A surface deformation monitoring system and method for reservoir safety monitoring

Abstract

The present application relates to the technical field of hydraulic engineering safety monitoring, and particularly relates to a surface deformation monitoring system and method for reservoir safety monitoring, comprising: configuring multiple types of monitoring units, the multiple types of monitoring units comprising an artificial observation unit, a global navigation satellite system (GNSS) automatic monitoring unit and a machine vision measurement unit; according to the characteristics of the geographical environment of the reservoir, the dam body characteristics, the risk level and the annual budget, at least two monitoring units are intelligently selected and combined to form a hierarchical monitoring mode. The present application configures multiple types of monitoring units including artificial observation, GNSS automatic monitoring and machine vision measurement, and intelligently selects at least two units to form a hierarchical monitoring mode in combination with the geographical environment of the reservoir, the dam body characteristics, the risk level and the budget, which not only solves the problems of low frequency, poor timeliness and dependence on manpower of artificial observation, but also reduces the high cost of GNSS alone, and is suitable for different scenarios such as small reservoirs.

Description

Technical Field

[0001] This invention relates to the field of water conservancy project safety monitoring technology, specifically to a surface deformation monitoring system and method for reservoir safety monitoring. Background Technology

[0002] As crucial water conservancy infrastructure, the structural safety of reservoir dams directly impacts the safety of life and property in downstream areas and the stable operation of the regional economy and society. During long-term operation, reservoirs are affected by various factors such as geological activity, water level changes, temperature effects, and material aging, causing deformations such as displacement and settlement on their surface structure. These deformations are often the most direct early manifestations of potential safety hazards within the dam body. Therefore, continuous, accurate, and timely monitoring of reservoir surface deformation is a key technical link in assessing the overall safety status of the dam and providing early warning of potential risks, and has become an indispensable foundation for modern reservoir safety management. Currently, a comprehensive reservoir safety monitoring system typically includes multiple dimensions such as deformation monitoring, seepage monitoring, environmental monitoring, and video surveillance. Among these, surface deformation monitoring, as the most direct indicator of dam structural stability, is particularly important.

[0003] Existing technologies for monitoring surface deformation in reservoirs mainly include manual observation, automatic monitoring via GPS, and machine vision measurement. Manual observation relies on specialized technicians periodically using precision instruments to measure at designated work points on-site. While the accuracy meets requirements, it suffers from low monitoring frequency, poor timeliness, high manpower requirements, and difficulty in implementation under adverse weather conditions. Automatic monitoring via GPS enables continuous automatic observation and remote data transmission through the establishment of reference stations and monitoring stations, improving data timeliness. However, the high cost of system construction and maintenance, the requirement for a suitable signal reception environment, and the need for stable foundation support for the reference stations limit its economical and effective application in large-scale small reservoir clusters. Machine vision measurement, as a new technology, offers advantages of non-contact and automation, but its accuracy and reliability are severely affected by environmental factors such as weather changes, lighting conditions, and visibility. Especially considering the large number of small reservoirs in my country, the limited investment per reservoir, and the diverse geographical environments, there is an urgent need to develop a surface deformation monitoring solution that can reduce construction and operation costs while ensuring monitoring accuracy and timeliness, and is highly adaptable to the environment and easy to deploy on a large scale. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a surface deformation monitoring system and method for reservoir safety monitoring. It solves the problems of poor timeliness of manual observation, high cost of GNSS, susceptibility of machine vision to environmental interference, and difficulty in balancing monitoring accuracy, timeliness, and cost compared to existing technologies.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a surface deformation monitoring method for reservoir safety monitoring, comprising: It is equipped with multiple types of monitoring units, including manual observation units, global navigation satellite system (GNSS) automatic monitoring units, and machine vision measurement units; Based on the reservoir's geographical environment, dam characteristics, risk level, and annual budget, at least two monitoring units are intelligently selected and combined to form a hierarchical monitoring mode to monitor the displacement, settlement, and other deformations on the reservoir dam surface and obtain multi-source raw deformation data. Establish a data fusion and intelligent analysis platform to receive and preprocess multi-source raw deformation data; By using intelligent processing algorithms for multi-source sensor data, the preprocessed multi-source raw deformation data is fused, analyzed, and cross-validated to obtain fused deformation data. Build a unified data transmission network to transmit the merged and transformed data to a remote centralized management platform; The remote centralized management platform aggregates, centrally stores, analyzes historical trends, and provides intelligent early warnings for fused deformation data in real time, and links with the video surveillance system to provide early warning responses.

[0006] Furthermore, the monitoring units are intelligently selected and combined, including: Obtain information on the reservoir's size, dam type, geological conditions, risk level, annual budget, and real-time environmental status; Based on the scale, dam type, geological conditions, risk level, annual budget, and real-time environmental status information, evaluate the combined benefits of monitoring modes including manual observation units, GNSS automatic monitoring units, and machine vision measurement units. Based on the combined benefits, monitoring units are intelligently selected and dynamically combined to form a customized hierarchical monitoring mode.

