A reservoir surrounding landslide monitoring method and system based on multi-source data
By generating and updating trend curves using multi-source data monitoring methods, the problem of insufficient early warning in landslide monitoring around reservoirs is solved, enabling dynamic and accurate prediction of landslide risks, and making it suitable for safe operation of reservoirs under complex geological conditions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 上海旭宇信息科技有限公司
- Filing Date
- 2026-03-10
- Publication Date
- 2026-06-09
AI Technical Summary
Existing technologies are insufficient in their ability to provide early warnings for landslide monitoring around reservoirs. In particular, they are slow to process time-delayed influencing factors, making it difficult to effectively capture the landslide gestation process. Furthermore, they ignore the inherent correlation between various factors, resulting in severely delayed early warning information and hindering disaster prevention and mitigation.
A monitoring method based on multi-source data is adopted. By acquiring multiple types of parameter information and image information, an initial trend curve is generated. When the risk status exceeds a preset threshold, the trend curve is updated. Combined with backtracking nodes and offset risk periods, future risk prediction information is generated.
It improves the accuracy and timeliness of landslide risk prediction, can sensitively capture early landslide precursors, and can dynamically correct trend curves through multi-dimensional data verification, overcoming the bias of single trend extrapolation.
Smart Images
Figure CN122176896A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of reservoir landslide monitoring technology, specifically relating to a method and system for monitoring landslides around reservoirs based on multi-source data. Background Technology
[0002] Reservoirs are critical infrastructures that integrate functions such as water supply, power generation, flood control, and navigation. The stability of the geological environment surrounding a reservoir is directly related to the overall safety of the dam and the reservoir area. Among these, landslides induced by factors such as rainfall and water level changes are one of the main geological hazards threatening the safe operation of reservoirs. Therefore, real-time and accurate monitoring and early warning of the stability of the slopes surrounding a reservoir are the core links to ensure the safe operation of the reservoir.
[0003] To achieve the aforementioned monitoring objectives, existing technologies typically employ the method of deploying sensors in landslide-prone areas to collect relevant environmental parameters. However, this monitoring method suffers from severely insufficient early warning capabilities, particularly in handling the time lag of influencing factors. Landslide formation is a gradual, cumulative process. For instance, the impact of soil moisture saturation caused by continuous rainfall on slope stability is not instantaneous but exhibits a significant lag effect. Using static threshold judgment logic based on a single parameter cannot effectively capture and model this complex evolutionary relationship, resulting in severely delayed early warning information and hindering effective disaster prevention and mitigation. Furthermore, landslides are often triggered by multiple factors, while existing solutions typically analyze each monitoring parameter in isolation, ignoring their inherent correlations and failing to comprehensively assess the true stability state of the slope.
[0004] In view of this, the industry urgently needs a method and system for monitoring landslides around reservoirs based on multi-source data. Summary of the Invention
[0005] The purpose of this invention is to provide a method and system for monitoring landslides around reservoirs based on multi-source data, so as to solve the problems existing in the prior art.
[0006] To achieve the above objectives, the present invention mainly employs the following technical solution: a method for monitoring landslides around reservoirs based on multi-source data, wherein when a node with a risk status exceeding a preset risk level threshold is detected within a time assessment period, the following steps are executed:
[0007] Based on the offset risk period defined by the node as the offset reference point and the determined backtracking node, update the trend curve used to characterize landslide risk.
[0008] Furthermore, based on the updated trend curve, risk prediction information for multiple future time periods is generated;
[0009] Among them, the node is a high-risk node whose risk status exceeds the preset risk level threshold. The high-risk node is directly marked as the offset reference point, and the offset risk period is defined with the offset reference point as the core.
[0010] For example, prior to the monitoring, the following is also included:
[0011] Acquire multi-source landslide monitoring data, which includes various parameter information and image information; and analyze the various parameter information to generate an initial trend curve.
