text messages, emails, and broadcasts) to ensure timely delivery of information. Emergency response: Activate emergency plans, organize evacuation and rescue operations, and minimize losses caused by the disaster.

[0130] The data storage and analysis module stores historical data to support subsequent analysis and system optimization. The database stores image data, recognition results, alarm information, and emergency response records. Data analysis tools analyze historical data to identify landslide patterns and optimize recognition algorithms and early warning strategies. Data visualization displays historical data and analysis results through charts and maps, facilitating user understanding and decision-making.

[0131] The user interface module provides a user-friendly interface for viewing monitoring data and alarm information. It displays real-time slope images and recognition results, allowing users to intuitively view the current slope status. It also displays current alarm information and emergency response status, allowing users to stay informed of disaster situations. It also displays historical data and analysis results, allowing users to conduct in-depth data analysis and trend forecasting.

[0132] Determine the monitoring area, select the slope areas to be monitored, and assess their geological conditions and landslide risks. Install data acquisition equipment and place cameras and sensors throughout the monitoring area to ensure comprehensive coverage of the slope area. Develop a data preprocessing module and, based on actual needs, develop denoising, enhancement, and feature extraction algorithms to improve image quality. Train an intelligent recognition model, collect historical landslide image data, and train machine learning models such as CNNs to accurately identify landslide signs. Build an early warning system and develop alert generation and dissemination capabilities to ensure timely alerts when landslide signs are identified. Deploy the data storage and analysis module, establish a database to store historical data, and develop data analysis tools to support subsequent optimization. Develop a user interface, designing user-friendly monitoring, alerting, and analysis interfaces to facilitate user operation and data viewing. System integration and testing: Integrate all modules and conduct system testing to ensure that they work together and achieve overall functionality. System launch and maintenance: After launch, perform regular maintenance and updates to ensure stable operation and continuous optimization.

[0133] The installation location and angle of cameras and sensors may affect data collection accuracy. Select appropriate installation locations and angles, and regularly inspect and adjust equipment to ensure accurate data collection. The accuracy and efficiency of image recognition algorithms may affect overall system performance. Select appropriate algorithms and improve their accuracy and efficiency through continuous optimization and training. The real-time performance of the system may affect the timeliness of alerts. Optimize data transmission and processing processes and adopt efficient algorithms and hardware to improve the system's real-time performance. The scalability of the system may affect future expansion and upgrades. Adopt a modular design to ensure system scalability and flexibility, facilitating future upgrades and expansion.

[0134] Through the above system architecture design and implementation steps, a slope and landslide geological disaster monitoring system based on intelligent image recognition can be realized. This system can monitor slope landslide risks in real time and issue timely alerts, providing strong support for disaster prevention and control. In actual application, the system design needs to be continuously optimized and adjusted according to specific needs and conditions to ensure stable operation and efficient performance.

[0135] Example 7

[0136] Figure 3 A block diagram is shown of an exemplary electronic device suitable for implementing embodiments of the present invention.

[0137] The electronic device may include a central processing unit / microprocessor / main control chip, etc. 4; a storage medium 5, coupled to the central processing unit / microprocessor / main control chip, etc. 4, and storing computer executable instructions therein for performing the steps of each method of an embodiment of the present invention when executed by the processor.

[0138] The central processing unit / microprocessor / main control chip 4 may include but is not limited to one or more processors or microprocessors.

[0139] The storage medium 5 may include, but is not limited to, for example, random access memory (RAM), read-only memory (ROM), flash memory, EPROM memory, EEPROM memory, registers, computer storage media (such as hard disk, floppy disk, solid-state drive, removable disk, CD-ROM, DVD-ROM, Blu-ray disc, etc.).

[0140] In addition, the electronic device may also include (but not limited to) a data bus 6, an input / output bus / external bus / device bus 7, a display 8, and input / output devices 9 (eg, keyboard, mouse, speaker, etc.).

[0141] The central processing unit / microprocessor / main control chip etc. 4 can communicate with external devices ( 8 , 9 etc.) via an I / O bus 7 via a wired or wireless network (not shown).

[0142] The storage medium 5 may also store at least one computer executable instruction for executing the various functions and / or method steps in the embodiments described in this technology when run by the central processing unit / microprocessor / main control chip 4.