[0007] Furthermore, the multi-source raw deformation data is preprocessed, including: Digital input and data format conversion of manually observed deformation data; Data calibration, denoising, and spatiotemporal benchmark unification are performed on GNSS displacement data; Perform image distortion correction, feature point matching, and initial displacement calculation on machine vision offset data.

[0008] Furthermore, the preprocessed multi-source raw deformation data are fused, analyzed, and cross-validated, including: By comparing GNSS displacement data and machine vision offset data, potential errors caused by environmental interference in the machine vision offset data can be identified. Using GNSS displacement data as a calibration benchmark, error correction is performed on machine vision offset data; The weighted fusion calculation is performed on the deformation data observed by humans, the displacement data of GNSS, and the corrected machine vision offset data to obtain the fused deformation data.

[0009] Furthermore, the remote centralized management platform analyzes and provides intelligent early warnings for the fused deformation data, including: Establish a real-time visualization interface for fused deformation data; Construct a long-term deformation trend model and a short-term abnormal fluctuation model for the surface of a reservoir dam; Based on the model and combined with preset risk assessment rules, potential safety hazards are intelligently identified and risk levels are classified. Based on the risk level, multi-level early warning information is sent when the fused deformation data exceeds the preset safety threshold.

[0010] Furthermore, the machine vision measurement unit includes: Industrial-grade cameras are used to collect image data of the surface of reservoir dams. Visual targets are placed on the surface of the reservoir dam as reference points for image feature matching; Edge computing units, configured near industrial-grade cameras, are used for preliminary coordinate transformation and anomaly screening of image data; Optimized low-light and rain / fog adaptation image algorithms are run in the edge computing unit.

[0011] Furthermore, the machine vision measurement unit performs adaptive image feature matching based on the surface features of the dam, including: Identify the texture features and visual targets on the dam surface; An adaptive image feature matching algorithm is used to calculate the initial offset of texture features and visual target relative to a preset reference position; When the GNSS automatic monitoring unit is available, the initial offset is registered and fused using GNSS displacement data to calculate the accurate coordinate change.

[0012] The present invention also provides a surface deformation monitoring system for reservoir safety monitoring, applied to the reservoir dam surface deformation monitoring method described in any one of the above-mentioned methods, comprising: The monitoring unit includes a manual observation unit, a Global Navigation Satellite System (GNSS) automatic monitoring unit, and a machine vision measurement unit, which are used to acquire multi-source raw deformation data; The mode selection and combination module is used to intelligently select and combine monitoring units based on the reservoir's geographical environment, dam characteristics, and safety risk level to form a hierarchical monitoring mode. The data fusion and analysis platform is used to receive and preprocess multi-source raw deformation data, and to perform fusion analysis and cross-validation on the preprocessed data to obtain fused deformation data. The data transmission network is used to transmit fused deformation data to the centralized management and early warning platform; The centralized management and early warning platform is used to collect, store, analyze, and intelligently warn about fused and deformed data, and to link with the video surveillance system for early warning response.

[0013] Furthermore, the data fusion and analysis platform includes: The data receiving module is used to receive multi-source raw deformation data from the monitoring unit; The data preprocessing module is used to digitize manually observed deformation data, calibrate GNSS displacement data, and correct image distortion of machine vision offset data. The fusion analysis module is used to perform fusion analysis and cross-validation on preprocessed data using intelligent processing algorithms for multi-source sensor data, including position registration and fusion calculation of machine vision offset data based on GNSS displacement data.

[0014] Furthermore, the centralized management and early warning platform includes: The data aggregation and storage module is used for real-time aggregation and centralized storage of fused and deformed data; The data analysis module is used to perform historical trend analysis on the fused deformation data and to build deformation trend models and abnormal fluctuation models. The intelligent early warning module is used to intelligently identify potential safety hazards and classify risk levels based on the model and preset risk assessment rules, and to trigger early warning information when the fused deformation data exceeds the preset safety threshold. The video linkage module is used to link with the video surveillance system for targeted image or video evidence collection.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention utilizes multiple monitoring units, including manual observation, GNSS automatic monitoring, and machine vision measurement. It intelligently selects at least two units based on the reservoir's geographical environment, dam characteristics, risk level, and budget to form a hierarchical monitoring mode. This addresses the problems of low frequency, poor timeliness, and reliance on manpower in manual observation, while also reducing the high cost of using GNSS alone. It is adaptable to different scenarios, such as small reservoirs. By preprocessing multi-source raw data to eliminate errors, and then calibrating machine vision data using GNSS data as a benchmark, the three types of data are weighted and fused to improve monitoring accuracy. Simultaneously, edge computing and low-light rain / fog adaptation algorithms of the machine vision units address their susceptibility to environmental interference. A remote centralized management platform enables real-time data aggregation, visualization, trend analysis, and multi-level intelligent early warning. It also links with the video surveillance system to respond to early warnings, ensuring timely and effective alerts. Overall, it balances monitoring accuracy, timeliness, and cost, facilitating large-scale deployment and meeting the safety monitoring needs of different reservoirs. Attached Figure Description