[0012] For example, the various parameter information includes hydrological factor data, surface deformation factor data, equipment alarm information, previous landslide records, surface temperature data, and soil moisture data;
[0013] The image information includes infrared surveillance images and visible light surveillance images.
[0014] For example, the detection of nodes whose risk status exceeds a preset risk level threshold within the time assessment period includes:
[0015] Each parameter in the multi-type parameter information is compared with the corresponding preset threshold, and the risk status corresponding to each time node is output based on the monitoring trend of each parameter.
[0016] And, determine whether there are any nodes within the time assessment period whose risk status exceeds a preset risk level threshold. For example, the determination of the backtracking node includes:
[0017] A similarity analysis is performed on the image information. If the final image similarity score is lower than the image similarity judgment threshold, the image acquisition time corresponding to that score is marked as a backtracking node.
[0018] Furthermore, the trend change is calculated from the trend curve, and the backtracking node is determined based on the comparison between the trend change and the reference offset.
[0019] For example, the similarity analysis of the image information includes: performing similarity analysis on the image information to generate an initial similarity score;
[0020] The image information is converted into a uniform grayscale distribution format, and the histogram distributions of each image are compared to generate histogram similarity.
[0021] Furthermore, the initial similarity score is weighted and fused with the histogram similarity to generate the final image similarity score.
[0022] For example, the step of updating the trend curve includes: traversing the backtracking nodes;
[0023] Data is populated for the offset risk period corresponding to each backtracking node;
[0024] Furthermore, the data within the offset risk period after data filling is weighted and fused with the trend curve to generate the updated trend curve.
[0025] This application also provides a reservoir perimeter landslide monitoring system based on multi-source data, including:
[0026] The data acquisition module is used to acquire multi-source landslide monitoring data, including various parameter information and image information.
[0027] The risk analysis module is used to generate and maintain a trend curve characterizing landslide risk based on the multi-source landslide monitoring data, and to monitor the trend curve to identify whether there are nodes where the risk status exceeds a preset risk level threshold within the time assessment period.
[0028] In addition, a risk prediction module is used to respond to nodes identified by the risk analysis module whose risk status exceeds a preset risk level threshold, update the trend curve based on the offset risk period defined by the node as the offset reference point and the backtracking node, and output risk prediction information for multiple future periods, wherein the updated trend curve is used for risk analysis in subsequent time assessment cycles.
[0029] Beneficial effects
[0030] This invention can generate an initial trend curve based on multi-source landslide monitoring data. Within the time assessment period, when a node with a risk status exceeding a preset risk level threshold is detected, the offset risk period associated with the node is determined. The trend curve is then updated by combining the backtracked node and the offset risk period to generate risk prediction information. By backtracking and integrating historical high-risk data associated with the offset risk period, dynamic correction of the trend curve is achieved, incorporating abrupt changes in landslide evolution into the update process. This overcomes the bias caused by extrapolating a single trend, thereby improving the accuracy of risk prediction.
[0031] This invention integrates multiple types of parameter information and image information to achieve multi-dimensional data verification. On the one hand, it determines the risk status by comparing parameters with preset thresholds. On the other hand, it identifies nodes with final image similarity scores lower than the image similarity judgment threshold as backtracking nodes by analyzing image information, providing a basis for updating the trend curve. By combining quantitative parameter monitoring with image change analysis, it can sensitively capture early landslide precursors, improving the timeliness of risk identification compared to monitoring methods based on a single data source. Attached Figure Description
[0032] Figure 1This is a flowchart of the method provided in Embodiment 1 of the present invention. Detailed Implementation
[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely for explaining the invention and are not intended to limit the scope of protection of the invention.
[0034] Example 1
[0035] Please see Figure 1 This embodiment provides a method for monitoring landslides around reservoirs based on multi-source data, specifically including the following steps:
[0036] When a node with a risk status exceeding a preset risk level threshold is detected within the time assessment period, the following steps are performed: based on the offset risk period defined by the node as the offset reference point and the determined backtracking node, the trend curve used to characterize landslide risk is updated; and, based on the updated trend curve, risk prediction information for multiple future time periods is generated.