[0143] In one embodiment, the at least one computer executable instruction may also be compiled into or constitute a software product, wherein one or more computer executable instructions are executed by a processor to perform the various functions and / or method steps in the embodiments described in the present technology.

[0144] Example 8

[0145] Figure 4 A schematic diagram of a computer-readable storage medium according to an embodiment of the present invention is shown.

[0146] like Figure 4 As shown, a non-transitory computer-readable storage medium 11 stores instructions, such as computer-readable instructions 10. When the computer-readable instructions 10 are executed by a processor, the various methods described above can be executed. Non-transitory computer-readable storage media include, but are not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. Non-transitory non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. For example, the non-transitory computer-readable storage medium 11 can be connected to a computing device such as a computer, and then, when the computing device executes the computer-readable instructions 10 stored on the computer-readable storage medium 11, the various methods described above can be performed.

[0147] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interface, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0148] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0149] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0150] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the various embodiments of the method of the present invention through a computer device (which can be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (English full name: Read-Only Memory, English abbreviation: ROM), random access memory (English full name: Random Access Memory, English abbreviation: RAM), magnetic disk or optical disk, and other media that can store program code.

[0151] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for monitoring landslide geological disasters based on image intelligent recognition, characterized in that: The following steps are involved: Collect multimodal data through drone hyperspectral cameras, ground image sensors, vibration sensors, displacement sensors, and meteorological sensors; De-noising, calibration and formatting of collected multimodal data; Extract landslide-related features from image data, hyperspectral data, vibration data, displacement data, and meteorological data; The multimodal data is fused through intelligent fusion algorithms to generate landslide risk assessment results; Generate early warning information based on the landslide risk assessment results and trigger the emergency response mechanism.

2. The method for monitoring landslide geological disasters based on image intelligent recognition according to claim 1, characterized in that: The multimodal data processing includes using low-pass filtering and median filtering to remove high-frequency noise and impulse noise, identifying and processing outliers through statistical methods, performing radiation calibration, zero point calibration and sensitivity calibration on hyperspectral cameras, vibration sensors and displacement sensors, unifying timestamps for data collected by sensors and equipment, normalizing data from different sensors to the same dimension and range, and storing the preprocessed data in a relational or non-relational database.

3. The method for monitoring landslide geological disasters based on image intelligent recognition according to claim 1, characterized in that: Perform fluctuation differential noise reduction on the collected image data, collect and establish a database based on the collected image data, cut the image data and unify the image size, count the image data in the database into the data set, and map it according to the timestamp; The denoising formula for image data is as follows; g(x)=ω -1 ·ω·f(x,y)·h(u,v), Among them, ω·f(x,y) is the wave differential transformation operation, ω -1 is the inverse wave differential transformation operation, and h(u,v) is the filter kernel.

4. The method for monitoring landslide geological disasters based on image intelligent recognition according to claim 1, characterized in that: The extraction of landslide-related features includes identifying cracks and edge features of collapsed areas in images, analyzing texture features of images, identifying possible landslide areas, detecting changes in vegetation cover through color changes, identifying potential landslide areas, performing radiation correction on hyperspectral data, eliminating the influence of the atmosphere and sensors, selecting characteristic bands related to landslides, identifying changes in the composition of vegetation, soil, and rock materials through spectral analysis, assessing landslide risks, analyzing time series of rainfall, identifying rainfall intensity and duration, assessing the impact of rainfall on landslides, fusing features extracted from images, hyperspectral, vibration, displacement and meteorological data, comprehensively assessing landslide risks, performing weighted fusion based on the importance of each feature, improving the accuracy and comprehensiveness of landslide identification, constructing a landslide risk assessment model based on the fused features, predicting the possibility of landslide occurrence, setting an early warning threshold based on the risk assessment results, and triggering an early warning mechanism.

5. The method for monitoring landslide geological disasters based on image intelligent recognition according to claim 1, characterized in that: Combine hyperspectral data with image data to analyze the correlation between vegetation cover and landslide areas, combine vibration data with displacement data to evaluate the dynamic stability of slopes, combine meteorological data with landslide risk assessment results, dynamically adjust the warning threshold, adjust the drone flight path and hyperspectral camera parameters according to terrain conditions, adjust the sampling frequency of vibration sensors and displacement sensors according to climatic conditions, and dynamically adjust the monitoring frequency and warning threshold according to the slope stability status.