[0016] Figure 1 This is a flowchart of the intelligent combination decision-making process for the monitoring unit of the present invention; Figure 2 This is a flowchart of the multi-source data acquisition and edge preprocessing process of the present invention; Figure 3 This is a flowchart of the multi-source data fusion and cross-validation algorithm of the present invention; Figure 4 This is a flowchart of the deformation early warning and linkage response closed loop of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figures 1-4 This invention provides a surface deformation monitoring method for reservoir safety monitoring, comprising: It is equipped with multiple types of monitoring units, including manual observation units, global navigation satellite system (GNSS) automatic monitoring units, and machine vision measurement units; Based on the reservoir's geographical environment, dam characteristics, risk level, and annual budget, at least two monitoring units are intelligently selected and combined to form a hierarchical monitoring mode to monitor the displacement, settlement, and other deformations on the reservoir dam surface and obtain multi-source raw deformation data. Establish a data fusion and intelligent analysis platform to receive and preprocess multi-source raw deformation data; By using intelligent processing algorithms for multi-source sensor data, the preprocessed multi-source raw deformation data is fused, analyzed, and cross-validated to obtain fused deformation data. Build a unified data transmission network to transmit the merged and transformed data to a remote centralized management platform; The remote centralized management platform aggregates, centrally stores, analyzes historical trends, and provides intelligent early warnings for fused deformation data in real time, and links with the video surveillance system to provide early warning responses.

[0019] Specifically, multiple types of monitoring units are configured. The manual observation unit uses precision instruments such as total stations and levels. The Global Navigation Satellite System (GNSS) automatic monitoring unit includes GNSS receivers, reference stations, and monitoring stations. The reference stations are located in stable bedrock areas around the dam, and the monitoring stations are arranged at key locations on the dam crest and abutments at intervals of 50-100 meters. The machine vision measurement unit consists of industrial-grade cameras, visual targets, and edge computing devices. The cameras are mounted on supports at high altitudes near the dam, and the targets are fixed in easily deformable areas of the dam surface.

[0020] Next, based on the characteristics of the reservoir, intelligent selection of combined monitoring units is made. For example, for a small reservoir with a limited budget and located in a foggy mountainous area, a hierarchical monitoring mode of "machine vision measurement unit + quarterly manual observation unit" is selected. The machine vision unit realizes continuous daily monitoring, and the manual observation unit calibrates the data every quarter. For a large, high-risk reservoir, a mode of "GNSS automatic monitoring unit + machine vision measurement unit + monthly manual observation unit" is adopted. GNSS ensures high-precision continuous monitoring, machine vision supplements coverage of areas with weak GNSS signals, and manual observation periodically verifies the reliability of the data, thereby obtaining multi-source raw deformation data.

[0021] Subsequently, a data fusion and intelligent analysis platform was established. The platform is equipped with a data preprocessing module, which first digitizes the manually observed data, converts paper records into electronic spreadsheet format and standardizes data units; it uses differential correction method to denoise GNSS displacement data and eliminates satellite signal interference; it performs image distortion correction on machine vision offset data and then extracts target displacement information through feature point matching algorithm.

[0022] The data is fused and analyzed using intelligent processing algorithms from multiple sensor sources, employing a weighted fusion formula: ,in To integrate deformation data (unit: millimeters). , , Deformation data observed manually GNSS displacement data Corrected machine vision offset data weights ( The settings are dynamically adjusted based on the accuracy of each data point; for example, when the accuracy of GNSS data is high... (Values ​​range from 0.4 to 0.6), and , , All data are normalized to the millimeter level to ensure dimensional consistency. The formula is used to calculate and combine cross-validation to identify and correct errors caused by fog in machine vision data, resulting in fused deformed data.

[0023] A unified data transmission network is constructed, employing a combination of 4G / 5G wireless networks and fiber optics. Edge computing devices transmit fused deformation data to a remote centralized management platform in real time. The remote platform houses a data storage server and analysis module, aggregating data and storing it in a database. A trend analysis module generates deformation curves and builds a risk assessment model. For example, when the daily change in fused deformation data exceeds a preset threshold, an early warning is automatically triggered. Simultaneously, the system is linked to the reservoir's video monitoring system, retrieving real-time video footage of the deformation area for staff to visually assess the situation. This approach solves the problems of high cost and poor environmental adaptability associated with single monitoring methods, balancing accuracy and timeliness.

[0024] In this embodiment, the intelligent selection and combination of monitoring units includes: Obtain information on the reservoir's size, dam type, geological conditions, risk level, annual budget, and real-time environmental status; Based on the scale, dam type, geological conditions, risk level, annual budget, and real-time environmental status information, evaluate the combined benefits of monitoring modes including manual observation units, GNSS automatic monitoring units, and machine vision measurement units. Based on the combined benefits, monitoring units are intelligently selected and dynamically combined to form a customized hierarchical monitoring mode.