[0037] Specifically, multi-source landslide monitoring data is acquired by retrieving historical monitoring records of the target landslide area from the database and continuously collecting data for the current period using a sensor network deployed on-site. Multi-source landslide monitoring data is a dataset obtained from sensors and records of various sources and types, used to comprehensively assess the landslide status. It specifically includes multiple types of parameter information and image information.
[0038] Multiple parameter information is used to comprehensively assess the stability of landslides from multiple physical dimensions. Specifically, it includes: hydrological factor data, such as reservoir water level, rainfall, and pore water pressure; surface deformation factor data, such as key point displacement and settlement data obtained through global navigation satellite systems or synthetic aperture radar interferometry; equipment alarm information, which records abnormal signals triggered by professional equipment such as stress gauges and inclinometers; previous landslide records, which are historical event data used to identify recurrent landslide patterns; and surface temperature data and soil moisture data.
[0039] Image information provides intuitive on-site condition verification, specifically including: infrared monitoring images, used to monitor surface temperature anomalies at night or under low visibility conditions; and visible light monitoring images, used to capture macroscopic changes such as surface cracks and vegetation destruction.
[0040] Furthermore, after acquiring image information, a validity assessment is performed. Specifically, the acquired images are analyzed to generate a final image similarity score. This score ranges from 0 to 1; a lower score indicates lower similarity between images or poor image quality. A validity assessment threshold is set. If the final image similarity score is lower than this threshold, the image is considered distorted, triggering an instruction to reacquire image information. If the final image similarity score is not lower than the threshold, the image is determined to be a valid image and is confirmed as image data that meets quality standards and can be used for subsequent analysis.
[0041] The final image similarity score is a computational model used to quantitatively evaluate the content similarity between two images while suppressing interference from non-surface morphology change factors such as illumination variations. Its specific formula is as follows:
[0042]
[0043] In the formula, The final image similarity score represents the comprehensive similarity score that combines image content similarity and brightness distribution similarity. This represents the initial similarity score, which is a similarity score calculated directly based on the image pixel content. This represents histogram similarity, which is a similarity score calculated based on the gray-level histogram distribution of the image, used to measure the similarity of overall brightness and contrast. This represents the weighting coefficient, which is a preset weighting coefficient used to adjust the relative importance of the initial similarity score and the histogram similarity; where the input is the initial similarity score of two adjacent frames. and the histogram similarity between the two images. The output is the final image similarity score. .
[0044] Each parameter in the acquired multi-parameter information is compared with the corresponding preset threshold, and the risk status corresponding to each time node is output based on the monitoring trend of each parameter. It refers to the qualitative assessment result of the stability of the landslide body at a specific time node based on the monitoring value and trend of multi-parameter information. The risk status includes risk levels of normal, mild risk and severe risk.
[0045] Furthermore, the specific determination method involves extracting reference parameters from a reference comparison sample. The reference comparison sample consists of one or more sets of data collected when the landslide is in a stable or known state, serving as a benchmark for subsequent risk assessment. The reference parameters are extracted from the reference comparison sample and represent the baseline values of various monitoring parameters under stable landslide conditions. The various parameter information at the current time point is compared item by item with these reference parameters, and the parameter difference for each parameter is calculated. This difference is the absolute value of the difference between a parameter value collected at the current time point and its corresponding reference parameter value, used to quantify the degree of parameter deviation. These parameter differences are then compared with their respective difference thresholds. For example, if the difference in displacement parameters exceeds its corresponding difference threshold, or the difference in pore water pressure exceeds its corresponding difference threshold, the risk status at the current time point is determined based on these comparison results.
[0046] Based on a set time assessment period, all risk states within that period are checked. Based on the results of the risk state checks, the corresponding pending data for that time assessment period is output. The time assessment period is a preset and fixed time length within which the landslide risk state is assessed and processed in stages. Specifically, the time assessment period is divided into several time periods. Within each time period, the risk states output in the above steps are checked. If there are time points within a time period where the risk state is abnormal, all data collected within that time period is compiled into pending data.