6. The method for monitoring landslide geological disasters based on image intelligent recognition according to claim 4, characterized in that: The early warning mechanism includes fusion analysis based on multimodal data to assess the landslide risk level of the slope, which is set as follows; Low risk, stable slopes, no signs of landslide; Medium risk: potential landslide risk exists on the slope and monitoring needs to be strengthened; High risk: the slope has obvious signs of landslide and emergency measures must be taken immediately; Alarm thresholds for different risk levels are set based on historical data. When the monitoring data exceeds the set threshold, the system automatically triggers an alarm and enters the early warning state. Based on real-time data from drones, ground sensors and meteorological equipment, the system dynamically updates the landslide risk assessment results, adjusts the alarm level and emergency measures, collects feedback information on emergency responses, evaluates the effectiveness of the early warning mechanism, and optimizes the early warning strategy.

7. The method for monitoring landslide geological disasters based on image intelligent recognition according to claim 1, characterized in that: The landslide risk assessment model is constructed by comparing the regions in the image with known objects or categories and using a machine learning algorithm to identify the actual objects in the image. The expression is as follows: Where ω is the normal vector of the hyperplane, α i is the Lagrange multiplier, y i is the class label, x i is the feature vector of the sample; Where b is the intercept of the hyperplane, y i is the category label, is the characteristic vector of the sample time point, x k is the support vector on the boundary, y k is the true label, T corresponds to the time point.

8. The method for monitoring landslide geological disasters based on image intelligent recognition according to claim 1, characterized in that: The generation of early warning information includes setting corresponding alarm thresholds according to the landslide risk level, setting up a multi-level alarm mechanism, generating specific early warning information, providing a real-time visual early warning interface on the monitoring platform, displaying landslide risk areas, alarm levels and emergency recommendations, launching corresponding emergency plans according to the alarm levels, dynamically updating landslide risk assessment results based on real-time data from drones, ground sensors and meteorological equipment, adjusting alarm levels and emergency measures, collecting feedback information on emergency responses, evaluating the effectiveness of the early warning mechanism, and optimizing early warning strategies. According to changes in terrain, climate and slope conditions, the alarm thresholds and emergency response plans are automatically adjusted to improve the adaptability of the system, all monitoring data, alarm records and emergency response information are stored in a database, and risk analysis reports are generated regularly.

9. The method for monitoring landslide geological disasters based on image intelligent recognition according to claim 1, characterized in that: The early warning strategy setting includes preprocessing hyperspectral data, image data, vibration data, displacement data and meteorological data and evaluating the landslide risk level, which is calculated as follows: y0=β0+β1·x1+β2·x2+β3·x3+…+β n ·x n +∈, Among them, y0 is the landslide risk score, β n is the regression coefficient, which indicates the influence of each characteristic on landslide risk, ∈ is the regression bias constant, and the regression coefficient β n When the risk score y0 exceeds the set threshold through least squares estimation, an alarm is triggered; The landslide risk is divided into three levels: low, medium and high. The support vector is used for division and the calculation is as follows: K(x,x′)=exp(-γ·‖x-x′‖ 2 ), Among them, γ is the kernel parameter, which controls the width of the kernel function.

10. A slope and landslide geological disaster monitoring system based on image intelligent recognition according to any one of claims 1 to 9, characterized in that: include: Data acquisition module, including a hyperspectral camera, ground image sensor, vibration sensor, displacement sensor, and meteorological sensor carried by the UAV; The data transmission module is used to transmit the collected multimodal data to the edge computing node or cloud server via a wireless network; The data processing module includes an image recognition submodule, a hyperspectral data analysis submodule, and a vibration / displacement data processing submodule, which is used to preprocess, extract features, and perform fusion analysis on multimodal data; Intelligent fusion algorithm module, used to perform multimodal fusion of image data, hyperspectral data, vibration data, displacement data and meteorological data to generate landslide risk assessment results; The early warning module generates early warning information based on the landslide risk assessment results and notifies users via mobile terminals, text messages or voice; Adaptive monitoring strategy module that dynamically adjusts monitoring frequency, sensor configuration, and data processing strategy based on terrain, climate, and slope conditions.

Images