[0025] Specifically, the process begins by obtaining reservoir characteristic information through the reservoir management system and on-site surveys. For example, regarding reservoir size, the total reservoir capacity, dam length, and dam height are determined; regarding dam type, earth-rock dams and concrete dams are distinguished; regarding geological conditions, the lithology of the dam foundation and the presence of faults in the surrounding area are recorded; the risk level is determined as low, medium, or high based on the assessment standards of the water resources department; the annual budget is used to calculate the procurement, installation, and operation and maintenance costs of monitoring equipment; and real-time environmental conditions are obtained through meteorological stations, including data on rainfall, visibility, and temperature.

[0026] Further evaluation of the combined benefits of monitoring units is needed. For example, for a concrete dam reservoir with a medium risk level and a medium budget, and a real-time environment that is mostly cloudy and rainy, the evaluation found that the GNSS unit has stable signal in cloudy and rainy weather but is expensive. The machine vision unit is affected by cloudy and rainy weather and requires additional algorithm optimization. The manual observation unit has high accuracy but low frequency. The combination of "GNSS unit + optimized machine vision unit + bi-monthly manual observation unit" is the most effective, ensuring both accuracy and cost control.

[0027] Finally, based on the combined benefits, the monitoring units are dynamically combined to form a customized hierarchical monitoring model. For example, during the rainy season, the image defogging processing of the machine vision unit is strengthened to improve data reliability; during the dry season, the risk of reservoir deformation is low, and the frequency of manual observation can be reduced from once every two months to once every quarter to achieve a reasonable allocation of monitoring resources.

[0028] In this embodiment, preprocessing the multi-source raw deformation data includes: Digital input and data format conversion of manually observed deformation data; Data calibration, denoising, and spatiotemporal benchmark unification are performed on GNSS displacement data; Perform image distortion correction, feature point matching, and initial displacement calculation on machine vision offset data.

[0029] Specifically, when preprocessing deformation data from manual observations, staff use dedicated data entry software to input angle and distance data measured by the total station and elevation data measured by the level into the system one by one. The system automatically converts the data format to the standard CSV format and verifies the data integrity. If any missing values ​​are found, the staff are contacted to supplement the records.

[0030] For GNSS displacement data preprocessing, differential positioning technology is used for data calibration to eliminate the effects of satellite orbit errors and atmospheric delays; noise is then removed using a Kalman filter algorithm, the formula of which is... ,in The data is the GNSS displacement data (in millimeters) after calibration at time k. Predict displacement data (in millimeters) for time k-1. For Kalman gain, The original GNSS observation data (in millimeters) at time k. The observation matrix is ​​set up, and all parameters are normalized to the millimeter level. Then, a unified spatiotemporal reference is established, and the data from each monitoring station are converted into the same coordinate system and time standard.

[0031] For machine vision offset data preprocessing, Zhang Zhengyou calibration method is used to correct image distortion and eliminate the influence of camera lens distortion; SIFT algorithm is used for feature point matching to identify the position of the visual target in different images; based on the actual size of the target and the pixel ratio of the image, the initial displacement data is calculated. For example, if the actual side length of the target is 10 cm, which corresponds to 200 pixels in the image, and the target is offset by 50 pixels in two frames, the initial displacement is calculated to be 2.5 cm.

[0032] In this embodiment, the preprocessed multi-source original deformation data undergoes fusion analysis and cross-validation, including: By comparing GNSS displacement data and machine vision offset data, potential errors caused by environmental interference in the machine vision offset data can be identified. Using GNSS displacement data as a calibration benchmark, error correction is performed on machine vision offset data; The weighted fusion calculation is performed on the deformation data observed by humans, the displacement data of GNSS, and the corrected machine vision offset data to obtain the fused deformation data.

[0033] Specifically, the first step is to compare GNSS displacement data with machine vision offset data. For example, on a sunny day, machine vision data is less affected by lighting, and the deviation from GNSS data is usually within 1-2 millimeters. However, in foggy weather, the deviation between machine vision data and GNSS data exceeds 5 millimeters, indicating that there is a potential error in the machine vision data caused by environmental interference.

[0034] Next, using GNSS displacement data as the calibration benchmark, error correction is applied to the machine vision offset data using a linear correction formula: Correction Among them, the correction The corrected machine vision offset data (in millimeters). This is the raw machine vision offset data (in millimeters). , The correction coefficients were obtained by fitting the relationship between GNSS data and machine vision data using the least squares method, and Both GNSS data and GNSS data are normalized to the millimeter level to ensure dimensional consistency.

[0035] Finally, a weighted fusion calculation is performed, with weights determined based on the accuracy of each data point. For example, when GNSS data has the highest accuracy (error ≤ 2 mm), the weight is... Take 0.5; the accuracy of the corrected machine vision data is second best (error ≤ 3 mm), weight... Take 0.3; manual observation data has high accuracy but low frequency, weight Take 0.2 and substitute it into the weighted fusion formula for correction. (All data are at the millimeter level), resulting in fused deformation data. This process improves data reliability and reduces the impact of single data errors.