[0047] The data to be processed output from the above steps are analyzed to determine the changing trend of each parameter over time, so as to generate a trend curve. The trend curve is a comprehensive curve drawn in chronological order based on all multi-source monitoring parameter data within the time assessment period and integrating the characteristic information of the data to be processed. It is used to intuitively show the changing pattern and trend of the comprehensive risk of landslides around the reservoir over time.
[0048] Furthermore, based on this trend curve, backtracking nodes and the corresponding time periods are determined. Backtracking nodes are key time points marked on the trend curve where parameters or image content undergo significant changes; these nodes are the starting point for analyzing the evolution of landslide risk. The determination of backtracking nodes can be achieved through several methods:
[0049] Based on trend curve morphology analysis, a baseline data point and the current data point are selected from the trend curve. The difference between the current data point and the baseline data point is calculated to obtain the trend change, which is used to quantify the magnitude of parameter change over a period of time. This trend change is compared with a preset reference offset. If the trend change exceeds the reference offset, the current data point is identified as a backtracking node. It should be noted that the reference offset is a preset threshold used for comparison with the trend change. When the trend change exceeds the reference offset, it indicates a significant parameter change, and the corresponding data point is identified as a backtracking node.
[0050] Based on image information analysis, similarity analysis is performed on the acquired valid images. Specifically, an initial similarity score is calculated between adjacent images. This initial similarity score is obtained by directly comparing the pixel content of two frames of images using an algorithm. This score may be affected by environmental factors such as lighting. To eliminate the influence of lighting changes, the image information is uniformly converted into a grayscale distribution format, and the histogram distribution of each image is compared to generate a histogram similarity score. This histogram similarity score is obtained by comparing the grayscale histogram distribution of two frames of images and mainly reflects the similarity of the overall brightness and contrast features of the images. The initial similarity score and the histogram similarity score are weighted and fused to generate the final image similarity score. An image similarity judgment threshold is set. If the final image similarity score of two frames of images is lower than the image similarity judgment threshold, it means that the surface morphology has changed significantly. In this case, the changed image is added as a reference comparison sample, and the acquisition time of the image is marked as a backtracking node.
[0051] After identifying all backtracking nodes, it is determined whether any nodes have a risk level exceeding a preset risk level threshold. If the risk level of a backtracking node exceeds this threshold, that node is designated as an offset benchmark point. The preset risk level threshold defines the critical level of risk severity; when the risk level of a backtracking node exceeds this threshold, the node is considered a critical risk point. The offset benchmark point is a backtracking node whose risk level exceeds the preset threshold, signifying that the landslide risk has entered a stage requiring focused analysis, and serves as a benchmark for defining the period of offset risk.
[0052] After determining the offset benchmark point, the offset risk period is determined. Specifically, all backtracking nodes and their corresponding trend changes are extracted. In the trend curve, the risk time points corresponding to these backtracking nodes are determined. The risk time points are the specific time points on the trend curve that correspond to the backtracking nodes. All risk time points are arranged in chronological order to generate a sequence of risk periods to be measured, which is used to analyze the risk evolution process. Taking the offset benchmark point as the core, the trend changes of adjacent backtracking nodes are first analyzed to determine the key period range of risk evolution. Then, based on this range, we trace back to the first backtracking node with a normal risk level, and trace back to the current time point or the next backtracking node with a significantly reduced risk level. The continuous time period that can fully reflect the landslide risk from its occurrence, development to its stage changes is finally defined as the offset risk period.