[0036] In this embodiment, the remote centralized management platform analyzes and provides intelligent early warning of the fused deformation data, including: Establish a real-time visualization interface for fused deformation data; Construct a long-term deformation trend model and a short-term abnormal fluctuation model for the surface of a reservoir dam; Based on the model and combined with preset risk assessment rules, potential safety hazards are intelligently identified and risk levels are classified. Based on the risk level, multi-level early warning information is sent when the fused deformation data exceeds the preset safety threshold.

[0037] Specifically, the remote centralized management platform establishes a real-time visual display interface. The interface is based on a 3D model of the reservoir, marking the location of each monitoring point and using different colors to indicate the magnitude of the fused deformation data. For example, green is used for deformation less than 1 mm, yellow for 1-3 mm, and red for greater than 3 mm. Staff can intuitively view the deformation situation in each area.

[0038] When constructing the deformation model, the long-term deformation trend model uses a linear regression algorithm, based on the fused deformation data from the past 1-3 years, to fit the relationship curve between deformation and time. The formula is as follows: ,in The deformation amount is in millimeters. For time (days) Deformation rate (mm / day). The initial deformation amount (mm); the short-term abnormal fluctuation model calculates the standard deviation of the fused deformation data over the past 7 days. When the deviation of a single day's data from the mean exceeds twice the standard deviation, it is judged as an abnormal fluctuation.

[0039] Based on the model and pre-set risk assessment rules, risk levels are classified, such as deformation rate. Less than 0.01 mm / day and without abnormal fluctuations is considered low risk; A value of 0.01-0.03 mm / day or a slight abnormal fluctuation is considered medium risk. A value greater than 0.03 mm / day or significant abnormal fluctuations indicate high risk. When the fused deformation data exceeds the preset safety threshold (e.g., daily deformation greater than 5 mm), the platform automatically sends multi-level early warning information. Low-risk warnings are sent via system message; medium-risk warnings are sent via SMS to reservoir management personnel; and high-risk warnings are simultaneously communicated to management personnel by phone and reported to the superior water resources department.

[0040] In this embodiment, the machine vision measurement unit includes: Industrial-grade cameras are used to collect image data of the surface of reservoir dams. Visual targets are placed on the surface of the reservoir dam as reference points for image feature matching; Edge computing units, configured near industrial-grade cameras, are used for preliminary coordinate transformation and anomaly screening of image data; Optimized low-light and rain / fog adaptation image algorithms are run in the edge computing unit.

[0041] Specifically, the industrial-grade camera in the machine vision measurement unit uses a device with a resolution of 20 million pixels and a frame rate of 15 frames per second. It supports low-light shooting (minimum illumination of 0.01 lux), has a lens focal length of 25 mm, and ensures that the shooting range covers the dam surface area within 50-100 meters. The camera housing adopts an IP67 waterproof rating design to adapt to the humid environment of the reservoir.

[0042] The visual target is designed as a square with a side length of 30 centimeters. The surface features a red and white checkerboard pattern. A strong magnet or expansion screw is installed on the back of the target, which can be firmly fixed to the protective layer of a concrete dam or earth-rock dam. The target is made of aging-resistant ABS plastic, ensuring that it will not fade or deform after long-term use.

[0043] The edge computing unit uses an embedded chip (such as the NVIDIA Jetson Nano) and is installed in a waterproof box near the industrial-grade camera. It is connected to the camera via a network cable to receive image data in real time. The unit runs optimized low-light and rain / fog adaptation image algorithms. The low-light algorithm uses multi-frame noise reduction technology to improve the brightness and clarity of the image by synthesizing multiple low-light images. The rain / fog adaptation algorithm uses a dark channel prior algorithm to remove fog particle interference in the image and enhance the target contour recognition. This unit can complete the initial data processing near the camera, reducing the amount of subsequent data transmission and improving the overall monitoring efficiency.

[0044] In this embodiment, the machine vision measurement unit performs adaptive image feature matching based on the surface features of the dam, including: Identify the texture features and visual targets on the dam surface; An adaptive image feature matching algorithm is used to calculate the initial offset of texture features and visual target relative to a preset reference position; When the GNSS automatic monitoring unit is available, the initial offset is registered and fused using GNSS displacement data to calculate the accurate coordinate change.

[0045] Specifically, the machine vision measurement unit first identifies the surface features of the dam body through image recognition algorithms. In terms of texture features, it identifies areas with obvious identifiability such as cracks, water stains, and concrete pouring joints on the dam body surface. The visual target recognition uses a template matching algorithm, which uses a preset target pattern as a template to search for matching areas in the captured image to determine the target position.

[0046] The initial offset is calculated using an adaptive image feature matching algorithm. The algorithm first extracts texture features and key feature points of the target (such as corner points and edge points), and records the coordinates of the feature points in the reference image. Then compare the coordinates of the same feature point in the current image. Based on the camera calibration parameters (focal length, image distance), using the formula Calculate the initial offset (millimeters), of which This represents the actual distance (in millimeters) from the camera to the dam surface. The camera pixel size (mm / pixel) is used, and all parameters are normalized to the millimeter level to ensure dimensional consistency.

[0047] When the GNSS automatic monitoring unit is available, GNSS displacement data of the same monitoring area can be acquired. (millimeters), the initial offset calculated by machine vision. and Position registration and fusion are performed using precise formula calculations. actuarial science For precise coordinate changes (in millimeters). Weighting coefficients (adjusted based on image sharpness; higher sharpness results in higher weighting coefficients). Use 0.6-0.8 for low resolution. (Take a value of 0.3-0.5). Through this fusion calculation, the accuracy of the machine vision measurement unit is improved, and the impact of environmental factors on the measurement results is reduced.

[0048] This invention also provides a surface deformation monitoring system for reservoir safety monitoring, applicable to the surface deformation monitoring method of reservoir dams described in any of the above-mentioned methods, comprising: The monitoring unit includes a manual observation unit, a Global Navigation Satellite System (GNSS) automatic monitoring unit, and a machine vision measurement unit, which are used to acquire multi-source raw deformation data; The mode selection and combination module is used to intelligently select and combine monitoring units based on the reservoir's geographical environment, dam characteristics, and safety risk level to form a hierarchical monitoring mode. The data fusion and analysis platform is used to receive and preprocess multi-source raw deformation data, and to perform fusion analysis and cross-validation on the preprocessed data to obtain fused deformation data. The data transmission network is used to transmit fused deformation data to the centralized management and early warning platform; The centralized management and early warning platform is used to collect, store, analyze, and intelligently warn about fused and deformed data, and to link with the video surveillance system for early warning response.

[0049] Specifically, the monitoring unit includes two high-precision total stations (accuracy 0.5 seconds) and one electronic level (accuracy 0.1 mm / km) for manual observation; three reference station receivers (supporting BeiDou and GPS dual-mode) and ten monitoring station receivers, with the reference stations located on stable bedrock 5 km away from the dam body and one monitoring station every 80 meters along the dam crest; and a machine vision measurement unit containing five industrial-grade cameras, 20 visual targets, and five edge computing devices, with the cameras installed at four observation points upstream and downstream of the dam body and at an observation point in the middle of the dam crest.

[0050] The mode selection and combination module is equipped with a decision tree algorithm. By inputting reservoir characteristic parameters (scale, dam type, risk level, etc.), it automatically outputs the optimal combination scheme of monitoring units. For example, if the input is "small reservoir, earth-rock dam, low risk, annual budget of 500,000 yuan", the module outputs a combination scheme of "4 machine vision measurement units + 1 manual observation unit per quarter". It also supports manual adjustment of the combination mode.

[0051] The data fusion and analysis platform uses an industrial server (CPU is Intel Xeon E3, memory is 32GB), equipped with data preprocessing and fusion analysis software. It is connected to each monitoring unit via network cable, receives multi-source raw deformation data in real time, and generates fused deformation data after preprocessing and fusion analysis.

[0052] The data transmission network adopts a dual-link design of "fiber optic + 4G". Monitoring units near the dam are connected to the platform through fiber optic cables, while monitoring units in remote areas transmit data through a 4G wireless network, ensuring the stability and real-time performance of data transmission.

[0053] The centralized management and early warning platform is deployed on a cloud server. Staff can access it via a web interface or mobile app. The platform aggregates and merges deformation data in real time and stores it in a MySQL database. The data analysis module generates deformation trend charts. When the intelligent early warning module triggers an early warning, it automatically links with the reservoir video monitoring system to retrieve camera footage from the deformation area and displays it synchronously on the platform interface, allowing staff to keep abreast of the situation on site.

[0054] In this embodiment, the data fusion and parsing platform includes: The data receiving module is used to receive multi-source raw deformation data from the monitoring unit; The data preprocessing module is used to digitize manually observed deformation data, calibrate GNSS displacement data, and correct image distortion of machine vision offset data. The fusion analysis module is used to perform fusion analysis and cross-validation on preprocessed data using intelligent processing algorithms for multi-source sensor data, including position registration and fusion calculation of machine vision offset data based on GNSS displacement data.

[0055] Specifically, the data receiving module of the data fusion and analysis platform adopts the TCP / IP communication protocol and receives data from various monitoring units through port 8080. It sets up a data buffer (capacity 10GB). When the network is interrupted, the buffer temporarily stores the data and automatically retransmits it after the network is restored to ensure that the data is not lost. The module also performs format verification on the received data. If the data format is found to be incorrect (such as missing key parameters or inconsistent units), it automatically sends a retransmission request to the monitoring unit.

[0056] The data preprocessing module processes different types of data separately. For manually observed deformation data, the software's built-in Excel import function is used to digitize the data and convert it to JSON format. For GNSS displacement data, precise single-point positioning technology is used for data calibration to remove errors such as satellite clock bias and ionospheric delay. Then, wavelet transform algorithm is used for noise reduction to eliminate high-frequency interference. For machine vision offset data, camera calibration tools (such as OpenCV calibration functions) are used to correct image distortion, removing radial and tangential distortion. Then, ORB algorithm is used for feature point matching to calculate the initial displacement data.