[0053] The trend curve is updated based on the determined backtracking nodes and the offset risk period, and risk prediction information for multiple future periods is output. The risk prediction information refers to the prediction of the landslide state for multiple future periods based on the corrected trend curve and through an extrapolation algorithm. The specific content includes the risk level and displacement of each future period. Specifically, all backtracking nodes within the offset risk period are traversed. Backtracking nodes are key time points from the occurrence to the development of the offset risk period, and there is an inclusive correspondence between them and the offset risk period. Data is filled into the sub-interval of the offset risk period where each backtracking node is located. Backtracking nodes are divided into global backtracking nodes and local backtracking nodes within the offset risk period. In this embodiment, the backtracking nodes used to update the trend curve specifically refer to the local backtracking nodes within the offset risk period. These nodes have a one-to-one sub-interval association relationship with the offset risk period.
[0054] The data filling process is not simply a matter of preprocessing missing or abnormal data from the original monitoring parameters, but rather a process of supplementing landslide risk characteristics based on multi-source data fusion. Specifically, it uses the original monitoring data within the offset risk period as a basis, combined with hydrological, deformation, and image data from historical high-risk periods, as well as the risk evolution characteristics of adjacent retrospective nodes, to supplement the weak data segments within the offset risk period that do not fully reflect the evolution of landslide risk and fit patterns. The supplemented data is landslide risk trend data, not just the original parameter data. This data has been verified by multiple sources and has the same or even higher reliability as the original trend curve.
[0055] Furthermore, the data within the offset risk period after data filling is weighted and fused with the trend curve to generate a corrected trend curve. This fusion process assigns higher weight to data within periods where the risk level exceeds a preset risk level threshold, enabling the corrected trend curve to more accurately reflect the current activity state of the landslide body. Specifically, risk trend data points within the offset risk period, after risk feature completion, are given higher weight, while the original trend curve data points at other time points have basic weights. The weight allocation is based on landslide risk trend feature data, not the original monitoring parameter data.
[0056] The corrected trend curve is a new curve generated by combining the data from the period of offset risk with the original trend curve through data filling and weighted fusion.
[0057] In the weighted fusion process, the risk contribution weights of parameters such as hydrology, deformation, and soil moisture are preset to calibrate the sub-item weights of the completed trend data, and then merge it with the comprehensive risk data of the original trend curve to generate an updated trend curve. The risk contribution weights of the sub-item parameters can be preset according to the geological characteristics around the reservoir and are consistent with the parameter contribution weights of the initial trend curve.
[0058] The original trend curve data points at other time points serve as the base weights, which preserve the overall evolution trend of landslide risk. The high-weight data points within the offset risk period highlight the core risk evolution characteristics. The weighted fusion ensures that the updated trend curve conforms to the overall monitoring pattern and accurately reflects the evolution characteristics of high-risk periods, thereby improving the accuracy of future risk prediction. Since the risk characteristics of the base weight data points have not changed significantly, there is no need to supplement the risk characteristics, so the base weights of the original trend are maintained.
[0059] Furthermore, the modified trend curve is applied to the monitoring of the next time assessment period, and an extrapolation algorithm is used to predict multiple future time periods, outputting risk prediction information on the risk level and displacement of each future time period, forming a risk prediction report. This report is a structured document that integrates risk prediction information and is the final output of the monitoring and prediction process of this invention.
[0060] In summary, this embodiment sets a time assessment period, identifies data within the time assessment period where the risk status is abnormal as data to be processed, analyzes the data to be processed to generate a trend curve, determines backtracking nodes based on the trend curve, and determines offset benchmark points and corresponding offset risk periods based on backtracking nodes where the risk level exceeds a preset risk level threshold. The trend curve is updated based on the backtracking nodes and offset risk periods, and risk prediction information for multiple future time periods is output.