[0057] The fusion and analysis module is equipped with a multi-source data fusion algorithm. First, it compares GNSS displacement data with machine vision offset data, identifies outliers in the machine vision data (such as data with a deviation exceeding 3 times the standard deviation), and corrects the outliers using GNSS data. Then, it substitutes the manually observed deformation data, the corrected GNSS data, and the corrected machine vision data into a weighted fusion formula to calculate the fused deformation data. The module also supports data cross-validation. For example, if the deviation between the fused deformation data and the manually observed data exceeds 2 millimeters, the fusion calculation is automatically re-performed to ensure data accuracy.

[0058] In this embodiment, the centralized management and early warning platform includes: The data aggregation and storage module is used for real-time aggregation and centralized storage of fused and deformed data; The data analysis module is used to perform historical trend analysis on the fused deformation data and to build deformation trend models and abnormal fluctuation models. The intelligent early warning module is used to intelligently identify potential safety hazards and classify risk levels based on the model and preset risk assessment rules, and to trigger early warning information when the fused deformation data exceeds the preset safety threshold. The video linkage module is used to link with the video surveillance system for targeted image or video evidence collection.

[0059] Specifically, the data aggregation and storage module of the centralized management and early warning platform adopts a distributed database (such as HBase) with a storage capacity of 100TB, supporting the writing and reading of 1,000 data entries per second. The data is classified and stored in the format of "reservoir number-monitoring date-monitoring point number". At the same time, a data backup mechanism is set up to automatically back up the data to a remote server every morning to prevent data loss.

[0060] The data analysis module is implemented using Python programming. Based on the fused deformation data from the past 5 years, it uses the ARIMA time series model to construct a long-term deformation trend model to predict the deformation trend for the next year. The short-term abnormal fluctuation model calculates the mean and standard deviation of the data over the past 30 days. When the data exceeds the range of "mean ± 2 × standard deviation", it is marked as abnormal data. The module also supports multi-dimensional analysis and can generate statistical reports by monitoring point, time period, and deformation type (displacement, settlement).

[0061] The intelligent early warning module presets three levels of safety thresholds. For example, the first level threshold (low risk) is ≤2 mm of daily deformation, the second level threshold (medium risk) is 2 mm < ≤4 mm of daily deformation, and the third level threshold (high risk) is >4 mm of daily deformation. When the fused deformation data exceeds the corresponding threshold, the module automatically generates early warning information, including the reservoir name, monitoring point location, deformation amount, risk level, etc., and sends it to the designated receiving end through the API interface.

[0062] The video linkage module interfaces with the reservoir's existing video surveillance system (such as the Hikvision monitoring platform), pre-establishing the association between monitoring points and cameras. When the intelligent early warning module triggers an early warning, the module automatically sends a command to the video surveillance system to retrieve the camera corresponding to the early warning monitoring point, switch to the real-time view and record video (recording time is 10 minutes). At the same time, a video window pops up on the platform interface, allowing staff to directly view the on-site situation without manual operation.

[0063] In summary, this invention utilizes multiple monitoring units, including manual observation, GNSS automatic monitoring, and machine vision measurement. It intelligently selects at least two units based on the reservoir's geographical environment, dam characteristics, risk level, and budget to form a hierarchical monitoring mode. This addresses the issues of low frequency, poor timeliness, and reliance on manpower in manual observation, while also reducing the high cost of using GNSS alone. It is adaptable to various scenarios, such as small reservoirs. By preprocessing multi-source raw data to eliminate errors, and then calibrating machine vision data using GNSS data as a benchmark, the three types of data are weighted and fused to improve monitoring accuracy. Furthermore, edge computing and low-light rain / fog adaptation algorithms in the machine vision units address their susceptibility to environmental interference. A remote centralized management platform enables real-time data aggregation, visualization, trend analysis, and multi-level intelligent early warning. It also links with the video surveillance system to respond to early warnings, ensuring timely and effective alerts. Overall, the invention balances monitoring accuracy, timeliness, and cost, facilitating large-scale deployment and meeting the safety monitoring needs of different reservoirs.

[0064] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0065] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A surface deformation monitoring method for reservoir safety monitoring, characterized in that, include: It is equipped with multiple types of monitoring units, including manual observation units, GNSS automatic monitoring units, and machine vision measurement units; Based on the reservoir's geographical environment, dam characteristics, risk level, and annual budget, at least two monitoring units are intelligently selected and combined to form a hierarchical monitoring mode to monitor the displacement, settlement, and other deformations on the reservoir dam surface and obtain multi-source raw deformation data. Establish a data fusion and intelligent analysis platform to receive and preprocess multi-source raw deformation data; By using intelligent processing algorithms for multi-source sensor data, the preprocessed multi-source raw deformation data is fused, analyzed, and cross-validated to obtain fused deformation data. Build a unified data transmission network to transmit the merged and transformed data to a remote centralized management platform; The remote centralized management platform aggregates, centrally stores, analyzes historical trends, and provides intelligent early warnings for fused deformation data in real time, and links with the video surveillance system to provide early warning responses.

2. The surface deformation monitoring method for reservoir safety monitoring as described in claim 1, characterized in that, Intelligent selection and combination of monitoring units, including: Obtain information on the reservoir's size, dam type, geological conditions, risk level, annual budget, and real-time environmental status; Based on the scale, dam type, geological conditions, risk level, annual budget, and real-time environmental status information, evaluate the combined benefits of monitoring modes including manual observation units, GNSS automatic monitoring units, and machine vision measurement units. Based on the combined benefits, monitoring units are intelligently selected and dynamically combined to form a customized hierarchical monitoring mode.

3. The surface deformation monitoring method for reservoir safety monitoring as described in claim 1, characterized in that, Preprocessing of multi-source raw deformation data includes: Digital input and data format conversion of manually observed deformation data; Data calibration, denoising, and spatiotemporal benchmark unification are performed on GNSS displacement data; Perform image distortion correction, feature point matching, and initial displacement calculation on machine vision offset data.

4. The surface deformation monitoring method for reservoir safety monitoring as described in claim 1, characterized in that, The preprocessed multi-source raw deformation data were fused, analyzed, and cross-validated, including: By comparing GNSS displacement data and machine vision offset data, potential errors caused by environmental interference in the machine vision offset data can be identified. Using GNSS displacement data as a calibration benchmark, error correction is performed on machine vision offset data; The weighted fusion calculation is performed on the deformation data observed by humans, the displacement data of GNSS, and the corrected machine vision offset data to obtain the fused deformation data.

5. The surface deformation monitoring method for reservoir safety monitoring as described in claim 1, characterized in that, The remote centralized management platform analyzes and provides intelligent early warnings for the fused deformation data, including: Establish a real-time visualization interface for fused deformation data; Construct a long-term deformation trend model and a short-term abnormal fluctuation model for the surface of a reservoir dam; Based on the model and combined with preset risk assessment rules, potential safety hazards are intelligently identified and risk levels are classified. Based on the risk level, multi-level early warning information is sent when the fused deformation data exceeds the preset safety threshold.

6. The surface deformation monitoring method for reservoir safety monitoring as described in claim 1, characterized in that, The machine vision measurement unit includes: Industrial-grade cameras are used to collect image data of the surface of reservoir dams. Visual targets are placed on the surface of the reservoir dam as reference points for image feature matching; Edge computing units, configured near industrial-grade cameras, are used for preliminary coordinate transformation and anomaly screening of image data; Optimized low-light and rain / fog adaptation image algorithms are run in the edge computing unit.

7. A surface deformation monitoring method for reservoir safety monitoring as described in claim 6, characterized in that, The machine vision measurement unit performs adaptive image feature matching based on the surface features of the dam, including: Identify the texture features and visual targets on the dam surface; An adaptive image feature matching algorithm is used to calculate the initial offset of texture features and visual target relative to a preset reference position; When the GNSS automatic monitoring unit is available, the initial offset is registered and fused using GNSS displacement data to calculate the accurate coordinate change.

8. A surface deformation monitoring system for reservoir safety monitoring, applied to the reservoir dam surface deformation monitoring method according to any one of claims 1-7, characterized in that, include: The monitoring unit includes a manual observation unit, a GNSS automatic monitoring unit, and a machine vision measurement unit, which are used to acquire multi-source raw deformation data; The mode selection and combination module is used to intelligently select and combine monitoring units based on the reservoir's geographical environment, dam characteristics, and safety risk level to form a hierarchical monitoring mode. The data fusion and analysis platform is used to receive and preprocess multi-source raw deformation data, and to perform fusion analysis and cross-validation on the preprocessed data to obtain fused deformation data. The data transmission network is used to transmit fused deformation data to the centralized management and early warning platform; The centralized management and early warning platform is used to collect, store, analyze, and intelligently warn about fused and deformed data, and to link with the video surveillance system for early warning response.

9. A surface deformation monitoring system for reservoir safety monitoring as described in claim 8, characterized in that, The data fusion and analysis platform includes: The data receiving module is used to receive multi-source raw deformation data from the monitoring unit; The data preprocessing module is used to digitize manually observed deformation data, calibrate GNSS displacement data, and correct image distortion of machine vision offset data. The fusion analysis module is used to perform fusion analysis and cross-validation on preprocessed data using intelligent processing algorithms for multi-source sensor data, including position registration and fusion calculation of machine vision offset data based on GNSS displacement data.

10. A surface deformation monitoring system for reservoir safety monitoring as described in claim 8, characterized in that, The centralized management and early warning platform includes: The data aggregation and storage module is used for real-time aggregation and centralized storage of fused and deformed data; The data analysis module is used to perform historical trend analysis on the fused deformation data and to build deformation trend models and abnormal fluctuation models. The intelligent early warning module is used to intelligently identify potential safety hazards and classify risk levels based on the model and preset risk assessment rules, and to trigger early warning information when the fused deformation data exceeds the preset safety threshold. The video linkage module is used to link with the video surveillance system for targeted image or video evidence collection.

Images