[0061] Example 2
[0062] This embodiment provides a reservoir perimeter landslide monitoring system based on multi-source data. This system is used to execute the reservoir perimeter landslide monitoring method described above. The system can be divided into the following collaborative modules:
[0063] The data acquisition module is responsible for continuously acquiring multi-source landslide monitoring data from multiple monitoring sources deployed around the reservoir. Specifically, it communicates with various sensors and data systems through standard data interfaces or dedicated protocols. The acquired data can be divided into two categories: The first category is multi-parameter information, which includes, but is not limited to, hydrological factor data acquired through water level gauges and rain gauges, surface deformation factor data acquired through GPS, interferometric radar (InSAR) or crack gauges, equipment alarm information reported by various sensors, previous landslide records in historical geological disaster databases, and surface temperature and soil moisture data acquired through surface sensors; the second category is image information, mainly including visible light monitoring images captured by high-definition cameras deployed at key monitoring points and infrared monitoring images used for observation at night or under special weather conditions. This module performs preliminary formatting and timestamp alignment processing on the acquired raw data for use by subsequent modules.
[0064] The risk analysis module is responsible for generating and continuously maintaining trend curves characterizing landslide risk, and identifying potential risk events in real time. This module's function can be divided into two phases:
[0065] In the initial stage, multiple types of initial parameter information are received. Through time series analysis, multivariate statistical models, or machine learning algorithms, these parameters are comprehensively analyzed and fused to generate a trend curve. This trend curve numerically or graphically represents the comprehensive risk evolution pattern of the monitored area over a period of time.
[0066] During the continuous monitoring phase, risk status determination is performed cyclically in preset time assessment cycles. Specifically, the values of each parameter in the real-time acquired multi-type parameter information are compared with their respective preset thresholds. At the same time, the monitoring trends of each parameter over time (such as accelerated deformation, rapid changes in water level, etc.) are combined to comprehensively determine the risk status corresponding to each time node. Within the current time assessment cycle, a check is performed to determine whether the risk status of any node exceeds the preset risk level threshold. If it does not exceed the threshold, the monitoring continues for the next cycle. If such a high-risk node is detected, the relevant information of the node (such as the time of occurrence, associated parameter data, etc.) is immediately transmitted to the risk prediction module for in-depth processing.
[0067] The risk prediction module is activated when the risk analysis module identifies high-risk nodes. Its core task is to correct historical knowledge and generate more reliable future predictions. Its workflow is as follows:
[0068] The determination of backtracking nodes employs a dual verification mechanism to locate key historical moments that led to the sudden change in the current risk state. On one hand, similarity analysis is performed on image information from a period prior to the occurrence of a high-risk node. Specifically, initial similarity scores are calculated between images. To eliminate interference such as changes in illumination, the image information is uniformly converted to a unified grayscale distribution format, and the histogram distributions of each image are compared to generate histogram similarity. The initial similarity scores and histogram similarity scores are then weighted and fused to obtain the final image similarity score. If this score is lower than a preset image similarity judgment threshold, it indicates that a significant change in the landform has occurred, and this point in time is marked as a backtracking node. On the other hand, the trend change within a specific time period is calculated from the trend curve maintained by the risk analysis module and compared with a reference offset representing normal fluctuations. If the change significantly exceeds the reference offset, this point of change is also determined as a backtracking node.
[0069] Based on the identified backtracking nodes and the offset risk periods associated with high-risk nodes, the trend curve is updated. All identified backtracking nodes are traversed, and data is filled into the offset risk period corresponding to each backtracking node. Data filling is a correction process used to correct or reconstruct data that may contain errors or fail to truly reflect the risk accumulation process in the period based on the results of the backtracking analysis. After data filling is completed, the data in the corrected offset risk period is weighted and fused with the trend curve to generate the corrected trend curve.
[0070] Based on the corrected trend curve, extrapolation calculations are performed using a prediction model (such as a time series prediction model) to generate risk prediction information for multiple future periods. Since this prediction information is based on a trend curve that has been backtracked and corrected, it has higher reliability. At the same time, the corrected trend curve will be fed back to the risk analysis module as a benchmark for risk analysis in subsequent time assessment cycles, thus forming a closed-loop adaptive optimization monitoring and prediction process.
[0071] This embodiment, through the collaborative work of the aforementioned data acquisition module, risk analysis module, and risk prediction module, enables dynamic, closed-loop monitoring and prediction of landslide risks around reservoirs. It can not only identify the current risk status, but more importantly, it can intelligently backtrack historical data and correct the understanding of the risk evolution process, thereby significantly improving the accuracy of prediction. This provides strong technical support for the safe operation of reservoirs and disaster prevention and mitigation work in the surrounding areas, and is suitable for large-scale reservoir projects with complex geological conditions and diverse monitoring data sources.
[0072] The above description is merely a preferred embodiment of this application and is not intended to limit this application. For those skilled in the art, this application can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.
Claims
1. A method for monitoring landslides around reservoirs based on multi-source data, characterized in that, When a node with a risk status exceeding a preset risk level threshold is detected within the time assessment period, the following actions are taken: Based on the offset risk period defined by the node as the offset reference point and the determined backtracking node, update the trend curve used to characterize landslide risk. Furthermore, based on the updated trend curve, risk prediction information for multiple future time periods is generated; Among them, the node is a high-risk node whose risk status exceeds the preset risk level threshold. The high-risk node is directly marked as the offset reference point, and the offset risk period is defined with the offset reference point as the core.
2. The method for monitoring landslides around reservoirs based on multi-source data according to claim 1, characterized in that, Prior to the monitoring, it also includes: Acquire multi-source landslide monitoring data, which includes various parameter information and image information; and analyze the various parameter information to generate an initial trend curve.
3. The method for monitoring landslides around reservoirs based on multi-source data according to claim 2, characterized in that, The various parameter information includes hydrological factor data, surface deformation factor data, equipment alarm information, previous landslide records, surface temperature data, and soil moisture data; The image information includes infrared surveillance images and visible light surveillance images.
4. The method for monitoring landslides around reservoirs based on multi-source data according to claim 3, characterized in that, The nodes whose risk status exceeds the preset risk level threshold within the monitored time assessment period include: Each parameter in the multi-type parameter information is compared with the corresponding preset threshold, and the risk status corresponding to each time node is output based on the monitoring trend of each parameter. In addition, it determines whether there are nodes whose risk status exceeds a preset risk level threshold within the time assessment period.
5. A method for monitoring landslides around reservoirs based on multi-source data according to claim 2, characterized in that, The determination of the backtracking node includes: A similarity analysis is performed on the image information. If the final image similarity score is lower than the image similarity judgment threshold, the image acquisition time corresponding to that score is marked as a backtracking node. Furthermore, the trend change is calculated from the trend curve, and the backtracking node is determined based on the comparison between the trend change and the reference offset.
6. A method for monitoring landslides around reservoirs based on multi-source data according to claim 5, characterized in that, The similarity analysis of the image information includes: The image information is subjected to similarity analysis to generate an initial similarity score; The image information is converted into a uniform grayscale distribution format, and the histogram distributions of each image are compared to generate histogram similarity. Furthermore, the initial similarity score is weighted and fused with the histogram similarity to generate the final image similarity score.
7. The method for monitoring landslides around reservoirs based on multi-source data according to claim 1, characterized in that, The step of updating the trend curve includes: Traverse the backtracked nodes; Data is populated for the offset risk period corresponding to each backtracking node; Furthermore, the data within the offset risk period after data filling is weighted and fused with the trend curve to generate the updated trend curve.
8. A reservoir perimeter landslide monitoring system based on multi-source data, characterized in that, include: The data acquisition module is used to acquire multi-source landslide monitoring data, including various parameter information and image information. The risk analysis module is used to generate and maintain a trend curve characterizing landslide risk based on the multi-source landslide monitoring data, and to monitor the trend curve to identify whether there are nodes where the risk status exceeds a preset risk level threshold within the time assessment period. In addition, a risk prediction module is used to respond to nodes identified by the risk analysis module whose risk status exceeds a preset risk level threshold, update the trend curve based on the offset risk period defined by the node as the offset reference point and the backtracking node, and output risk prediction information for multiple future periods, wherein the updated trend curve is used for risk analysis in subsequent time assessment cycles.